1 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6 * defaults.h (TARGET_COFF): Define.
7 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
9 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
11 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
13 2018-07-31 Alexandre Oliva <oliva@adacore.com>
15 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
16 * gimple-streamer-out.c (output_bb): Save it.
17 * lto-streamer-in.c (input_struct_function_base): Restore
18 instance discriminator if available. Create map on demand.
19 * lto-streamer-out.c (output_struct_function_base): Save it if
21 * final.c (decl_to_instance_map): Document LTO strategy.
23 2018-07-31 Alexandre Oliva <oliva@adacore.com>
24 Olivier Hainque <hainque@adacore.com>
26 * debug.h (decl_to_instance_map_t): New type.
27 (decl_to_instance_map): Declare.
28 (maybe_create_decl_to_instance_map): New inline function.
29 * final.c (bb_discriminator, last_bb_discriminator): New statics,
30 to track basic block discriminators.
31 (final_start_function_1): Initialize them.
32 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
34 (decl_to_instance_map): New variable.
35 (map_decl_to_instance, maybe_set_discriminator): New functions.
36 (notice_source_line): Set discriminator.
38 2018-07-31 Ian Lance Taylor <iant@golang.org>
40 * targhooks.c (default_have_speculation_safe_value): Add
43 2018-07-31 David Malcolm <dmalcolm@redhat.com>
45 * dump-context.h: Include "pretty-print.h".
46 (dump_context::refresh_dumps_are_enabled): New decl.
47 (dump_context::emit_item): New decl.
48 (class dump_context): Add fields "m_test_pp" and
50 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
51 (temp_dump_context::get_dumped_text): New decl.
52 (class temp_dump_context): Add field "m_pp".
53 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
54 (dump_context::refresh_dumps_are_enabled): ...and add a test for
56 (set_dump_file): Update for above change.
57 (set_alt_dump_file): Likewise.
58 (dump_loc): New overload, taking a pretty_printer *.
59 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
60 to any test pretty-printer.
61 (make_item_for_dump_gimple_stmt): New function, adapted from
62 optinfo::add_gimple_stmt.
63 (dump_context::dump_gimple_stmt): Call it, and use the result,
64 eliminating the direct usage of dump_file and alt_dump_file in
65 favor of indirectly using them via emit_item.
66 (make_item_for_dump_gimple_expr): New function, adapted from
67 optinfo::add_gimple_expr.
68 (dump_context::dump_gimple_expr): Call it, and use the result,
69 eliminating the direct usage of dump_file and alt_dump_file in
70 favor of indirectly using them via emit_item.
71 (make_item_for_dump_generic_expr): New function, adapted from
73 (dump_context::dump_generic_expr): Call it, and use the result,
74 eliminating the direct usage of dump_file and alt_dump_file in
75 favor of indirectly using them via emit_item.
76 (make_item_for_dump_printf_va): New function, adapted from
77 optinfo::add_printf_va.
78 (make_item_for_dump_printf): New function.
79 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
80 and use the result, eliminating the direct usage of dump_file and
81 alt_dump_file in favor of indirectly using them via emit_item.
82 (make_item_for_dump_dec): New function.
83 (dump_context::dump_dec): Call it, and use the result,
84 eliminating the direct usage of dump_file and alt_dump_file in
85 favor of indirectly using them via emit_item.
86 (make_item_for_dump_symtab_node): New function, adapted from
87 optinfo::add_symtab_node.
88 (dump_context::dump_symtab_node): Call it, and use the result,
89 eliminating the direct usage of dump_file and alt_dump_file in
90 favor of indirectly using them via emit_item.
91 (dump_context::begin_scope): Reimplement, avoiding direct usage
92 of dump_file and alt_dump_file in favor of indirectly using them
94 (dump_context::emit_item): New member function.
95 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
96 Set up test pretty-printer on the underlying context. Call
97 refresh_dumps_are_enabled.
98 (temp_dump_context::~temp_dump_context): Call
99 refresh_dumps_are_enabled.
100 (temp_dump_context::get_dumped_text): New member function.
101 (selftest::verify_dumped_text): New function.
102 (ASSERT_DUMPED_TEXT_EQ): New macro.
103 (selftest::test_capture_of_dump_calls): Run all tests twice, with
104 and then without optinfo enabled. Add uses of
105 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
106 * dumpfile.h: Update comment for the dump_* API.
107 * optinfo-emit-json.cc
108 (selftest::test_building_json_from_dump_calls): Update for new
109 param for temp_dump_context ctor.
110 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
112 (optinfo_item::~optinfo_item): Likewise.
113 (optinfo::add_item): New member function.
114 (optinfo::emit): Update comment.
115 (optinfo::add_string): Delete.
116 (optinfo::add_printf): Delete.
117 (optinfo::add_printf_va): Delete.
118 (optinfo::add_gimple_stmt): Delete.
119 (optinfo::add_gimple_expr): Delete.
120 (optinfo::add_tree): Delete.
121 (optinfo::add_symtab_node): Delete.
122 (optinfo::add_dec): Delete.
123 * optinfo.h (class dump_context): New forward decl.
124 (optinfo::add_item): New decl.
125 (optinfo::add_string): Delete.
126 (optinfo::add_printf): Delete.
127 (optinfo::add_printf_va): Delete.
128 (optinfo::add_gimple_stmt): Delete.
129 (optinfo::add_gimple_expr): Delete.
130 (optinfo::add_tree): Delete.
131 (optinfo::add_symtab_node): Delete.
132 (optinfo::add_dec): Delete.
133 (optinfo::add_poly_int): Delete.
134 (optinfo_item::optinfo_item): Remove "owned" param.
135 (class optinfo_item): Remove field "m_owned".
137 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
140 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
141 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
142 requested variable alignment.
143 (expand_one_ssa_partition): Likewise.
144 (expand_one_var): Likewise.
146 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
148 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
149 to speculation_safe_value_not_needed.
151 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
153 * targhooks.h (speculation_safe_value_not_needed): New prototype.
154 * targhooks.c (speculation_safe_value_not_needed): New function.
155 * target.def (have_speculation_safe_value): Update documentation.
156 * doc/tm.texi: Regenerated.
158 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
160 * config/aarch64/iterators.md (ALLI_TI): New iterator.
161 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
163 (despeculate_copy<ALLI:mode>_insn): New insn.
164 (despeculate_copyti_insn): New insn.
165 (despeculate_simple<ALLI:mode>): New insn
166 (despeculate_simpleti): New insn.
167 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
169 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
170 aarch64_speculation_safe_value.
171 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
173 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
175 * config/aarch64/aarch64-speculation.cc: New file.
176 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
177 before pass_reorder_blocks.
178 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
180 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
181 X14 and X15 when tracking speculation.
182 * config/aarch64/aarch64.md (register name constants): Add
183 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
184 (unspec): Add UNSPEC_SPECULATION_TRACKER.
185 (speculation_barrier): New insn attribute.
186 (cmp<mode>): Allow SP in comparisons.
187 (speculation_tracker): New insn.
188 (speculation_barrier): Add speculation_barrier attribute.
189 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
190 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
191 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
193 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
195 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
196 aarch64_track_speculation is true.
197 (tb<optab><mode>1): Likewise.
198 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
199 generate CB[N]Z when tracking speculation.
200 (aarch64_split_compare_and_swap): Likewise.
201 (aarch64_split_atomic_op): Likewise.
203 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
205 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
207 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
209 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
210 (speculation_barrier): New insn.
212 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
214 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
215 * config/arm/arm.md (speculation_barrier): New expand.
216 (speculation_barrier_insn): New pattern.
218 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
220 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
221 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
222 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
223 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
225 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
226 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
227 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
228 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
229 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
230 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
231 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
232 * builtins.c (expand_speculation_safe_value): New function.
233 (expand_builtin): Call it.
234 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
235 * doc/extend.texi: Document __builtin_speculation_safe_value.
236 * doc/md.texi: Document "speculation_barrier" pattern.
237 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
238 TARGET_HAVE_SPECULATION_SAFE_VALUE.
239 * doc/tm.texi: Regenerated.
240 * target.def (have_speculation_safe_value, speculation_safe_value): New
242 * targhooks.c (default_have_speculation_safe_value): New function.
243 (default_speculation_safe_value): New function.
244 * targhooks.h (default_have_speculation_safe_value): Add prototype.
245 (default_speculation_safe_value): Add prototype.
247 2018-07-31 David Malcolm <dmalcolm@redhat.com>
249 * dump-context.h (dump_context::dump_loc): New decl.
250 * dumpfile.c (dump_context::dump_loc): New member function.
251 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
252 and dump_gimple_stmt.
253 (dump_context::dump_gimple_expr_loc): Likewise, using
255 (dump_context::dump_generic_expr_loc): Likewise, using
257 (dump_context::dump_printf_loc_va): Likewise, using
259 (dump_context::begin_scope): Explicitly using the global function
260 "dump_loc", rather than the member function.
262 2018-07-31 Martin Sebor <msebor@redhat.com>
264 PR tree-optimization/86741
265 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
267 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
269 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
270 depend on whether prefetch instructions will be emitted or not.
271 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
272 will be emitted or not.
273 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
274 (TARGET_SETMEM_PFD): New macros.
276 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
278 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
279 (NULL_STMT_VEC_INFO): Delete.
280 (stmt_vec_info::operator*): Likewise.
281 (stmt_vec_info::operator gimple *): Likewise.
282 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
283 of NULL_STMT_VEC_INFO.
284 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
285 (vect_reassociating_reduction_p): Likewise.
286 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
287 (vectorizable_store): Likewise.
288 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
289 (vec_info::free_stmt_vec_infos): Likewise.
291 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
293 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
294 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
296 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
298 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
299 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
300 (vec_info::free_stmt_vec_info): New private member functions.
301 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
302 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
303 * tree-parloops.c (gather_scalar_reductions): Remove calls to
304 set_stmt_vec_info_vec and free_stmt_vec_infos.
305 * tree-vect-loop.c (_loop_vec_info): Remove call to
306 set_stmt_vec_info_vec.
307 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
308 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
309 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
310 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
311 (vec_info::free_stmt_vec_info): ...these new functions. Remove
312 assignments in {vec_info::,}new_stmt_vec_info that are redundant
313 with the clearing in the xcalloc.
314 (stmt_vec_info_vec): Delete.
315 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
316 (vectorize_loops): Likewise.
317 (vec_info::~vec_info): Remove argument from call to
319 (vec_info::add_stmt): Remove vinfo argument from call to
322 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
324 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
325 rather than a gimple stmt.
326 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
327 information for pattern statements when passed the original
328 statement; instead wait to be passed the pattern statement itself.
329 Don't call set_vinfo_for_stmt here.
330 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
331 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
333 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
334 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
335 stmt_vec_infos entry.
337 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
339 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
340 * tree-vectorizer.c (vec_info::replace_stmt): New function.
341 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
342 * tree-vect-stmts.c (vectorizable_call): Likewise.
343 (vectorizable_simd_clone_call): Likewise.
345 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
347 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
348 * tree-vectorizer.c (vec_info::remove_stmt): New function.
349 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
350 * tree-vect-loop.c (vect_transform_loop): Likewise.
351 * tree-vect-slp.c (vect_schedule_slp): Likewise.
352 * tree-vect-stmts.c (vect_remove_stores): Likewise.
354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
356 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
357 (vect_dr_stmt): Delete.
358 * tree-vectorizer.c (vec_info::lookup_dr): New function.
359 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
361 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
362 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
363 (vect_verify_datarefs_alignment, vect_peeling_supportable)
364 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
365 (vect_analyze_data_refs): Likewise.
366 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
368 (vect_find_same_alignment_drs): Likewise.
369 (vect_slp_analyze_node_dependences): Update calls accordingly.
370 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
371 instead of DR_VECT_AUX.
372 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
373 of a vector data references. Use vec_info::lookup_dr instead of
375 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
376 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
377 instead of DR_VECT_AUX.
379 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
381 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
383 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
385 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
386 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
387 (vect_gen_prolog_loop_niters): Likewise.
389 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
391 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
392 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
393 (vect_known_alignment_in_bytes, vect_dr_behavior)
394 (vect_get_scalar_dr_size): Take references as dr_vec_infos
395 instead of data_references. Update calls to other routines for
396 which the same change has been made.
397 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
398 dr_vec_infos instead of stmt_vec_infos.
399 (vect_analyze_data_ref_dependence): Update call accordingly.
400 (vect_slp_analyze_data_ref_dependence)
401 (vect_record_base_alignments): Use DR_VECT_AUX.
402 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
403 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
404 (vector_alignment_reachable_p, vect_get_data_access_cost)
405 (vect_peeling_supportable, vect_analyze_group_access_1)
406 (vect_analyze_group_access, vect_analyze_data_ref_access)
407 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
408 (vect_compile_time_alias, vect_small_gap_p)
409 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
410 (vect_supportable_dr_alignment): Take references as dr_vec_infos
411 instead of data_references. Update calls to other routines for
412 which the same change has been made.
413 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
414 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
415 (vect_slp_analyze_and_verify_node_alignment)
416 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
417 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
418 (vect_setup_realignment): Use dr_vec_infos. Update calls after
420 (_vect_peel_info::dr): Replace with...
421 (_vect_peel_info::dr_info): ...this new field.
422 (vect_peeling_hash_get_most_frequent)
423 (vect_peeling_hash_choose_best_peeling): Update accordingly.
424 (vect_peeling_hash_get_lowest_cost):
425 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
426 routines for which the same change has been made.
427 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
429 * tree-vect-loop-manip.c (get_misalign_in_elems)
430 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
432 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
433 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
434 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
435 (get_group_load_store_type, get_negative_load_store_type)
436 (vect_get_data_ptr_increment, vectorizable_store)
437 (vectorizable_load): Likewise.
438 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
439 Update calls to other routines for which the same change has been made.
441 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
443 * tree-vectorizer.h (vec_info::move_dr): New member function.
444 (dataref_aux): Rename to...
445 (dr_vec_info): ...this and add "dr" and "stmt" fields.
446 (_stmt_vec_info::dr_aux): Update accordingly.
447 (_stmt_vec_info::data_ref_info): Delete.
448 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
449 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
450 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
451 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
453 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
454 (STMT_VINFO_DR_INFO): New macro.
455 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
456 (set_dr_misalignment): Update after rename of dataref_aux.
457 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
458 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
459 initialization of STMT_VINFO_DATA_REF.
460 * tree-vectorizer.c (vec_info::move_dr): New function.
461 * tree-vect-patterns.c (vect_recog_bool_pattern)
462 (vect_recog_mask_conversion_pattern)
463 (vect_recog_gather_scatter_pattern): Use it.
464 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
465 the "dr" and "stmt" fields of dr_vec_info instead of
468 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
470 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
471 (is_pattern_stmt_p): Use it.
472 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
473 on pattern statements.
475 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
477 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
478 original stmt as a stmt_vec_info rather than a gimple stmt.
479 (vect_pattern_recog_1): Take the statement directly as a
480 stmt_vec_info, rather than via a gimple_stmt_iterator.
481 Update call to vect_mark_pattern_stmts.
482 (vect_pattern_recog): Update calls accordingly.
484 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
486 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
487 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
488 a vect_def_type for the first argument.
489 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
490 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
491 operand if it isn't defined by a vectorized statement.
492 (vect_build_gather_load_calls): Remove the mask_dt argument and
493 update calls to vect_get_vec_def_for_stmt_copy.
494 (vectorizable_bswap): Likewise the dt argument.
495 (vectorizable_call): Update calls to vectorizable_bswap and
496 vect_get_vec_def_for_stmt_copy.
497 (vectorizable_simd_clone_call, vectorizable_assignment)
498 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
499 (vectorizable_comparison): Update calls to
500 vect_get_vec_def_for_stmt_copy.
501 (vectorizable_store): Likewise. Remove now-unnecessary calls to
503 (vect_get_loop_based_defs): Remove dt argument and update call
504 to vect_get_vec_def_for_stmt_copy.
505 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
506 and vect_get_vec_def_for_stmt_copy.
507 (vectorizable_load): Update calls to vect_build_gather_load_calls
508 and vect_get_vec_def_for_stmt_copy.
509 * tree-vect-loop.c (vect_create_epilog_for_reduction)
510 (vectorizable_reduction, vectorizable_live_operation): Update calls
511 to vect_get_vec_def_for_stmt_copy.
513 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
515 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
516 and gimple stmt arguments with a stmt_vec_info.
517 (vect_record_base_alignments): Update calls accordingly.
518 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
519 and gimple stmt arguments with a stmt_vec_info.
520 (vect_build_slp_tree_1): Remove vinfo argument and update call
521 to vect_record_max_nunits.
522 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
523 and vect_record_max_nunits.
525 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
527 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
528 file and take a stmt_vec_info instead of a gimple stmt.
529 (supportable_widening_operation, vect_finish_replace_stmt)
530 (vect_finish_stmt_generation, vect_get_store_rhs)
531 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
532 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
533 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
534 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
535 (vect_create_data_ref_ptr, bump_vector_ptr)
536 (vect_permute_store_chain, vect_setup_realignment)
537 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
538 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
539 (vectorizable_reduction, vectorizable_induction)
540 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
541 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
542 than gimple stmts as arguments.
543 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
544 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
545 (can_group_stmts_p, vect_check_gather_scatter)
546 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
547 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
548 (vect_permute_load_chain, vect_shift_permute_load_chain)
549 (vect_transform_grouped_load)
550 (vect_record_grouped_load_vectors): Likewise.
551 * tree-vect-loop.c (vect_fixup_reduc_chain)
552 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
553 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
554 (vectorizable_reduction, vectorizable_induction)
555 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
556 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
557 (vect_get_load_store_mask): Likewise.
558 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
559 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
560 * tree-vect-stmts.c (vect_mark_relevant)
561 (is_simple_and_all_uses_invariant)
562 (exist_non_indexing_operands_for_use_p, process_use)
563 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
564 (vect_get_vec_def_for_operand, vect_get_vec_defs)
565 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
566 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
567 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
568 (get_negative_load_store_type, get_load_store_type)
569 (vect_check_load_store_mask, vect_check_store_rhs)
570 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
571 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
572 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
573 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
574 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
575 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
576 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
577 (supportable_widening_operation): Likewise.
579 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
581 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
582 a stmt_vec_info instead of a gcall.
583 (vect_check_gather_scatter): Update call accordingly.
584 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
586 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
587 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
588 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
589 instead of a gimple stmt.
590 (vect_transform_loop): Update calls accordingly.
591 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
592 stmt_vec_infos instead of gimple stmts.
593 (vect_analyze_slp_instance): Update use accordingly.
594 * tree-vect-stmts.c (read_vector_array, write_vector_array)
595 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
596 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
597 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
598 (vect_gen_widened_results_half, vect_get_loop_based_defs)
599 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
600 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
603 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
605 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
606 of the worklist from a vector of gimple stmts to a vector of
608 * tree-vect-stmts.c (vect_mark_relevant, process_use)
609 (vect_mark_stmts_to_be_vectorized): Likewise
611 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
613 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
614 statement before passing it to vect_analyze_stmt.
615 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
616 the chain of phi vector definitions. Track the exit phi via its
618 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
619 STMT_VINFO_REDUC_DEF.
620 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
621 stmt_vec_infos to handle the statement chains.
622 (vect_get_slp_defs): Record the first statement in the node
623 using a stmt_vec_info.
624 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
625 statements here and pass their stmt_vec_info down to subroutines.
626 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
627 down to vect_finish_stmt_generation.
628 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
629 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
630 stmt_vec_infos to subroutines.
631 (vect_remove_stores): Use stmt_vec_infos to handle the statement
634 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
636 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
637 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
638 (vect_permute_store_chain, vect_setup_realignment)
639 (vect_permute_load_chain, vect_shift_permute_load_chain)
640 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
641 stmts internally, and when passing values to other vectorizer routines.
642 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
643 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
644 (vect_analyze_loop_operations, get_initial_def_for_reduction)
645 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
646 (vectorizable_reduction, vectorizable_induction)
647 (vectorizable_live_operation, vect_transform_loop_stmt)
648 (vect_transform_loop): Likewise.
649 * tree-vect-patterns.c (vect_reassociating_reduction_p)
650 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
651 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
652 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
653 (vect_slp_analyze_node_operations_1): Likewise.
654 * tree-vect-stmts.c (vect_mark_relevant, process_use)
655 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
656 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
657 (vect_finish_stmt_generation_1, get_group_load_store_type)
658 (get_load_store_type, vect_build_gather_load_calls)
659 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
660 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
661 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
662 (vectorizable_store, vectorizable_load, vectorizable_condition)
663 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
664 (supportable_widening_operation): Likewise.
665 (vect_get_vector_types_for_stmt): Likewise.
666 * tree-vectorizer.h (vect_dr_behavior): Likewise.
668 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
670 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
671 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
672 (vect_permute_store_chain, vect_permute_load_chain)
673 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
674 repeated stmt_vec_info lookups.
675 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
676 (vect_update_ivs_after_vectorizer): Likewise.
677 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
678 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
679 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
680 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
681 (vect_bb_slp_scalar_cost): Likewise.
682 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
684 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
686 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
687 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
688 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
689 to get gassigns and gcalls, rather than operating on generc gimple
691 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
692 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
693 (vectorizable_load, vect_analyze_stmt): Likewise.
694 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
696 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
698 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
699 return stmt_vec_infos rather than gimple stmts. Do not accept
701 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
703 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
704 Update use of get_later_stmt.
705 (vect_get_constant_vectors): Update call accordingly.
706 (vect_schedule_slp_instance): Likewise
707 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
708 (vect_slp_analyze_instance_dependence): Likewise.
709 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
711 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
713 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
714 (stmt_info_for_cost::stmt_info): ...this new field.
715 (add_stmt_costs): Update accordingly.
716 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
717 (vect_get_known_peeling_cost): Likewise.
718 (vect_estimate_min_profitable_iters): Likewise.
719 * tree-vect-stmts.c (record_stmt_cost): Likewise.
721 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
723 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
724 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
725 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
727 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
729 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
731 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
732 a gimple stmt to a stmt_vec_info.
733 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
735 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
737 * tree-vectorizer.h (vec_info::grouped_stores): Change from
738 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
739 (_loop_vec_info::reduction_chains): Likewise.
740 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
742 * tree-vect-slp.c (vect_analyze_slp): Likewise.
744 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
746 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
747 a gimple stmt to a stmt_vec_info.
748 (_stmt_vec_info::next_element): Likewise.
749 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
750 (vect_slp_analyze_and_verify_node_alignment)
751 (vect_analyze_group_access_1, vect_analyze_group_access)
752 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
753 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
754 (vect_supportable_dr_alignment): Update accordingly.
755 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
756 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
757 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
758 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
759 * tree-vect-slp.c (vect_build_slp_tree_1)
760 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
761 (vect_split_slp_store_group, vect_analyze_slp_instance)
762 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
763 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
764 (get_group_load_store_type, get_load_store_type)
765 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
766 (vect_transform_stmt, vect_remove_stores): Likewise.
768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
770 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
772 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
773 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
774 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
775 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
776 (vector_alignment_reachable_p, vect_get_data_access_cost)
777 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
778 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
779 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
780 (vect_analyze_group_access_1, vect_analyze_group_access)
781 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
782 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
783 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
784 result of vect_dr_stmt and use the stmt_vec_info instead of
785 the associated gimple stmt.
786 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
787 (vect_gen_prolog_loop_niters): Likewise.
788 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
792 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
793 to a vec<stmt_vec_info>.
794 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
795 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
797 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
798 to a vec<stmt_vec_info>.
799 (bst_traits::value_type, bst_traits::value_type): Likewise.
800 (bst_traits::hash): Update accordingly.
801 (vect_get_and_check_slp_defs): Change the stmts parameter from
802 a vec<gimple *> to a vec<stmt_vec_info>.
803 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
804 (vect_build_slp_tree): Likewise.
805 (vect_build_slp_tree_2): Likewise. Update uses of
806 SLP_TREE_SCALAR_STMTS.
807 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
808 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
809 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
810 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
811 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
812 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
813 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
814 (vect_get_constant_vectors, vect_get_slp_defs)
815 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
816 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
817 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
818 instead of gimple stmts.
819 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
820 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
821 (vect_slp_analyze_instance_dependence): Update uses of
822 SLP_TREE_SCALAR_STMTS.
823 (vect_slp_analyze_and_verify_node_alignment): Likewise.
824 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
825 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
826 (get_initial_defs_for_reduction): Likewise.
827 (vect_create_epilog_for_reduction): Likewise.
828 (vectorize_fold_left_reduction): Likewise.
829 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
830 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
831 (can_vectorize_live_stmts): Likewise.
833 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
835 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
836 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
837 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
839 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
841 (gather_scalar_reductions): Update after above interface changes.
842 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
843 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
845 (vect_force_simple_reduction): Likewise.
846 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
847 LOOP_VINFO_REDUCTIONS.
848 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
850 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
852 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
853 a gimple stmt to a stmt_vec_info.
854 * tree-vect-loop.c (vect_active_double_reduction_p)
855 (vect_force_simple_reduction, vectorizable_reduction): Update
858 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
860 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
861 vec<gimple *> to a vec<stmt_vec_info>.
862 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
863 the reduction_phis argument from a vec<gimple *> to a
865 (vectorizable_reduction): Likewise the phis local variable that
866 is passed to vect_create_epilog_for_reduction. Update for new type
867 of SLP_TREE_VEC_STMTS.
868 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
869 (vectorizable_live_operation): Likewise.
870 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
871 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
873 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
875 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
876 a gimple stmt to a stmt_vec_info.
877 (vectorizable_condition, vectorizable_live_operation)
878 (vectorizable_reduction, vectorizable_induction): Pass back the
879 vectorized statement as a stmt_vec_info.
880 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
881 use of STMT_VINFO_VEC_STMT.
882 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
883 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
884 as stmt_vec_infos rather than gimple stmts.
885 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
887 (vectorizable_live_operation): Likewise.
888 (vectorizable_reduction, vectorizable_induction): Likewise,
889 updating use of STMT_VINFO_VEC_STMT.
890 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
891 of STMT_VINFO_VEC_STMT.
892 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
893 (vectorizable_simd_clone_call, vectorizable_conversion)
894 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
895 (vectorizable_store, vectorizable_load, vectorizable_condition)
896 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
897 from a gimple stmt to a stmt_vec_info.
898 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
899 pointer to a stmt_vec_info to the vectorizable_* routines.
901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
903 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
904 a gimple stmt to a stmt_vec_info.
905 (is_pattern_stmt_p): Update accordingly.
906 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
907 (vect_record_grouped_load_vectors): Likewise.
908 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
909 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
910 (vect_model_reduction_cost): Likewise.
911 (vect_create_epilog_for_reduction): Likewise.
912 (vectorizable_reduction, vectorizable_induction): Likewise.
913 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
914 Return the stmt_vec_info for the pattern statement.
915 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
916 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
917 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
918 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
919 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
920 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
921 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
922 (free_stmt_vec_info, vect_is_simple_use): Likewise.
924 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
926 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
927 (vect_finish_stmt_generation): Likewise.
928 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
929 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
930 (vect_build_gather_load_calls): Use the return value of the above
931 functions instead of a separate call to vinfo_for_stmt. Use narrow
932 scopes for the input gimple stmt and wider scopes for the associated
933 stmt_vec_info. Use vec_info::lookup_def when setting these
934 stmt_vec_infos from an SSA_NAME definition.
935 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
936 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
937 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
938 (vectorizable_store, vectorizable_load, vectorizable_condition)
939 (vectorizable_comparison): Likewise.
940 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
941 (vectorizable_reduction): Likewise.
943 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
945 * tree-vectorizer.h (vect_is_simple_use): Add an optional
946 stmt_vec_info * parameter before the optional gimple **.
947 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
948 (process_use, vect_get_vec_def_for_operand_1): Update callers.
949 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
950 * tree-vect-loop.c (vectorizable_reduction): Likewise.
951 (vectorizable_live_operation): Likewise.
952 * tree-vect-patterns.c (type_conversion_p): Likewise.
953 (vect_look_through_possible_promotion): Likewise.
954 (vect_recog_rotate_pattern): Likewise.
955 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
957 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
959 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
960 a typedef to a wrapper class.
961 (NULL_STMT_VEC_INFO): New macro.
962 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
963 (stmt_vec_info::operator*): New function.
964 (stmt_vec_info::operator gimple *): Likewise.
965 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
966 (add_stmt_costs): Likewise.
967 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
968 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
969 (vect_get_known_peeling_cost): Likewise.
970 (vect_estimate_min_profitable_iters): Likewise.
971 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
972 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
973 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
974 (vectorizable_store, free_stmt_vec_infos): Likewise.
975 (new_stmt_vec_info): Change return type of xcalloc to
978 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
980 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
981 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
982 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
983 a single_imm_use-based sequence.
984 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
986 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
988 * tree-vectorizer.h (vec_info::lookup_def): Declare.
989 * tree-vectorizer.c (vec_info::lookup_def): New function.
990 * tree-vect-patterns.c (vect_get_internal_def): Use it.
991 (vect_widened_op_tree): Likewise.
992 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
993 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
994 (vectorizable_reduction): Likewise.
995 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
997 (vect_is_slp_reduction): Update calls accordingly. Use
998 vec_info::lookup_def.
999 (vect_is_simple_reduction): Likewise
1000 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
1002 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1004 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
1005 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
1006 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
1008 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
1009 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
1010 (vect_update_vf_for_slp, vect_analyze_loop_operations)
1011 (vect_is_slp_reduction, vectorizable_induction)
1012 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
1013 * tree-vect-patterns.c (vect_init_pattern_stmt):
1014 (vect_determine_min_output_precision_1, vect_determine_precisions)
1015 (vect_pattern_recog): Likewise.
1016 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
1017 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
1018 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
1019 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
1020 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
1021 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
1022 info field from a loop to a loop_vec_info.
1024 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1026 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
1027 (vec_info::add_stmt): Declare.
1028 * tree-vectorizer.c (vec_info::add_stmt): New function.
1029 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
1030 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
1031 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1032 (vectorizable_induction): Likewise.
1033 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
1034 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1035 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
1036 (vectorizable_load): Likewise.
1037 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1038 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
1039 (vect_recog_gather_scatter_pattern): Likewise.
1040 (append_pattern_def_seq): Likewise. Remove a check that is
1041 performed by add_stmt itself.
1043 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1045 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
1046 which make_ssa_name was called with new_stmt before new_stmt
1049 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1051 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
1053 (vect_is_slp_reduction): ...here...
1054 (vect_is_simple_reduction): ...and here. Remove repetition of tests
1055 that are already known to be false.
1057 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1059 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
1060 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
1061 STMT_VINFO_NUM_SLP_USES when it's true.
1062 (vect_free_slp_instance): Add a final_p parameter and pass it to
1064 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
1065 (vect_analyze_slp_instance): Likewise.
1066 (vect_slp_analyze_operations): Likewise.
1067 (vect_slp_analyze_bb_1): Likewise.
1068 * tree-vectorizer.c (vec_info): Likewise.
1069 * tree-vect-loop.c (vect_transform_loop): Likewise.
1071 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1073 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1074 function is not called for second and subsequent members of
1077 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1079 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
1080 cases for nested loops from here to ...
1081 (vect_create_epilog_for_reduction): ...here. Only call
1082 vect_is_simple_use for inner-loop reductions.
1084 2018-07-31 Martin Liska <mliska@suse.cz>
1086 PR gcov-profile/85338
1087 PR gcov-profile/85350
1088 PR gcov-profile/85372
1089 * profile.c (struct location_triplet): New.
1090 (struct location_triplet_hash): Likewise.
1091 (output_location): Do not output a BB that
1092 is already recorded for a line.
1093 (branch_prob): Use streamed_locations.
1095 2018-07-31 Martin Liska <mliska@suse.cz>
1097 PR gcov-profile/85370
1098 * coverage.c (coverage_begin_function): Do not mark target
1099 clones as artificial functions.
1101 2018-07-31 Martin Liska <mliska@suse.cz>
1103 PR gcov-profile/83813
1104 PR gcov-profile/84758
1105 PR gcov-profile/85217
1106 PR gcov-profile/85332
1107 * profile.c (branch_prob): Do not record GOTO expressions
1108 for GIMPLE statements which locations are already streamed.
1110 2018-07-31 Olivier Hainque <hainque@adacore.com>
1112 * gcc.c (handle_spec_function): Accept a soft_matched_part
1113 argument, as do_spec_1. Pass it down to ...
1114 (eval_spec_function): Accept a soft_matched_part argument,
1115 and pass it down to ...
1116 (do_spec_2): Accept a soft_matched_part argument, and pass
1117 it down to do_spec_1.
1118 (do_spec_1): Pass soft_matched_part to handle_spec_function.
1119 (handle_braces): Update call to handle_spec_function.
1120 (driver::set_up_specs): Update calls to do_spec_2.
1121 (compare_debug_dump_opt_spec_function): Likewise.
1122 (compare_debug_self_opt_spec_function): Likewise.
1124 2018-07-31 Olivier Hainque <hainque@adacore.com>
1126 * common.opt (nolibc): New option.
1127 * doc/invoke.texi (Link Options): Document it.
1128 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
1129 * config/alpha/linux.h: Likewise.
1130 * config/arc/elf.h: Likewise.
1131 * config/arm/uclinux-elf.h: Likewise.
1132 * config/arm/unknown-elf.h: Likewise.
1133 * config/avr/avrlibc.h: Likewise.
1134 * config/bfin/bfin.h: Likewise.
1135 * config/bfin/linux.h: Likewise.
1136 * config/bfin/uclinux.h: Likewise.
1137 * config/darwin.h: Likewise.
1138 * config/darwin10.h: Likewise.
1139 * config/darwin12.h: Likewise.
1140 * config/gnu-user.h: Likewise.
1141 * config/lm32/uclinux-elf.h: Likewise.
1142 * config/pa/pa-hpux11.h: Likewise.
1143 * config/pa/pa64-hpux.h: Likewise.
1144 * config/sparc/sparc.h: Likewise.
1146 2018-07-31 Olivier Hainque <hainque@adacore.com>
1148 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
1149 undefined variables.
1151 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1154 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
1157 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1159 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1160 terminated string literal.
1162 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1164 PR rtl-optimization/85160
1165 * combine.c (is_just_move): New function.
1166 (try_combine): Allow combining two instructions into two if neither of
1167 the original instructions was a move.
1169 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
1172 * doc/extend.texi (Global Register Variables): Discourage use of type
1174 (Local Register Variables): Likewise.
1176 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
1178 PR tree-optimization/86506
1179 * hwint.h (ceil_log2): Resync with hwint.c implementation.
1181 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
1184 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
1185 hard_regno, make sure no insn between `from` and `to` clobbers it.
1187 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
1188 Tom de Vries <tdevries@suse.de>
1190 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
1191 (PTX_DEFAULT_RUNTIME_DIM): ... this.
1192 (nvptx_goacc_validate_dims): Set default worker and gang dims to
1193 PTX_DEFAULT_RUNTIME_DIM.
1194 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
1196 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
1198 * config/pa/pa.c (pa_output_addr_vec): Align address table.
1199 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
1200 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
1202 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
1204 * config/rs6000/constraints.md (wG constraint): Delete, no longer
1206 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
1207 predicate to reflect toc fusion has been deleted.
1208 (toc_fusion_mem_raw): Delete, no longer used.
1209 (toc_fusion_mem_wrapped): Likewise.
1210 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
1212 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
1213 Delete, no longer used.
1214 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
1215 meant to be used for toc fusion.
1216 (rs6000_debug_print_mode): Delete toc fusion debugging.
1217 (rs6000_debug_reg_global): Likewise.
1218 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
1219 fusion and secondary reload support that were never used.
1220 (rs6000_option_override_internal): Delete TOC fusion, that was only
1221 partially defined, and it did not work unless you also used the
1223 (rs6000_legitimate_address_p): Delete TOC fusion support.
1224 (rs6000_opt_masks): Likewise.
1225 (fusion_wrap_memory_address): Delete function, no longer used.
1226 (fusion_split_address); Delete TOC fusion support.
1227 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
1228 longer used with toc fusion being deleted.
1229 (TARGET_TOC_FUSION_FP): Likewise.
1230 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
1232 (toc fusion spliter): Delete TOC fusion support.
1233 (toc_fusionload_<mode>): Likewise.
1234 (toc_fusionload_di): Likewise.
1235 (fusion_gpr_load_<mode>): Delete generator function, this insn no
1236 longer needs to be named. Rename predicate to delete TOC fusion.
1237 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1238 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1239 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1240 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1241 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
1243 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
1245 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
1246 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
1247 __int128 in built-in function prototypes.
1248 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
1249 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
1251 2018-07-27 Martin Sebor <msebor@redhat.com>
1253 PR tree-optimization/86696
1254 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
1255 types, including enums.
1256 (handle_char_store): Be prepared for the above function to fail.
1258 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
1260 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
1261 when optimization level is lower than 2 or optimize for size.
1263 2018-07-26 Martin Sebor <msebor@redhat.com>
1265 PR tree-optimization/86043
1266 PR tree-optimization/86042
1267 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
1268 (get_string_cst_length): Rename...
1269 (get_min_string_length): ...to this. Add argument.
1270 (handle_char_store): Extend to handle multi-character stores by
1272 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
1273 * tree.h (initializer_zerop): Add argument.
1275 2018-07-26 Jakub Jelinek <jakub@redhat.com>
1278 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
1279 declare target to variables if they have always,{to,from,tofrom} map
1282 2018-07-26 Martin Liska <mliska@suse.cz>
1285 * lto-wrapper.c: Add linker_output as prefix
1286 for ltrans_output_file.
1288 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
1290 PR rtl-optimization/85805
1291 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
1292 value for hard registers if that was written in the same mode.
1294 2018-07-26 Martin Liska <mliska@suse.cz>
1296 PR gcov-profile/86536
1297 * gcov.c (format_gcov): Use printf format %.*f directly
1298 and do not handle special values.
1300 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1302 * common/config/arc/arc-common.c (arc_option_optimization_table):
1303 Update default optimizations for size.
1305 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1307 * config/arc/arc.md (movsf_insn): Add short instruction selection.
1308 * config/arc/constraints.md (CfZ): New constraint.
1309 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
1310 (subsf3_fpu): Likewise.
1311 (cmpsf_fpu): Likewise.
1312 (cmpsf_fpu_uneq): Likewise.
1314 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1316 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
1318 (arc_is_uncached_mem_p): uncached applies to both the variable and
1321 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1323 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
1326 2018-07-25 David Malcolm <dmalcolm@redhat.com>
1328 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
1329 field "m_scopes" from vec to auto_vec.
1331 2018-07-25 Martin Liska <mliska@suse.cz>
1333 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
1336 2018-07-25 Richard Biener <rguenther@suse.de>
1339 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
1340 special wrt context_die late.
1341 (gen_subprogram_die): Re-use DIEs in local scope.
1343 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
1345 PR tree-optimization/86644
1346 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
1348 2018-07-25 Martin Liska <mliska@suse.cz>
1351 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
1352 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
1354 2018-07-25 Martin Liska <mliska@suse.cz>
1357 * params.def: Explain ASan abbreviation and provide
1358 a documentation link.
1360 2018-07-24 Martin Sebor <msebor@redhat.com>
1362 PR tree-optimization/86622
1363 PR tree-optimization/86532
1364 * builtins.h (string_length): Declare.
1365 * builtins.c (c_strlen): Correct handling of non-constant offsets.
1366 (check_access): Be prepared for non-constant length ranges.
1367 (string_length): Make extern.
1368 * expr.c (string_constant): Only handle the minor non-constant
1369 array index. Use string_constant to compute the length of
1370 a generic string constant.
1372 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
1374 PR tree-optimization/86618
1375 * tree-vect-stmts.c (vectorizable_call): Don't take the address
1376 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
1378 2018-07-24 David Malcolm <dmalcolm@redhat.com>
1380 PR tree-optimization/86636
1381 * json.cc (json::object::set): Fix comment. Add assertions.
1382 (json::array::append): Move here from json.h. Add comment and an
1384 (json::string::string): Likewise.
1385 * json.h (json::array::append): Move to json.cc.
1386 (json::string::string): Likewise.
1387 * optinfo-emit-json.cc
1388 (optrecord_json_writer::impl_location_to_json): Assert that we
1389 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
1390 wrapper around it. Expand the location once, rather than three
1392 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
1393 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
1395 (optrecord_json_writer::optinfo_to_json): Likewise, in four
1396 places. Fix some overlong lines.
1398 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
1400 * config/aarch64/aarch64-simd.md
1401 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
1402 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
1403 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
1404 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
1405 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
1406 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
1407 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
1408 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
1409 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
1411 2018-07-24 Jakub Jelinek <jakub@redhat.com>
1414 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
1415 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
1416 and abs_d == d, do the power of two handling if profitable.
1418 2018-07-24 Richard Biener <rguenther@suse.de>
1420 * match.pd: Add BIT_FIELD_REF canonicalizations.
1422 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1425 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
1427 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1429 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
1430 terminated STRING_CST object.
1432 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1434 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
1436 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1438 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
1439 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
1440 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
1441 the elements into a register.
1442 (rs6000_split_v4si_init_di_reg): Delete.
1443 (rs6000_split_v4si_init): Delete.
1444 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
1445 (vsx_init_v4si): Rewrite as a define_expand.
1447 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1449 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
1450 zero_extend argument from memory): New.
1452 2018-07-22 Martin Sebor <msebor@redhat.com>
1455 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
1456 diagnosing calls with unknown arguments unless -Walloca-larger-than
1457 is restricted to less than PTRDIFF_MAX bytes.
1459 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
1461 * doc/gcov.texi (Invoking Gcov): Editorial changes.
1463 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1465 * pretty-print.c (text_info::set_location): Remove redundant
1466 "line_table" parameter from call to rich_location::set_range.
1468 2018-07-20 Martin Sebor <msebor@redhat.com>
1471 * builtins.c (expand_builtin_alloca): Adjust.
1472 * calls.c (alloc_max_size): Simplify.
1473 * cgraphunit.c (cgraph_node::expand): Adjust.
1474 * common.opt (larger_than_size, warn_frame_larger_than): Remove
1476 (frame_larger_than_size): Same.
1477 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
1478 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
1480 * doc/invoke.texi (GCC Command Options): Document option arguments.
1481 Explain byte-size arguments and suffixes.
1482 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
1483 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
1484 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
1485 * doc/options.texi (UInteger): Expand.
1486 (Host_Wide_Int, ByteSize): Document new properties.
1487 * final.c (final_start_function_1): Include sizes in an error message.
1488 * function.c (frame_offset_overflow): Same.
1489 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
1490 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
1491 Diagnose unbounded alloca calls only for limits of less than
1493 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
1494 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
1496 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
1497 only for limits of less than PTRDIFF_MAX.
1498 * langhooks-def.h (lhd_handle_option): Change function argument
1500 * langhooks.c (lhd_handle_option): Same.
1501 * langhooks.h (handle_option): Same.
1502 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
1504 (var_type, var_type_struct): Same.
1505 (var_set): Handle ByteSize flag.
1506 * optc-gen.awk: Add comments to output to ease debugging. Make
1507 use of HOST_WIDE_INT where appropriate.
1508 * opts-gen-save.awk: Use %lx to format unsigned long.
1509 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
1510 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
1511 arguments. Parse bytes-size suffixes.
1512 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
1513 (enum_value_to_arg): Same.
1514 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
1515 (handle_option): Adjust.
1516 (generate_option): Change function argument to HOST_WIDE_INT.
1517 (cmdline_handle_error): Adjust.
1518 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
1519 (set_option): Change function argument to HOST_WIDE_INT.
1520 (option_enabled): Handle cl_host_wide_int.
1521 (get_option_state): Handle CLVC_SIZE.
1522 (control_warning_option): Same.
1523 * opts.c (common_handle_option): Change function argument to
1524 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
1525 OPT_Wvla_larger_than_.
1526 * opts.h (enum cl_var_type): Add an enumerator.
1527 * stor-layout.c (layout_decl): Print a more meaningful warning.
1528 * toplev.c (output_stack_usage): Adjust.
1530 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
1532 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
1533 call to inline_expand_builtin_string_cmp.
1534 (expand_builtin_strcmp): Likewise.
1535 (expand_builtin_strncmp): Likewise.
1536 (inline_string_cmp): Delete the last parameter, change char_type_node
1537 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
1539 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
1540 the inlining expansion on target where the type of the call has same or
1541 narrower precision than unsigned char.
1543 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1545 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
1546 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
1547 * common.opt (fsave-optimization-record): New option.
1548 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
1549 * doc/invoke.texi (-fsave-optimization-record): New option.
1550 * dumpfile.c: Include "optinfo-emit-json.h".
1551 (struct kv_pair): Move to coretypes.h.
1552 (optgroup_options): Make non-static.
1553 (dump_context::end_scope): Call
1554 optimization_records_maybe_pop_dump_scope.
1555 * dumpfile.h (optgroup_options): New decl.
1556 * json.cc: New file.
1558 * optinfo-emit-json.cc: New file.
1559 * optinfo-emit-json.h: New file.
1560 * optinfo.cc: Include "optinfo-emit-json.h".
1561 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
1562 (optinfo_enabled_p): Check optimization_records_enabled_p.
1563 (optinfo_wants_inlining_info_p): Likewise.
1564 * optinfo.h: Update comment.
1565 * profile-count.c (profile_quality_as_string): New function.
1566 * profile-count.h (profile_quality_as_string): New decl.
1567 (profile_count::quality): New accessor.
1568 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
1569 and optinfo_emit_json_cc_tests.
1570 * selftest.h (selftest::json_cc_tests): New decl.
1571 (selftest::optinfo_emit_json_cc_tests): New decl.
1572 * toplev.c: Include "optinfo-emit-json.h".
1573 (compile_file): Call optimization_records_finish.
1574 (do_compile): Call optimization_records_start.
1575 * tree-ssa-live.c: Include optinfo.h.
1576 (remove_unused_scope_block_p): Retain inlining information if
1577 optinfo_wants_inlining_info_p returns true.
1579 2018-07-20 Richard Biener <rguenther@suse.de>
1582 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
1583 to cover -flto-partition=none.
1585 2018-07-20 Martin Liska <mliska@suse.cz>
1587 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
1588 (get_decl_source_range): Remove unused function.
1590 2018-07-20 Richard Biener <rguenther@suse.de>
1592 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
1593 (struct vn_phi_s): Likewise.
1594 (struct vn_reference_s): Likewise.
1595 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
1596 for searching the slot of an entry known to be in the hash itself.
1597 (vn_phi_hasher::equal): Likewise.
1598 (vn_reference_hasher::equal): Likewise.
1599 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
1601 (optimistic_info, current_info): Remove, keeping only valid_info.
1602 (vn_reference_lookup_1): Remove fallback lookup.
1603 (vn_reference_lookup_2): Likewise.
1604 (vn_nary_op_lookup_1): Likewise.
1605 (vn_phi_lookup): Likewise.
1606 (vn_nary_build_or_lookup_1): Make sure to not chain the built
1608 (vn_reference_insert): Adjust, chain the inserted hash element
1609 at last_inserted_ref.
1610 (vn_reference_insert_pieces): Likewise.
1611 (visit_reference_op_call): Likewise.
1612 (vn_nary_op_insert_into): Chain the inserted hash element at
1614 (vn_nary_op_insert_pieces): Adjust.
1615 (vn_nary_op_insert): Likewise.
1616 (vn_nary_op_insert_stmt): Likewise.
1617 (vn_phi_insert): Adjust, chain the inserted hash element at
1619 (process_scc): Remove clearing and copying the optimistic
1620 table. Instead remove elements inserted during an optimistic
1621 iteration from the single table we maintain.
1622 (init_scc_vn): Adjust.
1623 (free_scc_vn): Likewise.
1624 (sccvn_dom_walker::record_cond): Likewise.
1625 (sccvn_dom_walker::after_dom_children): Likewise.
1627 2018-07-19 Martin Sebor <msebor@redhat.com>
1629 PR tree-optimization/84047
1630 PR tree-optimization/83776
1631 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
1632 (check_array_bounds): Call it.
1634 2018-07-19 Martin Sebor <msebor@redhat.com>
1636 * align.h (align_flags): Use member initialization.
1638 2018-07-19 David Malcolm <dmalcolm@redhat.com>
1640 * Makefile.in (OBJS): Add optinfo.o.
1641 * coretypes.h (class symtab_node): New forward decl.
1642 (struct cgraph_node): New forward decl.
1643 (class varpool_node): New forward decl.
1644 * dump-context.h: New file.
1645 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
1647 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
1648 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
1649 (set_alt_dump_file): Likewise.
1650 (dump_context::~dump_context): New dtor.
1651 (dump_gimple_stmt): Move implementation to...
1652 (dump_context::dump_gimple_stmt): ...this new member function.
1653 Add the stmt to any pending optinfo, creating one if need be.
1654 (dump_gimple_stmt_loc): Move implementation to...
1655 (dump_context::dump_gimple_stmt_loc): ...this new member function.
1656 Start a new optinfo and add the stmt to it.
1657 (dump_gimple_expr): Move implementation to...
1658 (dump_context::dump_gimple_expr): ...this new member function.
1659 Add the stmt to any pending optinfo, creating one if need be.
1660 (dump_gimple_expr_loc): Move implementation to...
1661 (dump_context::dump_gimple_expr_loc): ...this new member function.
1662 Start a new optinfo and add the stmt to it.
1663 (dump_generic_expr): Move implementation to...
1664 (dump_context::dump_generic_expr): ...this new member function.
1665 Add the tree to any pending optinfo, creating one if need be.
1666 (dump_generic_expr_loc): Move implementation to...
1667 (dump_context::dump_generic_expr_loc): ...this new member
1668 function. Add the tree to any pending optinfo, creating one if
1670 (dump_printf): Move implementation to...
1671 (dump_context::dump_printf_va): ...this new member function. Add
1672 the text to any pending optinfo, creating one if need be.
1673 (dump_printf_loc): Move implementation to...
1674 (dump_context::dump_printf_loc_va): ...this new member function.
1675 Start a new optinfo and add the stmt to it.
1676 (dump_dec): Move implementation to...
1677 (dump_context::dump_dec): ...this new member function. Add the
1678 value to any pending optinfo, creating one if need be.
1679 (dump_context::dump_symtab_node): New member function.
1680 (dump_context::get_scope_depth): New member function.
1681 (dump_context::begin_scope): New member function.
1682 (dump_context::end_scope): New member function.
1683 (dump_context::ensure_pending_optinfo): New member function.
1684 (dump_context::begin_next_optinfo): New member function.
1685 (dump_context::end_any_optinfo): New member function.
1686 (dump_context::s_current): New global.
1687 (dump_context::s_default): New global.
1688 (dump_scope_depth): Delete global.
1689 (dumpfile_ensure_any_optinfo_are_flushed): New function.
1690 (dump_symtab_node): New function.
1691 (get_dump_scope_depth): Reimplement in terms of dump_context.
1692 (dump_begin_scope): Likewise.
1693 (dump_end_scope): Likewise.
1694 (selftest::temp_dump_context::temp_dump_context): New ctor.
1695 (selftest::temp_dump_context::~temp_dump_context): New dtor.
1696 (selftest::verify_item): New function.
1697 (ASSERT_IS_TEXT): New macro.
1698 (ASSERT_IS_TREE): New macro.
1699 (ASSERT_IS_GIMPLE): New macro.
1700 (selftest::test_capture_of_dump_calls): New test.
1701 (selftest::dumpfile_c_tests): Call it.
1702 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
1703 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
1704 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
1705 descriptive comment.
1706 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
1707 (dump_node, dump_bb): Move these unrelated decls.
1708 (class dump_manager): Add leading comment.
1709 * optinfo.cc: New file.
1710 * optinfo.h: New file.
1712 2018-07-19 Michael Collison <michael.collison@arm.com>
1713 Richard Henderson <rth@redhat.com>
1715 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
1716 (subti): Handle op1 zero.
1717 (subvti4, usub4ti4): New.
1718 (*sub<GPI>3_compare1_imm): New.
1719 (sub<GPI>3_carryinCV): New.
1720 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
1721 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
1723 2018-07-19 Michael Collison <michael.collison@arm.com>
1724 Richard Henderson <rth@redhat.com>
1726 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
1727 (addti3): Create simpler code if low part is already known to be 0.
1728 (addvti4, uaddvti4): New.
1729 (*add<GPI>3_compareC_cconly_imm): New.
1730 (*add<GPI>3_compareC_cconly): New.
1731 (*add<GPI>3_compareC_imm): New.
1732 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
1733 handle constants within this pattern..
1734 (*add<GPI>3_compareV_cconly_imm): New.
1735 (*add<GPI>3_compareV_cconly): New.
1736 (*add<GPI>3_compareV_imm): New.
1737 (add<GPI>3_compareV): New.
1738 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
1739 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
1740 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
1741 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
1742 with 'comparison' operator.
1743 (*add<GPI>3_compareV_cconly_imm): Ditto.
1744 (*add<GPI>3_compareV_cconly): Ditto.
1745 (*add<GPI>3_compareV_imm): Ditto.
1746 (add<GPI>3_compareV): Ditto.
1747 (add<mode>3_carryinC): Ditto.
1748 (*add<mode>3_carryinC_zero): Ditto.
1749 (*add<mode>3_carryinC): Ditto.
1750 (add<mode>3_carryinV): Ditto.
1751 (*add<mode>3_carryinV_zero): Ditto.
1752 (*add<mode>3_carryinV): Ditto.
1754 2018-07-19 Michael Collison <michael.collison@arm.com>
1755 Richard Henderson <rth@redhat.com>
1757 * config/aarch64/aarch64-modes.def (CC_V): New.
1758 * config/aarch64/aarch64-protos.h
1759 (aarch64_addti_scratch_regs): Declare
1760 (aarch64_subvti_scratch_regs): Declare.
1761 (aarch64_expand_subvti): Declare.
1762 (aarch64_gen_unlikely_cbranch): Declare
1763 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
1764 for signed overflow using CC_Vmode.
1765 (aarch64_get_condition_code_1): Handle CC_Vmode.
1766 (aarch64_gen_unlikely_cbranch): New function.
1767 (aarch64_addti_scratch_regs): New function.
1768 (aarch64_subvti_scratch_regs): New function.
1769 (aarch64_expand_subvti): New function.
1771 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1773 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1775 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1776 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1779 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1782 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1783 address check not strict.
1785 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1787 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1789 (store_pair_lanes<mode>): Likewise.
1790 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1791 enum value 'ADDR_QUERY_LDP_STP_N'.
1792 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1793 (aarch64_print_address_internal): Add declaration.
1794 (aarch64_print_ldpstp_address): Remove.
1795 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1796 (aarch64_print_operand): Change printing of 'y'.
1797 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1798 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1799 'true' rather than '1'.
1800 * config/aarch64/constraints.md (Uml): Likewise.
1801 (Uml): Rename to Umn.
1804 2018-07-19 Richard Biener <rguenther@suse.de>
1806 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1808 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1809 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1810 (vn_reference_hasher): Likewise.
1811 (struct vn_tables_s): Remove obstack and alloc-pool members.
1812 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1813 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1814 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1815 (vn_reference_insert_pieces): Likewise.
1816 (alloc_vn_nary_op_noinit): Adjust.
1817 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1818 (vn_phi_eq): Adjust.
1819 (shared_lookup_phiargs): Remove.
1820 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1821 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1822 (visit_reference_op_call): Likewise.
1823 (copy_nary, copy_phi, copy_reference): Remove.
1824 (process_scc): Rewind the obstack when iterating. Do not
1825 copy the elements to valid_info but just move them from one
1826 hashtable to the other.
1827 (allocate_vn_table): Adjust.
1828 (free_vn_table): Likewise.
1829 (init_scc_vn): Likewise.
1830 (free_scc_vn): Likewise.
1832 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1835 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1836 indirect_return as function type attribute.
1837 (ix86_attribute_table): Change indirect_return to function
1839 * doc/extend.texi: Update indirect_return attribute.
1841 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1843 * wide-int.h (widest2_int): New.
1844 * gimple-fold.c (arith_overflowed_p): Use it.
1845 * tree.h (widest2_int_cst): New.
1846 * tree-vrp.c (wide_int_binop_overflow): Rename from
1847 vrp_int_const_binop.
1848 Rewrite to work on trees.
1849 (extract_range_from_multiplicative_op_1): Abstract code to...
1850 (wide_int_range_min_max): ...here.
1851 (wide_int_range_cross_product): ...and here.
1852 (extract_range_from_binary_expr_1): Abstract overflow code to...
1853 (wide_int_range_mult_wrapping): ...here.
1854 * tree-vrp.h (wide_int_range_cross_product): New.
1855 (wide_int_range_mult_wrapping): New.
1857 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1858 Julia Koval <julia.koval@intel.com>
1860 * config/i386/x86-tune-costs.h (skylake_memcpy,
1861 skylake_memset): Replace rep_prefix with unrolling for size 512.
1863 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1866 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1869 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1871 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1872 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1873 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1874 previously known as "PowerPC AltiVec Built-in Functions". Move
1875 some material to new subsubsections "PowerPC AltiVec Built-in
1876 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1878 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1879 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1880 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1881 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1883 2018-07-18 Richard Biener <rguenther@suse.de>
1885 PR tree-optimization/86557
1886 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1889 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1891 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1893 2018-07-17 Jeff Law <law@redhat.com>
1895 * config/arm/arm.c (get_label_padding): Update for recent
1896 changes to label_to_alignment.
1898 PR tree-optimization/86010
1899 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1901 * config/mips/mips.c (vr4130_align_insns): Update for recent
1902 changes to label_to_alignment.
1904 * config/frv/frv.c (frv_label_align): Update for recent changes
1905 to label_to_alignment.
1907 * config/nios2/nios2.c (nios2_label_align): Update for recent
1908 changes which dropped ALIGN_LABELS_LOG.
1910 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1912 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1913 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1915 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1917 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1918 deprecated align_labels_log.
1920 2018-07-17 Richard Biener <rguenther@suse.de>
1923 * dwarf2out.c (init_sections_and_labels): Always generate
1924 a debug_line_str_section for early LTO debug.
1925 (dwarf2out_finish): Reset debug_line_str_hash output early.
1926 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1928 (dwarf2out_early_finish): Output debug_line_str.
1930 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1932 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1933 index register on z196 or later.
1935 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1937 * config/s390/s390.c (s390_default_align): Set default function
1939 (s390_override_options_after_change): Call s390_default align.
1940 (s390_option_override_internal): Call s390_default_align.
1941 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1943 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1946 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1947 to _looptemp_ clauses, other than the first two.
1949 2018-07-17 Martin Liska <mliska@suse.cz>
1951 * opts.c: Do not enable OPT_falign_* for -Os.
1953 2018-07-17 Martin Liska <mliska@suse.cz>
1955 * align.h (MAX_CODE_ALIGN): New.
1956 (MAX_CODE_ALIGN_VALUE): New.
1957 * common/config/i386/i386-common.c (ix86_handle_option):
1958 (MAX_CODE_ALIGN): Moved to align.h.
1959 * final.c (MAX_CODE_ALIGN): Likewise.
1960 * opts.c (parse_and_check_align_values):
1961 (MAX_CODE_ALIGN): Likewise.
1962 (MAX_CODE_ALIGN_VALUE): Likewise.
1964 2018-07-17 Martin Liska <mliska@suse.cz>
1966 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1967 in order to fulfil coding style.
1968 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1969 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1970 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1971 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1972 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1973 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1974 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1975 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1976 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1978 2018-07-17 Martin Liska <mliska@suse.cz>
1980 * align.h: New file.
1981 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1982 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1983 align_flags of label_to_alignment.
1984 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1986 * config/m68k/m68k.c: Do not use removed align_labels_value and
1988 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1989 (LOOP_ALIGN): Likewise.
1990 (LABEL_ALIGN): Likewise.
1991 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1992 Remove not used macro.
1993 (rs6000_loop_align): Change return type to align_flags.
1994 (rs6000_loop_align_max_skip): Remove.
1995 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1996 Change return type to align_flags.
1997 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1998 Remove not used macro.
1999 (rs6000_loop_align): Change return type to align_flags.
2000 (rs6000_loop_align_max_skip): Remove.
2001 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
2002 * config/rx/rx-protos.h (rx_align_for_label): Make it
2004 * config/rx/rx.c (rx_align_for_label): Change return type
2006 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
2008 into align_flags class.
2009 (LABEL_ALIGN): Likewise.
2010 (LOOP_ALIGN): Likewise.
2011 * config/s390/s390.c (s390_label_align): Use align_flags
2013 (s390_asm_output_function_label): Likewise.
2014 * config/sh/sh.c (sh_override_options_after_change):
2015 Use align_flags class directly without macros.
2016 (find_barrier): Likewise.
2017 (barrier_align): Likewise.
2018 (sh_loop_align): Likewise.
2019 * config/spu/spu.c (spu_option_override):
2020 Use align_flags_tuple::get_value instead of removed macros.
2021 (spu_sched_init): Likewise.
2022 * config/spu/spu.h (GTY): Likewise.
2023 * config/visium/visium.c (visium_option_override):
2024 Set "8" as default secondary alignment.
2025 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
2026 in order to guarantee secondary alignment of 8.
2027 * coretypes.h: Include align.h header file.
2028 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
2029 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
2030 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
2031 * doc/tm.texi.in: Likewise.
2032 * final.c (struct label_alignment): Remove not used structure.
2033 (LABEL_ALIGN): Change type to align_flags.
2034 (LOOP_ALIGN): Likewise.
2035 (JUMP_ALIGN): Likewise.
2036 (default_loop_align_max_skip): Remove.
2037 (default_label_align_max_skip): Likewise.
2038 (default_jump_align_max_skip): Likewise.
2039 (default_label_align_after_barrier_max_skip):
2040 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
2041 (LABEL_TO_MAX_SKIP): Remove.
2042 (label_to_alignment): Return align_flags type instead of integer.
2043 (label_to_max_skip): Remove.
2044 (align_fuzz): Use align_flags type.
2045 (compute_alignments): Use align_flags type and use align_flags::max
2046 to combine multiple alignments.
2047 (grow_label_align): Grow vec instead of C array.
2048 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
2049 (shorten_branches): Use align_flags type and use align_flags::max
2050 to combine multiple alignments.
2051 (final_scan_insn_1): Remove usage of secondary alignment that comes
2052 from label alignment, but instead use proper secondary alignment
2053 which is computed in grow_label_align.
2054 * flags.h (struct align_flags_tuple): Move to align.h.
2055 (struct align_flags): Likewise.
2056 (state_align_loops): Rename to align_loops.
2057 (state_align_jumps): Rename to align_jumps.
2058 (state_align_labels): Rename to align_labels.
2059 (state_align_functions): Rename to align_functions.
2060 (align_loops_log): Remove.
2061 (align_jumps_log): Remove.
2062 (align_labels_log): Remove.
2063 (align_functions_log): Remove.
2064 (align_loops_max_skip): Remove.
2065 (align_jumps_max_skip): Remove.
2066 (align_labels_max_skip): Remove.
2067 (align_functions_max_skip): Remove.
2068 (align_loops_value): Remove.
2069 (align_jumps_value): Remove.
2070 (align_labels_value): Remove.
2071 (align_functions_value): Remove.
2072 * output.h (label_to_alignment): Change return type to align_flags.
2073 (label_to_max_skip): Remove.
2074 * target.def: Remove loop_align_max_skip, label_align_max_skip,
2075 jump_align_max_skip macros.
2076 * targhooks.h (default_loop_align_max_skip): Remove.
2077 (default_label_align_max_skip): Likewise.
2078 (default_jump_align_max_skip): Likewise.
2079 (default_label_align_after_barrier_max_skip): Remove.
2080 * toplev.c (read_log_maxskip): Use ::normalize function.
2081 (parse_N_M): Remove not used argument and also call ::normalize.
2082 (parse_alignment_opts): Do not pass unused arguments.
2083 * varasm.c (assemble_start_function): Use directly align_functions
2084 instead of removed macros.
2085 * system.h: Do not poison removed macros.
2087 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2090 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
2091 and cond temporaries don't have reference type if iterator has
2092 pointer type. For init use &for_pre_body instead of pre_p if
2093 for_pre_body is non-empty.
2095 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2097 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
2098 double-double modes to SFmode directly directly.
2099 (trunc<mode>sf2_fprs): Delete.
2101 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2103 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
2104 for conversions between IFmode and the decimal floating point modes.
2105 (init_float128_ieee): Use the correct names for conversions between
2106 KFmode and the decimal floating point modes.
2108 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2110 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
2111 for the conversions between TDmode and IFmode.
2112 (init_float128_ieee): Use more correct names for the conversions
2113 between TDmode and KFmode.
2115 2018-07-16 Jakub Jelinek <jakub@redhat.com>
2117 PR tree-optimization/86526
2118 * builtins.c (expand_builtin_memcmp): Formatting fixes.
2119 (inline_expand_builtin_string_cmp): Likewise.
2120 (inline_string_cmp): Likewise. Use c_readstr instead of
2121 builtin_memcpy_read_str. Add unit_mode temporary.
2123 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2126 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
2127 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
2129 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2131 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
2132 Alphabetize prototypes of built-in functions, separating out
2133 built-in functions that are listed in this section but should be
2134 described elsewhere.
2136 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
2139 * expmed.c (emit_store_flag): Do not emit setcc followed by a
2140 conditional move when trapping comparison was split to a
2141 non-trapping one (and vice versa).
2143 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2145 * config/s390/s390.c (s390_function_profiler): Generate nops
2146 instead of profiler call sequences.
2147 * config/s390/s390.opt: Add the new option.
2149 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2151 * config/s390/s390.c (s390_function_profiler): Generate
2152 __mcount_loc section.
2153 * config/s390/s390.opt: Add the new option.
2155 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2157 * common.opt: Add the new warning.
2158 * config/s390/s390.c (s390_function_profiler): Emit "brasl
2159 %r0,__fentry__" when -mfentry is specified.
2160 (s390_option_override_internal): Disallow -mfentry for 31-bit
2162 * config/s390/s390.opt: Add the new option.
2164 2018-07-16 Richard Biener <rguenther@suse.de>
2167 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
2168 for function-local FUNCTION_DECL and RESULT_DECL immediately.
2170 2018-07-16 Martin Liska <mliska@suse.cz>
2173 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
2176 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2178 * config/arc/arcHS.md: Update ARCHS scheduling rules.
2180 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2182 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
2184 * config/arc/arc-cpus.def (hs4x): New cpu.
2186 * config/arc/arc-tables.opt: Regenerate.
2187 * config/arc/arc.c (arc_sched_issue_rate): New function.
2188 (TARGET_SCHED_ISSUE_RATE): Define.
2189 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
2190 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
2192 (attr tune): Add ARCHS4x tune values.
2193 (attr tune_dspmpy): Define.
2194 (*tst): Correct instruction type.
2195 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
2196 * config/arc/arcHS4x.md: New file.
2197 * config/arc/fpu.md: Update instruction type attributes.
2198 * config/arc/t-multilib: Regenerate.
2200 2018-07-16 Tom de Vries <tdevries@suse.de>
2203 * var-tracking.c (vt_initialize): Fix pre_dec handling.
2205 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2207 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
2210 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
2212 PR tree-optimization/86514
2213 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
2214 conversion to a boolean type from a type with greater precision.
2216 2018-07-16 Tom de Vries <tdevries@suse.de>
2218 * var-tracking.c (vt_initialize): Print adjusted insn slim if
2219 dump_flags request TDF_SLIM.
2221 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
2223 * fold-const.c (int_const_binop_1): Abstract...
2224 (wide_int_binop): ...wide int code here.
2225 (poly_int_binop): ...poly int code here.
2226 Abstract the rest of int_const_binop_1 into int_const_binop.
2227 * fold-const.h (wide_int_binop): New.
2228 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
2229 Remove useless PLUS/MINUS_EXPR case.
2230 (zero_nonzero_bits_from_vr): Move wide int code...
2231 (zero_nonzero_bits_from_bounds): ...here.
2232 (extract_range_from_binary_expr_1): Move mask optimization code...
2233 (range_easy_mask_min_max): ...here.
2234 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
2235 (range_easy_mask_min_max): New.
2237 2018-07-15 Jeff Law <law@redhat.com>
2240 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
2243 2018-07-14 Jim Wilson <jimw@sifive.com>
2245 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
2247 2018-07-14 Paul Koning <ni1d@arrl.net>
2249 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
2251 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
2253 * lto-streamer-out.c (copy_function_or_variable): Dump info about
2256 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
2257 Steve Munroe <munroesj52@gmail.com>
2259 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
2260 (_mm_andnot_si128): Likewise.
2261 (_mm_or_si128): Likewise.
2262 (_mm_xor_si128): Likewise.
2264 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
2267 * builtins.c (expand_builtin_memcmp): Inline the calls first
2268 when result_eq is false.
2269 (expand_builtin_strcmp): Inline the calls first.
2270 (expand_builtin_strncmp): Likewise.
2271 (inline_string_cmp): New routine. Expand a string compare
2272 call by using a sequence of char comparison.
2273 (inline_expand_builtin_string_cmp): New routine. Inline expansion
2274 a call to str(n)cmp/memcmp.
2275 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
2277 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
2279 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
2281 * config/arm/driver-arm.c: Include arm-native.h.
2282 (host_detect_local_cpu): Use auto-generated data tables.
2283 (vendors, arm_cpu_table): Delete. Move part information to ...
2284 * config/arm/arm-cpus.in: ... here.
2285 * config/arm/parsecpu.awk (gen_native): New function.
2286 (vendor, part): New CPU fields.
2287 (END): Add support for building the native CPU detection tables.
2288 * config/arm/t-arm (arm-native.h): Add build rule.
2289 (driver-arm.o): Add dependency on arm-native.h.
2291 2018-07-13 Richard Biener <rguenther@suse.de>
2294 * match.pd (addr1 - addr2): Allow either of the operand to
2297 2018-07-13 Tom de Vries <tdevries@suse.de>
2299 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
2302 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
2304 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
2305 arrays instead of numbered variables.
2307 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
2309 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
2310 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
2312 2018-07-13 Richard Biener <rguenther@suse.de>
2315 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
2316 instead of get_context_die.
2318 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
2319 Richard Biener <rguenther@suse.de>
2322 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
2323 that the loop latch destination where phi is defined.
2325 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
2327 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
2328 (riscv_expand_epilogue): Add assertion to check interrupt mode.
2329 (riscv_set_current_function): Extract getting interrupt type to new
2331 (riscv_get_interrupt_type): New function.
2332 (riscv_merge_decl_attributes): New function, checking interrupt type is
2334 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
2336 2018-07-12 Paul Koning <ni1d@arrl.net>
2338 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
2341 2018-07-12 Paul Koning <ni1d@arrl.net>
2343 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
2344 zero reference, add doloop_end instead.
2345 * doc/md.texi (decrement_and_branch_until_zero): Remove.
2346 (Looping patterns): Remove decrement_and_branch_until_zero. Add
2347 detail for doloop_end.
2349 2018-07-12 Martin Sebor <msebor@redhat.com>
2352 * attribs.c (decl_attributes): Reject conflicting attributes before
2353 calling attribute handlers.
2355 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2357 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
2359 (gcc::dump_manager::get_dump_file_name): likewise.
2360 (dump_begin): Likewise.
2361 * dumpfile.h (dump_begin): Update prototype.
2362 (gcc::dump_manager::get_dump_file_name,
2363 gcc::dump_manager::get_dump_file_name): Update prototype.
2365 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2367 * internal-fn.h (vectorizable_internal_fn_p): New function.
2368 * tree-vect-slp.c (compatible_calls_p): Likewise.
2369 (vect_build_slp_tree_1): Remove nops argument. Handle calls
2370 to internal functions.
2371 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
2373 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2375 * fold-const.h (inverse_conditions_p): Declare.
2376 * fold-const.c (inverse_conditions_p): New function.
2377 * match.pd: Use inverse_conditions_p. Add folds of view_converts
2378 that test the inverse condition of a conditional internal function.
2379 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
2380 * internal-fn.c (internal_fn_mask_index): Handle conditional
2382 (vectorized_internal_fn_supported_p): New function.
2383 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
2384 (any_pred_load_store): Replace with...
2385 (need_to_predicate): ...this new variable.
2386 (redundant_ssa_names): New variable.
2387 (ifcvt_can_use_mask_load_store): Move initial checks to...
2388 (ifcvt_can_predicate): ...this new function. Handle tree codes
2389 for which a conditional internal function exists.
2390 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
2391 instead of ifcvt_can_use_mask_load_store. Update after variable
2393 (predicate_load_or_store): New function, split out from
2394 predicate_mem_writes.
2395 (check_redundant_cond_expr): New function.
2396 (value_available_p): Likewise.
2397 (predicate_rhs_code): Likewise.
2398 (predicate_mem_writes): Rename to...
2399 (predicate_statements): ...this. Use predicate_load_or_store
2400 and predicate_rhs_code.
2401 (combine_blocks, tree_if_conversion): Update after above name changes.
2402 (ifcvt_local_dce): Handle redundant_ssa_names.
2403 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
2404 general conditional functions.
2405 * tree-vect-stmts.c (vectorizable_call): Likewise.
2407 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2408 Alan Hayward <alan.hayward@arm.com>
2409 David Sherwood <david.sherwood@arm.com>
2411 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
2412 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
2413 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
2414 plus and minus and convert them into IFN_COND_FMA-based sequences.
2415 (convert_mult_to_fma): Handle conditional plus and minus.
2417 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2419 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
2420 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
2421 (cond_fnms_optab): New optabs.
2422 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
2424 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
2425 * internal-fn.h (get_conditional_internal_fn): Declare.
2426 (get_unconditional_internal_fn): Likewise.
2427 * internal-fn.c (cond_ternary_direct): New macro.
2428 (expand_cond_ternary_optab_fn): Likewise.
2429 (direct_cond_ternary_optab_supported_p): Likewise.
2430 (FOR_EACH_COND_FN_PAIR): Likewise.
2431 (get_conditional_internal_fn): New function.
2432 (get_unconditional_internal_fn): Likewise.
2433 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
2434 (gimple_match_op::gimple_match_op): Add a new overload for 5
2436 (gimple_match_op::set_op): Likewise.
2437 (gimple_resimplify5): Declare.
2438 * genmatch.c (decision_tree::gen): Generate simplifications for
2440 * gimple-match-head.c (gimple_simplify): Define an overload for
2441 5 operands. Handle calls with 5 arguments in the top-level overload.
2442 (convert_conditional_op): Handle conversions from unconditional
2443 internal functions to conditional ones.
2444 (gimple_resimplify5): New function.
2445 (build_call_internal): Pass a fifth operand.
2446 (maybe_push_res_to_seq): Likewise.
2447 (try_conditional_simplification): Try converting conditional
2448 internal functions to unconditional internal functions.
2449 Handle 3-operand unconditional forms.
2450 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
2451 Define ternary equivalents of the current rules for binary conditional
2453 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
2455 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
2456 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
2457 (optab): Handle them.
2458 (SVE_COND_FP_TERNARY): New int iterator.
2459 (sve_fmla_op, sve_fmad_op): New int attributes.
2460 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
2461 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
2462 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
2464 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2466 * target.def (preferred_else_value): New target hook.
2467 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
2468 * doc/tm.texi: Regenerate.
2469 * targhooks.h (default_preferred_else_value): Declare.
2470 * targhooks.c (default_preferred_else_value): New function.
2471 * internal-fn.h (conditional_internal_fn_code): Declare.
2472 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
2473 (get_conditional_internal_fn): Use it.
2474 (conditional_internal_fn_code): New function.
2475 * gimple-match.h (gimple_match_cond): New struct.
2476 (gimple_match_op): Add a cond member function.
2477 (gimple_match_op::gimple_match_op): Update all forms to take a
2479 * genmatch.c (expr::gen_transform): Use the same condition as res_op
2480 for the suboperation, but don't specify a particular else_value.
2481 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
2482 (visit_nary_op, visit_reference_op_load): Pass
2483 gimple_match_cond::UNCOND to the gimple_match_op constructor.
2484 * gimple-match-head.c: Include tree-eh.h
2485 (convert_conditional_op): New function.
2486 (maybe_resimplify_conditional_op): Likewise.
2487 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
2488 (gimple_resimplify2): Likewise.
2489 (gimple_resimplify3): Likewise.
2490 (gimple_resimplify4): Likewise.
2491 (maybe_push_res_to_seq): Return null for conditional operations.
2492 (try_conditional_simplification): New function.
2493 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
2495 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
2497 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
2499 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
2501 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2503 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2505 (hash_tree): Do not hash DECL_FCONTEXT
2506 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
2507 Do not stream DECL_FCONTEXT.
2508 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
2509 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
2511 2018-07-12 Richard Biener <rguenther@suse.de>
2514 * dwarf2out.c (gen_block_die): Only output blocks when they have
2515 at least one !DECL_IGNORED_P variable.
2517 2018-07-12 Richard Biener <rguenther@suse.de>
2520 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2521 Remove -mieee-fp handling.
2523 2018-07-12 Richard Biener <rguenther@suse.de>
2525 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
2526 left-over from last patch.
2528 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2530 PR tree-optimization/86492
2531 * gimple-ssa-store-merging.c
2532 (imm_store_chain_info::coalesce_immediate_stores): Call
2533 check_no_overlap even for the merge_overlapping case. Formatting fix.
2535 2018-07-12 Richard Biener <rguenther@suse.de>
2538 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
2539 move possibly trapping operations into the conditional.
2541 2018-07-12 Richard Biener <rguenther@suse.de>
2543 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2544 (vn_lookup_simplify_result): Remove recursion limit applied
2546 (vn_nary_build_or_lookup_1): Adjust.
2547 (try_to_simplify): Likewise.
2548 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2550 (gimple_resimplify2): Likewise.
2551 (gimple_resimplify3): Likewise.
2552 (gimple_resimplify4): Likewise.
2554 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2556 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2557 Use __mmask64 type instead of __mmask8 for __M argument.
2558 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2559 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2561 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2562 __mmask16 for __M argument.
2563 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2564 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2565 to __mmask16 instead of __mmask8.
2566 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2567 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2568 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2569 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2570 instead of __mmask16 for __U argument.
2571 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2572 __mmask16 instead of __mmask8 for __U argument.
2573 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2575 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2577 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2579 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2581 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2583 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2585 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2587 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2589 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2591 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2593 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2594 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
2595 return type as well as __M argument type and all casts from __mmask8
2597 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
2598 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
2599 return type as well as __M argument type and all casts from __mmask8
2601 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
2602 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
2603 return type as well as __M argument type and all casts from __mmask8
2605 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
2606 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
2607 return type as well as __M argument type and all casts from __mmask8
2609 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
2610 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
2613 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
2615 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
2616 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
2619 2018-07-11 Paul Koning <ni1d@arrl.net>
2621 * doc/md.texi (define_subst): Document how multiple occurrences of
2622 the same argument in the replacement pattern are handled.
2624 2018-07-11 Paul Koning <ni1d@arrl.net>
2626 * doc/extend.texi (Common Variable Attributes): Move "mode" into
2628 (Common Type Attributes): Add "mode" attribute.
2630 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
2632 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
2633 stream DECL_ORIGINAL_TYPE.
2634 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
2635 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
2636 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
2637 Do not walk original type.
2638 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
2640 (write_ts_decl_non_common_tree_pointers): Do not stream
2642 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
2643 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
2645 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2647 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
2648 thread twice from the same starting edge.
2650 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2652 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
2653 * gimple.c (gimple_call_nonnull_result_p): ...here...
2654 (gimple_call_nonnull_arg): ...and here.
2655 * gimple.h (gimple_call_nonnull_result_p): New.
2656 (gimple_call_nonnull_arg): New.
2658 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
2660 * config/arm/arm-cpus.in: Move information from fpu field of each
2661 cpu definition to the isa field.
2662 * config/arm/parsecpu.awk (fpu): Delete match rule.
2663 (gen_comm_data): Don't add bits from the CPU's FPU entry.
2665 2018-07-11 Richard Biener <rguenther@suse.de>
2668 * dwarf2out.c (init_sections_and_labels): Use
2669 output_asm_line_debug_info consistently.
2670 (dwarf2out_early_finish): Likewise.
2671 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
2674 2018-07-11 Richard Biener <rguenther@suse.de>
2676 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
2677 Rework father_bb setting in a way to avoid propagating constants
2678 multiple times on a loop body.
2680 2018-07-10 Mark Wielaard <mark@klomp.org>
2683 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
2685 2018-07-10 Richard Biener <rguenther@suse.de>
2687 * hash-map.h (hash_map::iterator::operator*): Return
2688 references to key and value.
2690 2018-07-10 Jakub Jelinek <jakub@redhat.com>
2693 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
2694 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
2695 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
2696 (gimplify_omp_for): For composite loops, move outer
2697 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
2698 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
2699 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
2700 TREE_LIST for both the original class iterator and the "last" helper
2701 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
2702 loop, remember has_decl_expr from outer composite loops for the
2703 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
2705 2018-07-09 Martin Sebor <msebor@redhat.com>
2709 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
2710 accessing implicitly initialized array elements.
2711 * expr.c (string_constant): Handle string initializers of
2712 character arrays within aggregates.
2713 * gimple-fold.c (fold_array_ctor_reference): Add argument.
2714 Store element offset. As a special case, handle zero size.
2715 (fold_nonarray_ctor_reference): Same.
2716 (fold_ctor_reference): Add argument. Store subobject offset.
2717 * gimple-fold.h (fold_ctor_reference): Add argument.
2719 2018-07-09 Paul Koning <ni1d@arrl.net>
2721 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
2722 (pdp11_insn_cost): New function.
2723 (pdp11_md_asm_adjust): New function.
2724 (TARGET_INVALID_WITHIN_DOLOOP): Define.
2725 (pdp11_rtx_costs): Update to match machine better.
2726 (output_addr_const_pdp11): Correct format mismatch warnings.
2727 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
2728 * config/pdp11/pdp11.md: General change to add base_cost and/or
2729 length attributes for use by new pdp11_insn_cost function.
2730 (MIN_BRANCH): Correct definition.
2732 (doloop_end): Use standard pattern name for looping pattern.
2733 (doloop_end_nocc): New.
2734 (movsf): Add another constraint alternative.
2735 (zero_extendqihi2): Add constraint alternatives for not in place
2737 (zero_extendhisi2): Remove.
2738 (shift patterns): Add CC handling variants.
2742 (define_peephole2): New peephole to recognize mov that sets CC for
2745 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2747 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
2748 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
2751 2018-07-09 Jakub Jelinek <jakub@redhat.com>
2754 * real.c (real_nextafter): Return true if result is denormal.
2756 2018-07-09 Martin Liska <mliska@suse.cz>
2758 * common.opt: Add back wrongly removed attribute.
2760 2018-07-09 Richard Biener <rguenther@suse.de>
2763 * dwarf2out.c (gen_block_die): For an early generated DIE
2764 always output high/low PC attributes.
2766 2018-07-09 Tom de Vries <tdevries@suse.de>
2768 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2769 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2770 onto VAR_DECL with abstract origin.
2772 2018-07-07 Jim Wilson <jimw@sifive.com>
2774 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2776 2018-07-07 Tom de Vries <tdevries@suse.de>
2778 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2780 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2782 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2784 (combine_bound): Use wide-int overflow calculation instead of
2786 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2788 * fold-const.c (int_const_binop_2): Same.
2789 (extract_muldiv_1): Same.
2790 (fold_div_compare): Same.
2791 (fold_abs_const): Same.
2793 * poly-int.h (add): Same.
2797 * predict.c (predict_iv_comparison): Same.
2798 * profile-count.c (slow_safe_scale_64bit): Same.
2799 * simplify-rtx.c (simplify_const_binary_operation): Same.
2800 * tree-chrec.c (tree_fold_binomial): Same.
2801 * tree-data-ref.c (split_constant_offset_1): Same.
2802 * tree-if-conv.c (idx_within_array_bound): Same.
2803 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2804 * tree-ssa-phiopt.c (minmax_replacement): Same.
2805 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2806 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2807 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2808 * wide-int.cc (wi::add_large): Same.
2809 (wi::mul_internal): Same.
2810 (wi::sub_large): Same.
2811 (wi::divmod_internal): Same.
2812 * wide-int.h: Change overflow type to overflow_type for neg, add,
2813 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2814 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2815 mul_internal, divmod_internal.
2816 (overflow_type): New enum.
2817 (accumulate_overflow): New.
2819 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2821 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2822 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2824 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2826 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2827 argument is checked for zero before entering loop, avoid checking again.
2829 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2831 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2832 * gimplify.h (generic_expr_could_trap_p): Likwise.
2833 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2835 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2837 PR tree-optimization/86401
2838 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2839 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2840 (fold_bit_and_mask): ... here. New helper function for match.pd.
2841 * fold-const.h (fold_bit_and_mask): Declare.
2842 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2844 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2847 * target.def (translate_mode_attribute): New hook.
2848 * targhooks.h (default_translate_mode_attribute): Declare.
2849 * targhooks.c (default_translate_mode_attribute): New function.
2850 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2851 * doc/tm.texi: Regenerate.
2852 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2853 (rs6000_translate_mode_attribute): New function.
2855 2018-07-06 Paul Koning <ni1d@arrl.net>
2857 * doc/md.texi (define_split): Document DONE and FAIL.
2858 (define_peephole2): Ditto.
2860 2018-07-05 Jeff Law <law@redhat.com>
2862 PR tree-optimization/86010
2863 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2864 both the head and tail of mem* and str* calls.
2866 2018-07-05 Jim Wilson <jimw@sifive.com>
2868 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2870 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2872 * config/aarch64/aarch64-simd.md: correct flags text for
2873 MIN_EXPR replacement.
2875 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2877 * configure: Regenerated.
2879 2018-07-05 Carl Love <cel@us.ibm.com>
2881 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2882 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2883 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2884 VSX_BUILTIN_DOUBLEL_V4SF.
2886 2018-07-05 Martin Sebor <msebor@redhat.com>
2889 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2890 than its domain to compute its the upper bound of a char array.
2892 2018-07-05 Nathan Sidwell <nathan@acm.org>
2894 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2895 * doc/cpp.texi: Update comment.
2896 * doc/tm.texi: Rebuilt.
2897 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2898 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2899 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2900 headers affected by extern "C".
2901 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2902 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2903 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2904 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2905 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2906 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2907 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2908 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2909 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2910 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2911 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2912 config/sparc/sp64-elf.h, config/spu/spu.h,
2913 config/stormy16/stormy16.h, config/v850/v850.h,
2914 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2915 define NO_IMPLICIT_EXTERN_C.
2916 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2918 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2921 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2922 instead of GET_MODE_SIZE when comparing Units.
2924 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2927 * rtlanal.c (set_noop_p): Constrain on mode change,
2928 include hard-reg-set.h
2930 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2932 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2934 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2937 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2940 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2942 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2944 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2947 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2949 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2951 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2954 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2956 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2958 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2959 neg_*_op* variables.
2961 2018-07-04 Martin Liska <mliska@suse.cz>
2963 * tree-switch-conversion.c: Define
2964 max_ratio_for_speed and max_ratio_for_size constants.
2966 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2967 Martin Liska <mliska@suse.cz>
2972 * common.opt: Rename align options with 'str_' prefix.
2973 * common/config/i386/i386-common.c (set_malign_value): New
2975 (ix86_handle_option): Use it to set -falign-* options/
2976 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2977 type from int to string.
2978 * config/aarch64/aarch64.c: Update default values from int
2980 * config/alpha/alpha.c (alpha_override_options_after_change):
2982 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2983 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2984 max skip conditionally.
2985 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2986 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2987 max skip conditionally.
2988 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2989 max skip conditionally.
2990 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2991 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2992 max skip conditionally.
2993 * config/i386/i386.c (struct ptt): Change type from int to
2995 (ix86_default_align): Set default values.
2996 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2997 max skip conditionally.
2998 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2999 (ASM_OUTPUT_MAX_SKIP_ALIGN):
3000 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
3001 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3002 max skip conditionally.
3003 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
3004 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
3005 * config/i386/x86-64.h (SUBALIGN_LOG): New.
3006 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3007 max skip conditionally.
3008 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3009 * config/ia64/ia64.c (ia64_option_override): Set default values
3010 for alignment options.
3011 * config/m68k/m68k.c: Handle new str_align_* options.
3012 * config/mips/mips.c (mips_set_compression_mode): Change
3014 (mips_option_override): Set default values for options.
3015 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3017 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3019 * config/rx/rx.c (rx_option_override): Likewise.
3020 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
3021 (LABEL_ALIGN): Use align_labels_log.
3022 (LOOP_ALIGN): Use align_loops_align.
3023 * config/s390/s390.c (s390_asm_output_function_label): Use new
3025 * config/sh/sh.c (sh_override_options_after_change):
3026 Change type of constants.
3027 * config/spu/spu.c (spu_sched_init): Likewise.
3028 * config/sparc/sparc.c (sparc_option_override): Set default
3030 * config/visium/visium.c (visium_option_override): Likewise.
3031 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
3032 emit p2align format with last argument if it's not needed.
3033 * doc/invoke.texi: Document extended format of -falign-*.
3034 * final.c: Use align_labels alignment.
3035 * flags.h (struct target_flag_state): Change type to use
3037 (struct align_flags_tuple): New.
3038 (struct align_flags): Likewise.
3039 (align_loops_log): Redefine macro to use new types.
3040 (align_loops_max_skip): Redefine macro to use new types.
3041 (align_jumps_log): Redefine macro to use new types.
3042 (align_jumps_max_skip): Redefine macro to use new types.
3043 (align_labels_log): Redefine macro to use new types.
3044 (align_labels_max_skip): Redefine macro to use new types.
3045 (align_functions_log): Redefine macro to use new types.
3046 (align_loops): Redefine macro to use new types.
3047 (align_jumps): Redefine macro to use new types.
3048 (align_labels): Redefine macro to use new types.
3049 (align_functions): Redefine macro to use new types.
3050 (align_functions_max_skip): Redefine macro to use new types.
3051 (align_loops_value): New macro.
3052 (align_jumps_value): New macro.
3053 (align_labels_value): New macro.
3054 (align_functions_value): New macro.
3055 * function.c (invoke_set_current_function_hook): Propagate
3056 alignment values from flags to global variables default in
3058 * ipa-icf.c (sem_function::equals_wpa): Use
3059 cl_optimization_option_eq instead of memcmp.
3060 * lto-streamer.h (cl_optimization_stream_out): Support streaming
3062 (cl_optimization_stream_in): Likewise.
3063 * optc-save-gen.awk: Support strings in cl_optimization.
3064 * opth-gen.awk: Likewise.
3065 * opts.c (finish_options): Remove error checking of invalid
3067 (MAX_CODE_ALIGN): Remove.
3068 (MAX_CODE_ALIGN_VALUE): Likewise.
3069 (parse_and_check_align_values): New function.
3070 (check_alignment_argument): Likewise.
3071 (common_handle_option): Use check_alignment_argument.
3072 * opts.h (parse_and_check_align_values): Declare.
3073 * toplev.c (init_alignments): Remove.
3074 (read_log_maxskip): New.
3075 (parse_N_M): Likewise.
3076 (parse_alignment_opts): Likewise.
3077 (backend_init_target): Remove usage of init_alignments.
3078 * toplev.h (parse_alignment_opts): Declare.
3079 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
3081 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
3082 * tree.c (cl_option_hasher::equal): New.
3083 * varasm.c: Use new global macros.
3085 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
3087 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
3088 Use a simpler align directive also if MAXSKIP = ALIGN-1.
3089 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3090 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3091 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3092 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3093 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
3094 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
3095 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
3096 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3097 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3098 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3099 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3101 2018-07-04 Martin Liska <mliska@suse.cz>
3102 Jonathan Wakely <jwakely@redhat.com>
3104 * coverage.c: Use correct type.
3105 * doc/invoke.texi: Language correction.
3107 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
3110 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
3111 ENDBRANCH for non-tail call which may return via indirect branch.
3112 * doc/extend.texi: Document indirect_return attribute.
3114 2018-07-03 Martin Sebor <msebor@redhat.com>
3116 PR tree-optimization/86274
3117 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
3119 (format_floating): Correct handling of infinities and NaNs.
3121 2018-07-03 Martin Sebor <msebor@redhat.com>
3123 * print-tree.c (print_real_cst): New function.
3124 (print_node_brief): Call it.
3125 (print_node): Ditto.
3127 2018-07-03 Jeff Law <law@redhat.com>
3129 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
3130 into a single pattern.
3132 * config/h8300/h8300.md (ors code_iterator): New.
3133 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
3134 a single pattern and single splitter.
3135 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
3136 (iorqi3_1, xorqi3_1): Likewise.
3137 (iorqi3, xorqi3 expanders): Similarly.
3139 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
3140 (movmd_internal) into a single pattern using the P mode iterator.
3141 (movmd splitters): Similarly.
3142 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
3143 (movsd splitters): Similarly.
3145 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
3146 ADDB, ADDW and ADDL into a single ADD attribute which selects the
3147 right table based on the size of the operand.
3148 * config/h8300/h8300.md (length_table): Corresponding changes. All
3149 references to "addb", "addw" and "addl" changed to "add".
3150 (btst patterns): Merge two variants into a single pattern.
3151 (tstqi, tsthi): Likewise.
3152 (addhi3_incdec, addsi3_incdec): Likewise.
3153 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
3154 (mulhi3, mulsi3): Likewise.
3155 (udivhi3, udivsi3): Likewise.
3156 (divhi3, divsi3): Likewise.
3157 (andorqi3, andorhi3, andorsi3): Likewise.
3159 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
3162 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
3163 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
3165 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3167 PR tree-optimization/85694
3168 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
3169 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
3171 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
3172 (<u>avg<mode>3_ceil): New patterns.
3174 2018-07-03 David Malcolm <dmalcolm@redhat.com>
3176 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
3177 scan-tree-dump directive.
3178 * gcc.dg/vect/slp-perm-2.c: Likewise.
3179 * gcc.dg/vect/slp-perm-3.c: Likewise.
3180 * gcc.dg/vect/slp-perm-5.c: Likewise.
3181 * gcc.dg/vect/slp-perm-6.c: Likewise.
3182 * gcc.dg/vect/slp-perm-7.c: Likewise.
3183 * gcc.dg/vect/slp-perm-8.c: Likewise.
3185 2018-07-03 Marek Polacek <polacek@redhat.com>
3188 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
3190 2018-07-03 Richard Biener <rguenther@suse.de>
3193 * tree-ssa-structalias.c (find_func_clobbers): Properly
3194 handle indirect calls.
3196 2018-07-03 Jeff Law <law@redhat.com>
3198 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
3199 (shifts): New code iterator.
3200 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
3201 expander. Fix HImode handling on H8/SX.
3202 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
3203 (subqi3, subhi3, subsi3 expanders): Likewise.
3204 (andqi3, andhi3, andsi3 expanders): Likewise.
3205 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
3206 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
3207 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
3208 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
3209 (zero_extendqihi2, zero_extendqisi2): Likewise.
3210 (extendqihi2, extendqisi2): Likewise.
3211 (rotlqi3, rotlhi3, rotlsi3): Likewise.
3212 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
3213 (rotlqi3_1, rotlhi3_1): Likewise.
3214 (logicalhi3_sn, logicalsi3_sn): Likewise.
3215 (logicalhi3, logicalsi3): Likewise.
3217 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3219 * tree-vect-patterns.c (vect_recog_rotate_pattern)
3220 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3221 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
3222 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
3223 type to append_pattern_def_seq instead of creating a stmt_vec_info
3225 (build_mask_conversion): Likewise. Remove vinfo argument.
3226 (vect_add_conversion_to_patterm): Likewise, renaming to...
3227 (vect_add_conversion_to_pattern): ...this.
3228 (vect_recog_mask_conversion_pattern): Update call to
3229 build_mask_conversion. Pass the vector type to
3230 append_pattern_def_seq here too.
3231 (vect_recog_gather_scatter_pattern): Update call to
3232 vect_add_conversion_to_pattern.
3234 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3236 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
3237 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3238 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
3239 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
3240 STMT_VINFO_PATTERN_DEF_SEQ to null here.
3241 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
3242 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
3243 append_pattern_def_seq instead of new_pattern_def_seq.
3244 (vect_recog_divmod_pattern): Do both of the above.
3245 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
3248 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3250 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
3251 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
3252 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
3253 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
3254 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
3255 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3256 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
3257 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3258 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3259 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
3260 parameter with a single stmt_vec_info.
3261 (vect_recog_func_ptr): Likewise.
3262 (vect_recog_gather_scatter_pattern): Likewise, folding in...
3263 (vect_try_gather_scatter_pattern): ...this.
3264 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
3265 the stmt_vec_info of the statement to be matched. Don't clear
3266 STMT_VINFO_RELATED_STMT.
3267 (vect_pattern_recog): Update call accordingly.
3269 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3271 PR tree-optimization/85694
3272 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
3273 (uavgM3_ceil): Document new optabs.
3274 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
3275 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
3277 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
3278 (savg_ceil_optab): New optabs.
3279 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
3280 (vect_vect_recog_func_ptrs): Add it.
3281 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
3282 constant directly from the associated lhs.
3284 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3286 * tree-vect-patterns.c (vect_split_statement): New function.
3287 (vect_convert_input): Use it to try to split an existing cast.
3289 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3291 * poly-int.h (print_hex): New function.
3292 * dumpfile.h (dump_dec, dump_hex): Declare.
3293 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
3294 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
3295 min_input_precision, operation_precision and operation_sign.
3296 * tree-vect-patterns.c (vect_get_range_info): New function.
3297 (vect_same_loop_or_bb_p, vect_single_imm_use)
3298 (vect_operation_fits_smaller_type): Delete.
3299 (vect_look_through_possible_promotion): Add an optional
3300 single_use_p parameter.
3301 (vect_recog_over_widening_pattern): Rewrite to use new
3302 stmt_vec_info infomration. Handle one operation at a time.
3303 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
3304 (vect_truncatable_operation_p, vect_set_operation_type)
3305 (vect_set_min_input_precision): New functions.
3306 (vect_determine_min_output_precision_1): Likewise.
3307 (vect_determine_min_output_precision): Likewise.
3308 (vect_determine_precisions_from_range): Likewise.
3309 (vect_determine_precisions_from_users): Likewise.
3310 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
3311 (vect_vect_recog_func_ptrs): Put over_widening first.
3313 (vect_pattern_recog): Call vect_determine_precisions.
3315 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3317 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
3318 statements that have been replaced by further pattern statements.
3319 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
3321 2018-07-03 Richard Biener <rguenther@suse.de>
3323 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
3324 always set *dt. Dump vectype in vectype overload.
3325 * dumpfile.h (dump_gimple_expr): New function.
3326 (dump_gimple_expr_loc): Likewise.
3327 * dumpfile.c (dump_gimple_expr): New function.
3328 (dump_gimple_expr_loc): Likewise.
3330 2018-07-02 Jeff Law <law@redhat.com>
3332 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
3333 the H8/300, H8/300H and H8/S variants into a single pattern.
3334 (movhi_h8300, movqi_h8300hs): Similarly.
3335 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
3336 (QHI mode iterator): New.
3338 * config/h8300/h8300.md: Remove trailing whitespace.
3340 2018-07-02 Jim Wilson <jimw@sifive.com>
3342 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
3343 instead of emit_insn for interrupt returns.
3344 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
3345 (riscv_sret, riscv_uret): Likewise.
3347 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3349 * pretty-print.c (selftest::test_pp_format): Move save and restore
3350 of quotes to class auto_fix_quotes, and add an instance.
3351 * selftest.c: Include "intl.h".
3352 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
3353 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
3354 * selftest.h (selftest::auto_fix_quotes): New class.
3356 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
3358 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
3359 (aarch64_sve_prepare_conditional_op): Remove.
3360 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
3361 Allow aarch64_simd_reg_or_zero as select operand; remove
3362 the aarch64_sve_prepare_conditional_op call.
3363 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
3364 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3365 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
3366 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
3367 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
3368 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
3369 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
3370 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
3371 and a splitters to match all of the *_any patterns.
3372 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
3374 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
3375 (SVE_COND_FP_BINARY_REV): Remove.
3376 (sve_int_op_rev, sve_fp_op_rev): New.
3377 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
3378 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
3379 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
3380 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
3381 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
3382 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
3383 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
3384 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
3385 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
3387 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
3388 Remove match_dup 1 from the inner unspec.
3389 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3391 * config/aarch64/aarch64.md (movprfx): New attr.
3392 (length): Default movprfx to 8.
3393 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
3394 (*madd<SVE_I>, *msub<SVE_I): Likewise.
3395 (*<su>mul<SVE_I>3_highpart): Likewise.
3396 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
3397 (*v<ASHIFT><SVE_I>3): Likewise.
3398 (*<su><MAXMIN><SVE_I>3): Likewise.
3399 (*<su><MAXMIN><SVE_F>3): Likewise.
3400 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
3401 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
3402 (*div<SVE_F>4): Likewise.
3404 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
3406 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
3409 2018-07-02 Richard Biener <rguenther@suse.de>
3411 PR tree-optimization/86363
3412 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
3413 memset argument refers to a non-variable address.
3415 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3417 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
3418 {PLUS,MINUS}_EXPR code to...
3419 (adjust_symbolic_bound): ...here,
3420 (combine_bound): ...here,
3421 (set_value_range_with_overflow): ...and here.
3423 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3425 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
3427 (extract_range_from_abs_expr): ...here.
3429 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3431 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3432 -fno-omit-frame-pointer when not optimizing.
3434 2018-07-02 Martin Liska <mliska@suse.cz>
3437 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
3438 (propagate_nothrow): Likewise.
3440 2018-07-02 Martin Liska <mliska@suse.cz>
3443 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
3445 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3447 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
3448 function in r262149, changing "loc" param from source_location to
3449 const dump_location_t &.
3450 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
3451 declaration, as above.
3453 2018-07-01 Paul Koning <ni1d@arrl.net>
3455 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3456 -munit-asm, -mgnu-asm, -mdec-asm.
3457 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
3458 (pdp11_output_labelref): New.
3459 (pdp11_output_def): New.
3460 (pdp11_output_addr_vec_elt): New.
3461 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
3462 %# and %@ format codes.
3463 (pdp11_option_override): New.
3464 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
3465 (pdp11_output_ident): New.
3466 (pdp11_asm_named_section): New.
3467 (pdp11_asm_init_sections): New.
3468 (pdp11_file_start): New.
3469 (pdp11_file_end): New.
3470 (output_ascii): Use .ascii/.asciz for -mdec-asm.
3471 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
3472 %o, like %c but octal.
3473 (pdp11_option_override): New.
3474 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
3476 (DATA_SECTION_ASM_OP): Ditto.
3477 (READONLY_DATA_SECTION_ASM_OP): New.
3478 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
3479 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
3480 (ASM_OUTPUT_LABELREF): Ditto.
3481 (ASM_OUTPUT_DEF): Ditto.
3482 (ASM_OUTPUT_EXTERNAL): New.
3483 (ASM_OUTPUT_SOURCE_FILENAME): New.
3484 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
3485 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
3486 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
3487 %# and %@ format codes.
3488 * config/pdp11/pdp11.opt (mgnu-asm): New.
3489 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
3490 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
3491 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
3493 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
3495 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
3496 dereferencing path[] beyond its length.
3498 (debug_all_paths): New.
3499 (rewire_first_differing_edge): New.
3500 (adjust_paths_after_duplication): New.
3501 (duplicate_thread_path): Call adjust_paths_after_duplication.
3503 (thread_through_all_blocks): Add new argument to
3504 duplicate_thread_path.
3506 2018-06-30 Jim Wilson <jimw@sifive.com>
3508 * config/riscv/predicates.md (p2m1_shift_operand): New.
3509 (high_mask_shift_operand): New.
3510 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
3511 pattern using p2m1_shift_operand.
3512 (lshsi3_zero_extend_3+2): New combiner pattern using
3513 high_mask_shift_operand.
3515 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3517 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
3519 (vect_recog_rotate_pattern): ...here.
3520 (vect_convert_input): Try to insert casts of invariants in the
3522 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
3523 preheader to be empty.
3525 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3527 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
3528 vector type. If given, install it in the new statement's
3530 (vect_element_precision): New function.
3531 (vect_unpromoted_value): New struct.
3532 (vect_unpromoted_value::vect_unpromoted_value): New function.
3533 (vect_unpromoted_value::set_op): Likewise.
3534 (vect_look_through_possible_promotion): Likewise.
3535 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
3536 (vect_widened_op_tree, vect_convert_input): Likewise.
3537 (vect_convert_inputs, vect_convert_output): Likewise.
3538 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
3539 to handle the optional cast of the multiplication result and
3540 vect_widened_op_tree to detect the widened multiplication itself.
3541 Do not require the input and output of promotion casts to have
3542 the same sign, but base the signedness of the operation on the
3543 input rather than the result. If the pattern includes two
3544 promotions, check that those promotions have the same sign.
3545 Do not restrict the MULT_EXPR handling to a double-width result;
3546 handle quadruple-width results and wider. Use vect_convert_inputs
3547 to convert the inputs to the common type.
3548 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
3549 to handle the optional cast of the ABS result. Also allow a sign
3550 change or a sign extension between the ABS and MINUS.
3551 Use vect_widened_op_tree to detect the widened subtraction and use
3552 vect_convert_inputs to convert the inputs to the common type.
3553 (vect_handle_widen_op_by_const): Delete.
3554 (vect_recog_widen_op_pattern): New function.
3555 (vect_recog_widen_mult_pattern): Use it.
3556 (vect_recog_widen_shift_pattern): Likewise.
3557 (vect_recog_widen_sum_pattern): Use
3558 vect_look_through_possible_promotion to handle the promoted
3561 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3563 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
3564 the containing gimple_seq *.
3565 * gimple-iterator.h (gsi_for_stmt): Declare it.
3566 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
3567 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
3568 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3569 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3570 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
3572 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
3574 (vect_mark_pattern_stmts): ...here. Handle cases in which the
3575 statement being replaced is part of an existing pattern
3576 definition sequence, inserting the new pattern statements before
3578 (vect_pattern_recog_1): Don't return a bool. If the statement
3579 is already part of a pattern, instead apply pattern matching
3580 to the pattern definition statements. Don't clear the
3581 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
3582 (vect_pattern_recog): Don't break after the first match;
3583 continue processing the pattern definition statements instead.
3584 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
3586 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3588 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3589 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3590 (vect_recog_widen_sum_pattern): Use it.
3592 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3594 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3595 phi is not a pattern statement and has not been replaced by
3596 a pattern statement.
3597 * tree-vect-patterns.c (type_conversion_p): Don't check
3598 STMT_VINFO_IN_PATTERN_P.
3599 (vect_recog_vector_vector_shift_pattern): Likewise.
3600 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
3601 the pattern statement rather than the original statement; check
3602 directly for a WIDEN_MULT_EXPR here.
3603 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
3604 vect_is_simple_use to return the pattern statement rather
3605 than the original statement; use is_pattern_stmt_p to check
3606 for such a pattern statement.
3607 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
3608 to return the pattern statement rather than the original statement;
3609 don't do the same transformation here.
3610 (vect_is_simple_use): If the defining statement has been replaced
3611 by a pattern statement, return the pattern statement instead.
3612 Remove the corresponding (local) transformation from the vectype
3615 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3617 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
3618 end and default to null.
3619 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3620 (vectorizable_reduction): Update calls accordingly, dropping the
3621 gimple ** argument if the passed-back statement isn't needed.
3622 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
3623 (vect_recog_rotate_pattern): Likewise.
3624 (vect_recog_mask_conversion_pattern): Likewise.
3625 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3626 (vect_mask_constant_operand_p): Likewise.
3627 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
3628 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
3629 (get_group_load_store_type, get_load_store_type): Likewise.
3630 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
3631 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
3632 (vectorizable_conversion, vectorizable_assignment): Likewise.
3633 (vectorizable_shift, vectorizable_operation): Likewise.
3634 (vectorizable_store, vect_is_simple_cond): Likewise.
3635 (vectorizable_condition, vectorizable_comparison): Likewise.
3636 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
3637 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
3638 and move it to the end. Cope with null def_stmt_outs.
3640 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3642 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
3644 2018-06-29 Jeff Law <law@redhat.com>
3646 * config/v850/v850.c (v850_legitimate_address_p): Handle large
3647 displacements for TARGET_V850E2V3 and newer.
3648 (TARGET_LRA_P): Remove. Defaults to LRA now.
3649 * config/v850/v850.md (sign23byte_load): Remove.
3650 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
3651 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
3653 2018-06-29 Martin Liska <mliska@suse.cz>
3656 * coverage.c (coverage_init): Mangle full path name.
3657 * doc/invoke.texi: Document the change.
3658 * gcov-io.c (mangle_path): New.
3659 * gcov-io.h (mangle_path): Likewise.
3660 * gcov.c (mangle_name): Use mangle_path for path mangling.
3662 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3664 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
3665 if starting source register is not even.
3667 2018-06-29 Martin Liska <mliska@suse.cz>
3669 PR tree-optimization/86263
3670 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
3671 Make edge redirection.
3673 2018-06-29 David Malcolm <dmalcolm@redhat.com>
3675 * dumpfile.c (dump_loc): Add indentation based on scope depth.
3676 (dump_scope_depth): New variable.
3677 (get_dump_scope_depth): New function.
3678 (dump_begin_scope): New function.
3679 (dump_end_scope): New function.
3680 * dumpfile.h (get_dump_scope_depth): New declaration.
3681 (dump_begin_scope): New declaration.
3682 (dump_end_scope): New declaration.
3683 (class auto_dump_scope): New class.
3684 (AUTO_DUMP_SCOPE): New macro.
3685 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
3688 2018-06-29 Richard Biener <rguenther@suse.de>
3690 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
3691 compute_all_dependences succeeds.
3692 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
3693 exceed --param loop-max-datarefs-for-datadeps.
3695 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3697 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3699 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
3702 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
3703 alternative 0 in preferred_for_speed attribute.
3705 2018-06-28 Paul Koning <ni1d@arrl.net>
3707 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
3708 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
3709 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
3710 * config/pdp11/pdp11.md: Correct "length" attribute calculation
3711 for shift insn patterns.
3713 2018-06-28 David Malcolm <dmalcolm@redhat.com>
3715 * cgraph.c (cgraph_node::get_body): Replace assignments to
3716 "dump_file" with calls to set_dump_file.
3717 * dumpfile.c (alt_dump_file): Make static, and group with...
3718 (alt_flags): ...this definition.
3719 (dumps_are_enabled): New variable.
3720 (refresh_dumps_are_enabled): New function.
3721 (set_dump_file): New function.
3722 (set_alt_dump_file): New function.
3723 (gcc::dump_manager::dump_start): Replace assignments to
3724 "dump_file" and "alt_dump_file" with calls to set_dump_file and
3726 (gcc::dump_manager::dump_finish): Likewise.
3727 * dumpfile.h (alt_dump_file): Delete decl.
3728 (dumps_are_enabled): New variable decl.
3729 (set_dump_file): New function decl.
3730 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
3732 * tree-nested.c (lower_nested_functions): Replace assignments to
3733 "dump_file" with calls to set_dump_file.
3735 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3737 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
3738 goto_locus of each outgoing edge of each basic block.
3740 2018-06-28 Richard Biener <rguenther@suse.de>
3742 * dwarf2out.c (decl_scope_table): Remove.
3743 (push_decl_scope): Likewise.
3744 (pop_decl_scope): Likewise.
3745 (gen_type_die_for_member): Do not call push/pop_decl_scope.
3746 (gen_struct_or_union_type_die): Likewise.
3747 (gen_tagged_type_die): Likewise.
3748 (dwarf2out_init): Do not initialize decl_scope_table.
3749 (dwarf2out_c_finalize): Do not free it.
3751 2018-06-28 Richard Biener <rguenther@suse.de>
3753 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
3754 deciding whether to not re-use a DIE.
3756 2018-06-28 Richard Biener <rguenther@suse.de>
3758 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
3759 DW_AT_abstract_origin attribute.
3761 2018-06-28 Martin Liska <mliska@suse.cz>
3763 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3764 Use newly introduced constants.
3765 * tree-switch-conversion.h (struct jump_table_cluster):
3766 Define max_ratio_for_size and max_ratio_for_speed.
3768 2018-06-28 Martin Liska <mliska@suse.cz>
3770 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3771 Add new checking assert to catch invalid state.
3772 (jump_table_cluster::can_be_handled): Handle single case
3774 (jump_table_cluster::is_beneficial): Bail out for such case.
3775 (bit_test_cluster::find_bit_tests):
3776 Add new checking assert to catch invalid state.
3777 (bit_test_cluster::can_be_handled): Handle single case
3779 (bit_test_cluster::is_beneficial): Bail out for such case.
3780 (switch_decision_tree::analyze_switch_statement):
3783 2018-06-28 Martin Liska <mliska@suse.cz>
3785 * common.opt: Introduce -completion option.
3786 * gcc.c (driver_handle_option): Handle it.
3787 (driver::main): Print completions if completion
3789 * opt-suggestions.c (option_proposer::get_completions):
3791 (option_proposer::suggest_completion): Likewise.
3792 (option_proposer::find_param_completions): Likewise.
3793 (verify_autocompletions): Likewise.
3794 (test_completion_valid_options): Likewise.
3795 (test_completion_valid_params): Likewise.
3796 (in_completion_p): Likewise.
3797 (empty_completion_p): Likewise.
3798 (test_completion_partial_match): Likewise.
3799 (test_completion_garbage): Likewise.
3800 (opt_proposer_c_tests): Likewise.
3801 * opt-suggestions.h: Declare new functions.
3802 * opts.c (common_handle_option): Handle OPT__completion_.
3803 * selftest-run-tests.c (selftest::run_tests): Add
3804 opt_proposer_c_tests.
3805 * selftest.c (assert_str_startswith): New.
3806 * selftest.h (assert_str_startswith): Likewise.
3807 (opt_proposer_c_tests): New.
3808 (ASSERT_STR_STARTSWITH): Likewise.
3810 2018-06-28 Martin Liska <mliska@suse.cz>
3812 * Makefile.in: Add opt-suggestions.o.
3813 * gcc-main.c: Include opt-suggestions.h.
3814 * gcc.c (driver::driver): Likewise.
3815 (driver::~driver): Remove m_option_suggestions.
3816 (driver::build_option_suggestions): Moved to option_proposer.
3817 (driver::suggest_option): Likewise.
3818 (driver::handle_unrecognized_options): Use option_proposer.
3819 * gcc.h (class driver): Add new memver m_option_proposer.
3820 * opt-suggestions.c: New file.
3821 * opt-suggestions.h: New file.
3823 2018-06-28 Martin Liska <mliska@suse.cz>
3825 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3826 (auto_string_vec::~auto_string_vec): Likewise.
3828 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3830 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3831 prevent_decl_creation_for_types fields up and add reset_location field.
3832 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3833 statement if id->reset_location is true.
3834 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3835 id->reset_location is true.
3836 (copy_phis_for_bb): Force input_location on the arguments if
3837 id->reset_location is true.
3838 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3839 is set on the function to be inlined.
3841 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3843 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3845 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3847 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3848 registers for Pmode.
3849 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3850 hard registers for the clobbered pseudo.
3852 2018-06-27 Paul Koning <ni1d@arrl.net>
3854 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3855 mutually exclusive options.
3856 * config/pdp11/constraints.md (h): New constraint.
3857 (O): Update definition to match shift code generation.
3858 (D): New constraint.
3859 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3861 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3863 (output_jump): Change arguments.
3864 (pdp11_fixed_cc_regs): New function.
3865 (pdp11_cc_mode): Ditto.
3866 (pdp11_expand_shift): Ditto.
3867 (pdp11_assemble_shift): Ditto.
3868 (pdp11_small_shift): Ditto.
3869 (pdp11_branch_cost): Remove.
3870 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3872 (pdp11_register_move_cost): Update for CC registers.
3873 (pdp11_rtx_costs): Add case for LSHIFTRT.
3874 (pdp11_output_jump): Add CCNZ mode conditional branches.
3875 (notice_update_cc_on_set): Remove.
3876 (pdp11_cc_mode): New function.
3877 (simple_memory_operand): Correct pre/post decrement case.
3878 (no_side_effect_operand): New function.
3879 (pdp11_regno_reg_class): Add CC_REGS class.
3880 (pdp11_fixed_cc_regs): New function.
3881 (pdp11_small_shift): New function.
3882 (pdp11_expand_shift): New function to expand shift insns.
3883 (pdp11_assemble_shift): New function to output shifts.
3884 (pdp11_branch_cost): Remove.
3885 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3886 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3887 (WCHAR_TYPE): Ditto.
3888 (PTRDIFF_TYPE): Ditto.
3889 (ADJUST_INSN_LENGTH): New macro.
3890 (FIXED_REGISTERS): Add CC registers.
3891 (CALL_USED_REGISTERS): Ditto.
3893 (REG_CLASS_NAMES): Ditto.
3894 (REG_CLASS_CONTENTS): Ditto.
3895 (SELECT_CC_MODE): Use new function.
3896 (TARGET_FLAGS_REGNUM): New macro.
3897 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3898 (cc0_reg_rtx): Remove.
3899 (CC_STATUS_MDEP): Remove.
3900 (CC_STATUS_MDEFP_INIT): Remove.
3901 (CC_IN_FPU): Remove.
3902 (NOTICE_UPDATE_CC): Remove.
3903 (REGISTER_NAMES): Add CC registers.
3904 (BRANCH_COST): Change to constant 1.
3905 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3907 * config/pdp11/pdp11.opt (mbcopy): Remove.
3908 (mbcopy-builtin): Remove.
3909 (mbranch-cheap): Remove.
3910 (mbranch-expensive): Remove.
3911 * config/pdp11/predicates.md (expand_shift_operand): Update to
3912 match shift code generation.
3913 (ccnz_operator): New predicate.
3914 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3915 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3916 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3918 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3919 description of O constraint.
3921 2018-06-27 Jeff Law <law@redhat.com>
3922 Austin Law <austinklaw@gmail.com>
3924 * config/v850/v850.md (addsi3_set_flags): New pattern.
3925 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3926 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3927 (zero_extendhisi2_v850_set_flags): Likewise.
3928 (zero_extendqisi2_v850_set_flags): Likewise.
3929 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3930 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3931 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3933 * config/v850/v850-protos.h (notice_update_cc): Remove.
3934 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3935 (v850_print_operand): Handle 'D' and "d".
3936 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3937 Add handling of arithmetic/logical operations compared against zero.
3938 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3939 Do not look at v850_compare_op, instead get mode from last argument.
3940 (v850_gen_compare): Remove
3941 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3942 after reload for prologue insns.
3943 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3945 (construct_save_jarl): Likewise.
3946 (TARGET_FLAGS_REGNUM): Define.
3947 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3948 (NOTICE_UPDATE_CC): Remove.
3949 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3950 than cc0. Conditionalize on reload_completed.
3951 (cmpsi_insn, setfcc_insn): Likewise.
3952 (tst1 splitter): Turn into define_and_split which sets the flags
3954 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3955 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3956 (cstoresf4, cstoredf4): Clobber the flags.
3957 (cmpsi, cmpsf, cmpdf): Remove expanders.
3958 (setf_insn): Remove pattern.
3959 (addsi3): Turn into define_and_split which clobbers the flags after
3960 reload and a suitable pattern (addsi3_clobber_flags) for use after
3962 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3963 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3964 (ashrsi3, ashrsi3_v850e2): Likewise.
3965 (bins): Clobber the flags.
3966 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3967 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3968 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3969 (call_value_internal_short, call_value_internal_long): Likewise.
3970 (callt_save_interrupt, callt_return_interrupt): Likewise.
3971 (save_interrupt, return_interrupt): Likewise.
3972 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3973 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3974 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3975 (All FP comparisons): Only allow after reload has completed.
3977 (divh, divhu): Tweak output template.
3978 (branch_z_normal, branch_z_invert): Remove
3979 (branch_nz_normal, branch_nz_invert): Likewise.
3980 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3982 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3983 * config/v850/v850.c (notice_update_cc): Remove.
3984 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3985 (CC_NO_CARRY): Likewise.
3986 (NOTICE_UPDATE_CC): Define to nothing.
3987 * config/v850/v850.md: Remove block comment on cc0 handling
3988 Remove "cc" attribute from all patterns. Remove cc_status handling
3989 from all patterns. Minor formatting fixes.
3991 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3993 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3994 (cortex-a76.cortex-a55): Likewise.
3995 * config/aarch64/aarch64-tune.md: Regenerate.
3996 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3997 cortex-a76.cortex-a55.
3999 2018-06-27 Jeff Law <law@redhat.com>
4001 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
4002 (MULTILIB_DIRNAMES): Similarly.
4004 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
4006 * gimple.h (gimple_return_retbnd): Delete.
4007 (gimple_return_set_retbnd): Likewise.
4008 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
4009 gimple_return_set_retbnd.
4010 * gimple-pretty-print.c (dump_gimple_return): Remove call to
4011 gimple_return_retbnd and adjust.
4012 * tree-inline.h (struct copy_body_data): Remove retbnd field.
4013 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
4014 Explicitly return NULL in a couple more cases. Move assertion
4015 on debug statements and remove unreachable code.
4016 (reset_debug_binding): Do not test id->retbnd.
4017 (expand_call_inline): Do not set it.
4019 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4021 * configure.ac: Add --disable-gcov option.
4022 * configure: Regenerate.
4023 * Makefile.in: Honour @enable_gcov@.
4024 * doc/install.texi: Document --disable-gcov.
4026 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4028 * config/arm/arm-cpus.in (cortex-a76): New entry.
4029 (cortex-a76.cortex-a55): Likewise.
4030 * config/arm/arm-tables.opt: Regenerate.
4031 * config/arm/arm-tune.md: Likewise.
4032 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
4033 * doc/invoke.texi (ARM Options): Document cortex-a76 and
4034 cortex-a76.cortex-a55.
4036 2018-06-27 Tamar Christina <tamar.christina@arm.com>
4039 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
4041 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
4043 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
4045 (EPILOGUE_USES): Likewise.
4047 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
4049 * tree-inline.c (remap_location): New function extracted from...
4050 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
4051 (copy_phis_for_bb): ...here. Call remap_location.
4052 (copy_cfg_body): Adjust call to copy_edges_for_bb.
4054 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
4056 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
4057 unaligned vsx for 16B memset.
4059 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
4062 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
4063 ieee128_float_type_node to long_double_type_node unless
4064 TARGET_LONG_DOUBLE_128 is set.
4066 2018-06-26 David Malcolm <dmalcolm@redhat.com>
4068 * cfgloop.c (get_loop_location): Convert return type from
4069 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
4070 by implicit construction from rtx_insn *, and using
4071 dump_user_location_t::from_function_decl for the fallback case.
4072 * cfgloop.h (get_loop_location): Convert return type from
4073 location_t to dump_user_location_t.
4074 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
4075 dump_printf_loc to pass in a dump_location_t rather than a
4076 location_t, via the gimple stmt.
4077 * coverage.c (get_coverage_counts): Update calls to
4078 dump_printf_loc to pass in dump_location_t rather than a
4080 * doc/optinfo.texi (Dump types): Convert example of
4081 dump_printf_loc from taking "locus" to taking "insn". Update
4082 description of the "_loc" calls to cover dump_location_t.
4083 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
4085 (dump_user_location_t::dump_user_location_t): New constructors,
4086 from gimple *stmt and rtx_insn *.
4087 (dump_user_location_t::from_function_decl): New function.
4088 (dump_loc): Make static.
4089 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
4090 const dump_location_t &.
4091 (dump_generic_expr_loc): Delete.
4092 (dump_printf_loc): Convert param "loc" from location_t to
4093 const dump_location_t &.
4094 (selftest::test_impl_location): New function.
4095 (selftest::dumpfile_c_tests): New function.
4096 * dumpfile.h: Include "profile-count.h".
4097 (class dump_user_location_t): New class.
4098 (struct dump_impl_location_t): New struct.
4099 (class dump_location_t): New class.
4100 (dump_printf_loc): Convert 2nd param from source_location to
4101 const dump_location_t &.
4102 (dump_generic_expr_loc): Delete.
4103 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
4104 const dump_location_t &.
4105 * gimple-fold.c (fold_gimple_assign): Update call to
4106 dump_printf_loc to pass in a dump_location_t rather than a
4107 location_t, via the gimple stmt.
4108 (gimple_fold_call): Likewise.
4109 * gimple-loop-interchange.cc
4110 (loop_cand::analyze_iloop_reduction_var): Update for change to
4111 check_reduction_path.
4112 (tree_loop_interchange::interchange): Update for change to
4114 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
4115 change in return-type of find_loop_location.
4116 (graphite_regenerate_ast_isl): Likewise.
4117 * graphite-optimize-isl.c (optimize_isl): Likewise.
4118 * graphite.c (graphite_transform_loops): Likewise.
4119 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
4120 pass in a dump_location_t rather than a location_t, via the
4122 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4123 * ipa.c (walk_polymorphic_call_targets): Likewise.
4124 * loop-unroll.c (report_unroll): Convert "locus" param from
4125 location_t to dump_location_t.
4126 (decide_unrolling): Update for change to get_loop_location's
4128 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
4129 location_t to dump_user_location_t.
4130 (grid_find_single_omp_among_assignments_1): Updates calls to
4131 dump_printf_loc to pass in a dump_location_t rather than a
4132 location_t, via the gimple stmt.
4133 (grid_parallel_clauses_gridifiable): Convert "tloc" from
4134 location_t to dump_location_t. Updates calls to dump_printf_loc
4135 to pass in a dump_location_t rather than a location_t, via the
4137 (grid_inner_loop_gridifiable_p): Likewise.
4138 (grid_dist_follows_simple_pattern): Likewise.
4139 (grid_gfor_follows_tiling_pattern): Likewise.
4140 (grid_target_follows_gridifiable_pattern): Likewise.
4141 (grid_attempt_target_gridification): Convert initialization
4142 of local "grid" from memset to zero-initialization; FIXME: does
4143 this require C++11? Update call to dump_printf_loc to pass in a
4144 optinfo_location rather than a location_t, via the gimple stmt.
4145 * profile.c (read_profile_edge_counts): Updates call to
4146 dump_printf_loc to pass in a dump_location_t rather than a
4148 (compute_branch_probabilities): Likewise.
4149 * selftest-run-tests.c (selftest::run_tests): Call
4151 * selftest.h (dumpfile_c_tests): New decl.
4152 * tree-loop-distribution.c (pass_loop_distribution::execute):
4153 Update for change in return type of find_loop_location.
4154 * tree-parloops.c (parallelize_loops): Likewise.
4155 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
4156 "locus" from location_t to dump_user_location_t.
4157 (canonicalize_loop_induction_variables): Likewise.
4158 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
4159 for change in return type of find_loop_location.
4160 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
4161 to dump_printf_loc to pass in a dump_location_t rather than a
4162 location_t, via the stmt.
4163 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
4165 * tree-vect-loop-manip.c (find_loop_location): Convert return
4166 type from source_location to dump_user_location_t.
4167 (vect_do_peeling): Update for above change.
4168 (vect_loop_versioning): Update for change in type of
4170 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
4171 from location_t to dump_user_location_t.
4172 (vect_estimate_min_profitable_iters): Update for change in type
4174 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
4175 location_t to dump_location_t.
4176 (vect_slp_bb): Update for change in type of vect_location.
4177 * tree-vectorizer.c (vect_location): Convert from source_location
4178 to dump_user_location_t.
4179 (try_vectorize_loop_1): Update for change in vect_location's type.
4180 (vectorize_loops): Likewise.
4181 (increase_alignment): Likewise.
4182 * tree-vectorizer.h (vect_location): Convert from source_location
4183 to dump_user_location_t.
4184 (find_loop_location): Convert return type from source_location to
4185 dump_user_location_t.
4186 (check_reduction_path): Convert 1st param from location_t to
4187 dump_user_location_t.
4188 * value-prof.c (check_counter): Update call to dump_printf_loc to
4189 pass in a dump_user_location_t rather than a location_t; update
4190 call to error_at for change in type of "locus".
4191 (check_ic_target): Update call to dump_printf_loc to
4192 pass in a dump_user_location_t rather than a location_t, via the
4195 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
4197 * config/s390/s390.h (enum processor_flags): Do not use
4198 default tune parameter when -march was specified.
4200 2018-06-26 Jakub Jelinek <jakub@redhat.com>
4203 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
4204 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
4207 2018-06-26 Richard Biener <rguenther@suse.de>
4209 PR tree-optimization/86287
4211 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
4212 (vect_analyze_loop): Initialize n_stmts.
4214 2018-06-26 Richard Biener <rguenther@suse.de>
4217 * fold-const.c (fold_convertible_p): Pointer extension
4220 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
4223 * dwarf2out.c (loc_list_has_views): Adjust comments.
4224 (dw_loc_list): Split single cross-partition range with
4227 2018-06-25 Jeff Law <law@redhat.com>
4229 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
4230 on -mbig-switch by default.
4232 * config/v850/predicates.md (const_float_1_operand): Fix match_code
4234 (const_float_0_operand): Remove unused predicate.
4235 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
4236 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
4237 (recipsf2): New expander. Original pattern now called
4239 (recipdf2, recipdf2_insn): Similarly.
4240 (rsqrtsf2, rsqrtsf2_insn): Similarly
4241 (rsqrtdf2, rsqrtdf2_insn): Similarly
4243 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
4245 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
4246 Simplify logic for FreeBSD (twice).
4248 2018-06-25 Martin Sebor <msebor@redhat.com>
4250 PR tree-optimization/86204
4251 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
4252 a strnlen result if it's less than the length of the string.
4254 2018-06-25 Martin Sebor <msebor@redhat.com>
4256 PR tree-optimization/85700
4257 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
4258 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
4259 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
4261 2018-06-25 Martin Sebor <msebor@redhat.com>
4263 * doc/extend.texi (Zero-length arrays): Update and clarify.
4265 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
4267 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
4268 added IEEE/IBM long double multilib support on PowerPC little
4269 endian Linux systems.
4270 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
4271 (MULTILIB_DEFAULTS): Likewise.
4272 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4274 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
4275 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
4276 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
4278 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
4281 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
4282 (REORDER_45): Likewise.
4284 2018-06-25 Jeff Law <law@redhat.com>
4286 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
4287 dividend to 32 bits. Adjust length.
4288 (udivmodhi4): Cleanup output template. Fix length.
4290 2018-06-25 Carl Love <cel@us.ibm.com>
4292 * config/rs6000/vsx.md: Change word selector to prefered location.
4294 2018-06-25 Richard Biener <rguenther@suse.de>
4296 PR tree-optimization/86304
4297 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
4298 epilogue-if-converted loops as well.
4300 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
4302 * lto-section-out.c (lto_begin_section): Do not print section
4303 name for noaddr and unnumbered dumps.
4305 2018-06-25 Richard Biener <rguenther@suse.de>
4307 * tree-vectorizer.h (struct vec_info_shared): New structure
4308 with parts split out from struct vec_info and loop_nest from
4309 struct _loop_vec_info.
4310 (struct vec_info): Adjust accordingly.
4311 (struct _loop_vec_info): Likewise.
4312 (LOOP_VINFO_LOOP_NEST): Adjust.
4313 (LOOP_VINFO_DATAREFS): Likewise.
4314 (LOOP_VINFO_DDRS): Likewise.
4315 (struct _bb_vec_info): Likewise.
4316 (BB_VINFO_DATAREFS): Likewise.
4317 (BB_VINFO_DDRS): Likewise.
4318 (struct _stmt_vec_info): Add dr_aux member.
4319 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
4320 (DR_MISALIGNMENT_UNINITIALIZED): New.
4321 (set_dr_misalignment): Adjust.
4322 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
4323 (vect_analyze_loop): Adjust prototype.
4324 (vect_analyze_loop_form): Likewise.
4325 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4326 Compute dependences lazily.
4327 (vect_record_base_alignments): Use shared datarefs/ddrs.
4328 (vect_verify_datarefs_alignment): Likewise.
4329 (vect_analyze_data_refs_alignment): Likewise.
4330 (vect_analyze_data_ref_accesses): Likewise.
4331 (vect_analyze_data_refs): Likewise.
4332 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
4333 constructor parameter for shared part.
4334 (vect_analyze_loop_form): Pass in shared part and adjust.
4335 (vect_analyze_loop_2): Pass in storage for the number of
4336 stmts. Move loop nest finding to the caller. Compute
4338 (vect_analyze_loop): Pass in shared part.
4339 (vect_transform_loop): Verify shared datarefs are unchanged.
4340 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
4341 constructor parameter for shared part.
4342 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
4343 (vect_slp_bb): Verify shared datarefs are unchanged before
4345 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
4347 (new_stmt_vec_info): Initialize DR_AUX misalignment to
4348 DR_MISALIGNMENT_UNINITIALIZED.
4349 * tree-vectorizer.c (vec_info::vec_info): Add constructor
4350 parameter for shared part.
4351 (vec_info::~vec_info): Adjust.
4352 (vec_info_shared::vec_info_shared): New.
4353 (vec_info_shared::~vec_info_shared): Likewise.
4354 (vec_info_shared::save_datarefs): Likewise.
4355 (vec_info_shared::check_datarefs): Likewise.
4356 (try_vectorize_loop_1): Construct shared part live for analyses
4357 of a single loop for multiple vector sizes.
4358 * tree-parloops.c (gather_scalar_reductions): Adjust.
4360 2018-06-25 Richard Biener <rguenther@suse.de>
4362 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
4363 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
4364 (vect_analyze_data_refs): Remove similar code from here and
4365 simplify accordingly.
4367 2018-06-25 Richard Biener <rguenther@suse.de>
4369 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
4370 for reverse storage order accesses rather than asserting
4371 they cannot happen here.
4373 2018-06-25 Tom de Vries <tdevries@suse.de>
4376 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
4377 Use data16 instead of .byte for insn prefix.
4379 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
4382 * parser.c (make_char_string_pack): Pass this literal chars
4383 through cpp_interpret_string.
4384 (cp_parser_userdef_numeric_literal): Check the result of
4385 make_char_string_pack.
4387 2018-06-24 Maya Rashish <coypu@sdf.org>
4389 * ginclude/stddef.h: Simplify conditions around avoiding
4390 re-definition of __size_t.
4392 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4394 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
4395 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
4397 2018-06-22 Maya Rashish <coypu@sdf.org>
4399 * doc/invoke.texi (mno-fancy-math-387): Update for changes
4400 made to OpenBSD and NetBSD through the years.
4402 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4404 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4405 behavior of vec_pack (vector double, vector double) to match
4406 behavior of vec_float2 (vector double, vector double).
4408 2018-06-22 Olivier Hainque <hainque@adacore.com>
4410 * gimplify.c (gimplify_function_tree): Prevent creation
4411 of a trampoline for the address of the current function
4412 passed to entry/exit instrumentation hooks.
4414 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
4417 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
4420 2018-06-22 Martin Liska <mliska@suse.cz>
4422 PR tree-optimization/86263
4423 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4424 Bail out if is_enabled is false.
4425 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
4427 (jump_table_cluster::is_enabled): New function.
4429 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4431 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4432 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
4433 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
4434 (lto_input_ts_binfo_tree_pointers): Likewise.
4435 * tree-streamer-out.c (streamer_write_tree_bitfields,
4436 write_ts_binfo_tree_pointers): Likewise.
4437 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
4439 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4441 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
4443 2018-06-22 Martin Liska <mliska@suse.cz>
4445 * symbol-summary.h (get): Make it pure and inline move
4446 functionality from ::get function.
4447 (get): Remove and inline into ::get and ::get_create.
4448 (get_create): Move code from ::get function.
4450 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4453 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
4454 -x assembler-with-cpp.
4456 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4458 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
4459 _FILE_OFFSET_BITS=64 for C++.
4461 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
4463 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
4464 conversion insn that shows up when pr85657-3.c is compiled using
4465 IEEE 128-bit long double.
4466 (neg<mode>2_internal): Use the correct mode to check whether the
4467 mode is IBM extended.
4468 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
4469 multiply and divide external functions from being created more
4472 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4474 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
4476 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
4477 the edge can be forwarded.
4478 (cfg_layout_merge_blocks): Likewise.
4480 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4482 * except.c (finish_eh_generation): Commit edge insertions only after
4483 the EH edges have been redirected from post-landing to landing pads.
4485 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4487 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
4488 create_tmp_var_for to create the FRAME decl.
4489 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
4491 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4493 * tree-inline.c (copy_edges_for_bb): Minor tweak.
4494 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
4495 debug statement when resetting its value.
4496 (expand_call_inline): Copy the locus of the call onto the assignment
4497 of the return value, if any. Use local variable in more cases.
4499 2018-06-21 Martin Liska <mliska@suse.cz>
4501 * ipa-pure-const.c (propagate_nothrow): Use
4502 funct_state_summaries->get.
4503 (dump_malloc_lattice): Likewise.
4504 (propagate_malloc): Likewise.
4506 2018-06-21 Richard Biener <rguenther@suse.de>
4508 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
4509 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
4510 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
4512 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
4513 BLOCK_ABSTRACT_ORIGIN unconditionally.
4515 2018-06-21 David Malcolm <dmalcolm@redhat.com>
4517 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
4519 * ipa-reference.c (ipa_reference_c_finalize): Delete
4520 ipa_ref_opt_sum_summaries and set it to NULL.
4522 2018-06-21 Tom de Vries <tdevries@suse.de>
4524 PR tree-optimization/85859
4525 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
4526 test with comment from bb_no_side_effects_p.
4528 2018-06-21 Richard Biener <rguenther@suse.de>
4530 PR tree-optimization/86232
4531 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
4532 max for constant niter.
4534 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4536 * config/aarch64/aarch64-simd.md
4537 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
4539 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4541 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
4542 Make opernads of the unspec commutative.
4544 2018-06-21 Richard Biener <rguenther@suse.de>
4546 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4547 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
4548 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
4549 (vect_analyze_data_ref_dependence): Re-order checks to deal with
4551 (vect_record_base_alignments): Do not record base alignment
4552 for gathers or scatters.
4553 (vect_compute_data_ref_alignment): Drop return value that is always
4554 true. Bail out early for gathers or scatters.
4555 (vect_enhance_data_refs_alignment): Bail out early for gathers
4557 (vect_find_same_alignment_drs): Likewise.
4558 (vect_analyze_data_refs_alignment): Remove dead code.
4559 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4560 (vect_analyze_data_refs): For possible gathers or scatters do
4561 not create an alternate DR, just check their possible validity
4562 and mark them. Adjust DECL_NONALIASED handling to not rely
4564 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4565 update inits of gathers or scatters.
4566 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4567 Also copy gather/scatter flag to pattern vinfo.
4569 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
4571 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4572 behavior of vec_packsu (vector unsigned long long, vector unsigned
4573 long long) to match behavior of vec_packs with same signature.
4575 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
4576 Thomas Schwinge <thomas@codesourcery.com>
4577 Cesar Philippidis <cesar@codesourcery.com>
4579 * gimplify.c (gimplify_scan_omp_clauses): Add support for
4580 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4581 (gimplify_adjust_omp_clauses): Likewise.
4582 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
4583 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
4584 (gimplify_omp_target_update): Update handling of acc update and
4586 * omp-low.c (install_var_field): Remove unused parameter
4587 base_pointers_restrict.
4588 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
4589 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
4591 (omp_target_base_pointers_restrict_p): Delete.
4592 (scan_omp_target): Update call to scan_sharing_clauses.
4593 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
4595 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4596 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4597 (convert_local_omp_clauses): Likewise.
4598 * tree-pretty-print.c (dump_omp_clause): Likewise.
4599 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
4601 (omp_clause_code_name): Likewise.
4603 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4606 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4609 PR tree-optimization/86231
4610 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4611 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4613 2018-06-20 Tom de Vries <tdevries@suse.de>
4615 PR tree-optimization/86097
4616 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
4617 iv type if signedness of iv type is not the same as that of *nit.
4619 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4621 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
4622 EDGE_EH edges, verify they are all EDGE_EH.
4624 2018-06-20 Maya Rashish <coypu@sdf.org>
4626 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
4628 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4630 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
4631 * config/aarch64/aarch64.c (xgene1_tunings): Add
4632 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
4633 (aarch64_mode_valid_for_sched_fusion_p):
4634 Allow 16-byte modes.
4635 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
4636 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
4638 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
4640 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
4641 * config/aarch64/iterators.md (VQ2): New mode iterator.
4643 2018-06-20 Martin Liska <mliska@suse.cz>
4645 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4646 Change default ratio from 10 to 8.
4648 2018-06-20 Martin Liska <mliska@suse.cz>
4650 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4652 (bit_test_cluster::find_bit_tests): Likewise.
4653 (switch_decision_tree::analyze_switch_statement): Find clusters.
4654 * tree-switch-conversion.h (struct jump_table_cluster): Document
4657 2018-06-20 Martin Liska <mliska@suse.cz>
4659 * tree-switch-conversion.c (switch_conversion::collect):
4660 Record m_uniq property.
4661 (switch_conversion::expand): Bail out for special conditions.
4662 (group_cluster::~group_cluster): New.
4663 (group_cluster::group_cluster): Likewise.
4664 (group_cluster::dump): Likewise.
4665 (jump_table_cluster::emit): New.
4666 (switch_decision_tree::fix_phi_operands_for_edges): New.
4667 (struct case_node): Remove struct.
4668 (jump_table_cluster::can_be_handled): New.
4669 (case_values_threshold): Moved to header.
4670 (reset_out_edges_aux): Likewise.
4671 (jump_table_cluster::is_beneficial): New.
4672 (bit_test_cluster::can_be_handled): Likewise.
4673 (add_case_node): Remove.
4674 (bit_test_cluster::is_beneficial): New.
4675 (case_bit_test::cmp): New.
4676 (bit_test_cluster::emit): New.
4677 (expand_switch_as_decision_tree_p): Remove.
4678 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
4679 (fix_phi_operands_for_edge): Likewise.
4680 (switch_decision_tree::analyze_switch_statement): New.
4681 (compute_cases_per_edge): Move ...
4682 (switch_decision_tree::compute_cases_per_edge): ... here.
4683 (try_switch_expansion): Likewise.
4684 (switch_decision_tree::try_switch_expansion): Likewise.
4685 (record_phi_operand_mapping): Likewise.
4686 (switch_decision_tree::record_phi_operand_mapping): Likewise.
4687 (emit_case_decision_tree): Likewise.
4688 (switch_decision_tree::emit): Likewise.
4689 (balance_case_nodes): Likewise.
4690 (switch_decision_tree::balance_case_nodes): Likewise.
4691 (dump_case_nodes): Likewise.
4692 (switch_decision_tree::dump_case_nodes): Likewise.
4693 (emit_jump): Likewise.
4694 (switch_decision_tree::emit_jump): Likewise.
4695 (emit_cmp_and_jump_insns): Likewise.
4696 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
4697 (emit_case_nodes): Likewise.
4698 (switch_decision_tree::emit_case_nodes): Likewise.
4699 (conditional_probability): Remove.
4700 * tree-switch-conversion.h (enum cluster_type): New.
4702 (struct cluster): Likewise.
4703 (cluster::cluster): Likewise.
4704 (struct simple_cluster): Likewise.
4705 (simple_cluster::simple_cluster): Likewise.
4706 (struct group_cluster): Likewise.
4707 (struct jump_table_cluster): Likewise.
4708 (struct bit_test_cluster): Likewise.
4709 (struct min_cluster_item): Likewise.
4710 (struct case_tree_node): Likewise.
4711 (case_tree_node::case_tree_node): Likewise.
4712 (jump_table_cluster::case_values_threshold): Likewise.
4713 (struct case_bit_test): Likewise.
4714 (struct switch_decision_tree): Likewise.
4715 (struct switch_conversion): Likewise.
4716 (switch_decision_tree::reset_out_edges_aux): Likewise.
4718 2018-06-20 Martin Liska <mliska@suse.cz>
4720 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
4721 (hoist_edge_and_branch_if_true): Likewise.
4722 (expand_switch_using_bit_tests_p): Likewise.
4723 (struct case_bit_test): Likewise.
4724 (case_bit_test_cmp): Likewise.
4725 (emit_case_bit_tests): Likewise.
4726 (switch_conversion::switch_conversion): New class.
4727 (struct switch_conv_info): Remove old struct.
4728 (collect_switch_conv_info): More to ...
4729 (switch_conversion::collect): ... this.
4730 (check_range): Likewise.
4731 (switch_conversion::check_range): Likewise.
4732 (check_all_empty_except_final): Likewise.
4733 (switch_conversion::check_all_empty_except_final): Likewise.
4734 (check_final_bb): Likewise.
4735 (switch_conversion::check_final_bb): Likewise.
4736 (create_temp_arrays): Likewise.
4737 (switch_conversion::create_temp_arrays): Likewise.
4738 (free_temp_arrays): Likewise.
4739 (gather_default_values): Likewise.
4740 (switch_conversion::gather_default_values): Likewise.
4741 (build_constructors): Likewise.
4742 (switch_conversion::build_constructors): Likewise.
4743 (constructor_contains_same_values_p): Likewise.
4744 (switch_conversion::contains_same_values_p): Likewise.
4745 (array_value_type): Likewise.
4746 (switch_conversion::array_value_type): Likewise.
4747 (build_one_array): Likewise.
4748 (switch_conversion::build_one_array): Likewise.
4749 (build_arrays): Likewise.
4750 (switch_conversion::build_arrays): Likewise.
4751 (gen_def_assigns): Likewise.
4752 (switch_conversion::gen_def_assigns): Likewise.
4753 (prune_bbs): Likewise.
4754 (switch_conversion::prune_bbs): Likewise.
4755 (fix_phi_nodes): Likewise.
4756 (switch_conversion::fix_phi_nodes): Likewise.
4757 (gen_inbound_check): Likewise.
4758 (switch_conversion::gen_inbound_check): Likewise.
4759 (process_switch): Use the newly created class.
4760 (switch_conversion::expand): New.
4761 (switch_conversion::~switch_conversion): New.
4762 * tree-switch-conversion.h: New file.
4764 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4766 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4767 tree-vect-patterns.c.
4768 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4769 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4770 (vect_recog_sad_pattern): Likewise.
4771 (vect_recog_widen_sum_pattern): Likewise.
4772 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4773 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4774 (vect_recog_rotate_pattern): Likewise.
4775 (vect_recog_mult_pattern): Likewise.
4776 (vect_recog_vector_vector_shift_pattern): Likewise.
4777 (vect_recog_divmod_pattern): Likewise.
4778 (vect_recog_mixed_size_cond_pattern): Likewise.
4779 (vect_recog_bool_pattern): Likewise.
4780 (vect_recog_mask_conversion_pattern): Likewise.
4781 (vect_try_gather_scatter_pattern): Likewise.
4782 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4783 (vect_recog_over_widening_pattern): Likewise.
4784 (vect_recog_gather_scatter_pattern): Likewise.
4785 (vect_recog_func_ptr): Move from tree-vectorizer.h
4786 (vect_vect_recog_func_ptrs): Move further down the file.
4787 (vect_recog_func): Likewise. Remove the third argument.
4788 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4789 (vect_pattern_recog_1): Expect the pattern function to do any
4790 necessary target tests. Also expect it to provide a vector type.
4791 Remove the type_in handling.
4793 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4795 * tree-vect-patterns.c (vect_pattern_detected): New function.
4796 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4797 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4798 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4799 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4800 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4801 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4802 (vect_recog_mask_conversion_pattern)
4803 (vect_try_gather_scatter_pattern): Likewise.
4805 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4807 * tree-vect-patterns.c (vect_get_internal_def): New function.
4808 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4809 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4810 (search_type_for_mask_1): Use it.
4812 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4814 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4815 redundant WIDEN_SUM_EXPR handling.
4816 (vect_recog_sad_pattern): Likewise.
4818 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4820 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4821 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4822 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4823 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4824 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4826 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4828 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4829 use the stmt_vec_info of the original bb statement for the
4830 new zero assignment, even if the call is part of a pattern.
4832 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4834 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4835 that the sequence is attached to the original statement rather
4836 than the pattern statement.
4837 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4838 PATTERN_DEF_SEQ from the original statement rather than
4839 the main pattern statement.
4840 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4841 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4842 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4844 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4846 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4847 definition statements before the early exit for statements that aren't
4849 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4851 (vect_transform_loop): ...here. Process pattern definition
4852 statements without first checking whether the main pattern
4853 statement is live or relevant.
4855 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4857 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4858 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4860 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4862 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4863 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4864 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4865 (expand_block_compare): Change select_block_compare_mode call.
4866 (expand_strncmp_align_check): Use new functions, fix comment.
4867 (emit_final_str_compare_gpr): New function.
4868 (expand_strn_compare): Refactor and clean up code.
4869 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4871 2018-06-19 Tony Reix <tony.reix@atos.com>
4872 Damien Bergamini <damien.bergamini@atos.com>
4873 David Edelsohn <dje.gcc@gmail.com>
4875 * collect2.c (static_obj): New variable.
4876 (static_libs): New variable.
4877 (is_in_list): Uncomment declaration.
4878 (main): Track AIX libraries linked statically.
4879 (is_in_list): Uncomment definition.
4880 (scan_prog_file): Don't add AIX shared libraries initializer
4881 to constructor list if linking statically.
4883 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4885 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4887 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4889 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4891 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4894 2018-06-19 Martin Liska <mliska@suse.cz>
4896 * config/i386/i386.c (ix86_can_inline_p): Do not use
4897 ipa_fn_summaries::get_create.
4898 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4900 (devirtualization_time_bonus): Likewise.
4901 (ipcp_propagate_stage): Likewise.
4902 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4903 (edge_set_predicate): Likewise.
4904 (evaluate_conditions_for_known_args): Likewise.
4905 (evaluate_properties_for_edge): Likewise.
4906 (ipa_call_summary::reset): Tranform to ...
4907 (ipa_call_summary::~ipa_call_summary): ... this.
4908 (ipa_fn_summary::reset): Transform to ...
4909 (ipa_fn_summary::~ipa_fn_summary): ... this.
4910 (ipa_fn_summary_t::remove): Rename to ...
4911 (ipa_fn_summary_t::remove_callees): ... this.
4912 (ipa_fn_summary_t::duplicate): Use placement new
4913 instead of memory copy.
4914 (ipa_call_summary_t::duplicate): Likewise.
4915 (ipa_call_summary_t::remove): Remove.
4916 (dump_ipa_call_summary): Change get_create to get.
4917 (ipa_dump_fn_summary): Dump only when summary exists.
4918 (analyze_function_body): Use symbol_summary::get instead
4920 (compute_fn_summary): Likewise.
4921 (estimate_edge_devirt_benefit): Likewise.
4922 (estimate_edge_size_and_time): Likewise.
4923 (inline_update_callee_summaries): Likewise.
4924 (remap_edge_change_prob): Likewise.
4925 (remap_edge_summaries): Likewise.
4926 (ipa_merge_fn_summary_after_inlining): Likewise.
4927 (write_ipa_call_summary): Likewise.
4928 (ipa_fn_summary_write): Likewise.
4929 (ipa_free_fn_summary): Likewise.
4930 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4931 (struct ipa_call_summary): Likewise.
4932 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4934 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4935 (estimate_size_after_inlining): Likewise.
4936 (estimate_growth): Likewise.
4937 (growth_likely_positive): Likewise.
4938 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4939 (inline_call): Likewise.
4940 * ipa-inline.c (caller_growth_limits): Likewise.
4941 (can_inline_edge_p): Likewise.
4942 (can_inline_edge_by_limits_p): Likewise.
4943 (compute_uninlined_call_time): Likewise.
4944 (compute_inlined_call_time): Likewise.
4945 (want_inline_small_function_p): Likewise.
4946 (edge_badness): Likewise.
4947 (update_caller_keys): Likewise.
4948 (update_callee_keys): Likewise.
4949 (inline_small_functions): Likewise.
4950 (inline_to_all_callers_1): Likewise.
4951 (dump_overall_stats): Likewise.
4952 (early_inline_small_functions): Likewise.
4953 (early_inliner): Likewise.
4954 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4955 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4956 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4957 * ipa-split.c (execute_split_functions): Likewise.
4958 * symbol-summary.h: Likewise.
4959 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4961 2018-06-19 Richard Biener <rguenther@suse.de>
4963 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4964 (vectorize_loops): ... here. Fix dbgcnt handling.
4965 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4967 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4970 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4971 ieee128 argument takes up only one (vector) register, not two (floating
4974 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4976 * gimplify.c (gimplify_init_constructor): Really never clear for an
4977 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4979 2018-06-19 Richard Biener <rguenther@suse.de>
4981 PR tree-optimization/86179
4982 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4983 after failed recognition.
4985 2018-06-18 Martin Sebor <msebor@redhat.com>
4988 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4989 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4990 Handle integer subtraction.
4991 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4992 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4994 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4996 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4997 param from rtx to rtx_insn *.
4998 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
5000 (frv_ifcvt_modify_insn): Likwise.
5001 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
5002 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
5003 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
5004 as_a <rtx_insn *> cast to local "unprotected_region" once
5005 it's been established that it's not NULL or pc_rtx.
5006 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
5007 param "sethi" from rtx to rtx_insn *.
5008 (nds32_group_float_insns): Likewise for param "insn".
5009 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
5011 (vax_output_int_subtract): Likewise.
5012 * config/vax/vax.c (vax_output_int_add): Likewise for param
5014 (vax_output_int_subtract): Likewise.
5015 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
5016 (emit_pattern_after): Likewise for param "after".
5017 (emit_insn_after): Likewise.
5018 (emit_jump_insn_after): Likewise.
5019 (emit_call_insn_after): Likewise.
5020 (emit_debug_insn_after): Likewise.
5021 (emit_pattern_before): Likewise for param "before".
5022 (emit_insn_before): Likewise.
5023 (emit_jump_insn_before): Likewise.
5024 * final.c (get_insn_template): Likewise for param "insn", removing
5026 * output.h (get_insn_template): Likewise for 2nd param.
5027 * rtl.h (emit_insn_before): Likewise.
5028 (emit_jump_insn_before): Likewise.
5029 (emit_debug_insn_before_noloc): Likewise.
5030 (emit_insn_after): Likewise.
5031 (emit_jump_insn_after): Likewise.
5032 (emit_call_insn_after): Likewise.
5033 (emit_debug_insn_after): Likewise.
5034 (set_insn_deleted): Likewise for param.
5036 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
5039 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
5040 floating point modes, so that IFmode is numerically greater than
5041 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
5042 to declare the ordering. This prevents IFmode from being
5043 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
5044 machine. Include rs6000-modes.h to share the fractional values
5045 between genmodes* and the rest of the compiler.
5049 * config/rs6000/rs6000-modes.h: New file.
5050 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
5051 meaning of rs6000_long_double_size so that 126..128 selects an
5052 appropriate 128-bit floating point type.
5053 (rs6000_option_override_internal): Likewise.
5054 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
5055 (TARGET_LONG_DOUBLE_128): Change the meaning of
5056 rs6000_long_double_size so that 126..128 selects an appropriate
5057 128-bit floating point type.
5058 (LONG_DOUBLE_TYPE_SIZE): Update comment.
5059 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
5060 source and destination to match the standard usage.
5061 (truncifkf2): Likewise.
5062 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
5063 ISA 2.07 to use an explicit clobber, instead of passing in a
5065 (copysign<mode>3_soft): Likewise.
5067 2018-06-18 David Malcolm <dmalcolm@redhat.com>
5069 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5070 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
5071 (vect_slp_analyze_instance_dependence): Likewise.
5072 (vect_enhance_data_refs_alignment): Likewise.
5073 (vect_analyze_data_refs_alignment): Likewise.
5074 (vect_slp_analyze_and_verify_instance_alignment
5075 (vect_analyze_data_ref_accesses): Likewise.
5076 (vect_prune_runtime_alias_test_list): Likewise.
5077 (vect_analyze_data_refs): Likewise.
5078 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
5079 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5080 (vect_analyze_scalar_cycles_1): Likewise.
5081 (vect_get_loop_niters): Likewise.
5082 (vect_analyze_loop_form_1): Likewise.
5083 (vect_update_vf_for_slp): Likewise.
5084 (vect_analyze_loop_operations): Likewise.
5085 (vect_analyze_loop): Likewise.
5086 (vectorizable_induction): Likewise.
5087 (vect_transform_loop): Likewise.
5088 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
5089 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5090 (vect_make_slp_decision): Likewise.
5091 (vect_detect_hybrid_slp): Likewise.
5092 (vect_slp_analyze_operations): Likewise.
5093 (vect_slp_bb): Likewise.
5094 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
5095 (vectorizable_bswap): Likewise.
5096 (vectorizable_call): Likewise.
5097 (vectorizable_simd_clone_call): Likewise.
5098 (vectorizable_conversion): Likewise.
5099 (vectorizable_assignment): Likewise.
5100 (vectorizable_shift): Likewise.
5101 (vectorizable_operation): Likewise.
5102 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
5104 2018-06-18 Martin Sebor <msebor@redhat.com>
5106 PR tree-optimization/81384
5107 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
5108 * builtins.c (expand_builtin_strnlen): New function.
5109 (expand_builtin): Call it.
5110 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
5111 * builtins.def (BUILT_IN_STRNLEN): New.
5112 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
5113 Warn for bounds in excess of maximum object size.
5114 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
5115 single-value ranges. Handle strnlen.
5116 (handle_builtin_strlen): Handle strnlen.
5117 (strlen_check_and_optimize_stmt): Same.
5118 * doc/extend.texi (Other Builtins): Document strnlen.
5120 2018-06-18 Maya Rashish <coypu@sdf.org>
5122 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
5123 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
5124 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
5126 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
5128 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
5130 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5132 * tree.c (escaped_string::escape): Replace cast to char * by
5133 const_cast<char *> (unescaped).
5135 2018-06-18 Nick Clifton <nickc@redhat.com>
5138 * tree.c (escaped_string): New class. Converts an unescaped
5139 string into its escaped equivalent.
5140 (warn_deprecated_use): Use the new class to convert the
5141 deprecation message, if present.
5142 (test_escaped_strings): New self test.
5143 (test_c_tests): Add test_escaped_strings.
5144 * doc/extend.texi (deprecated): Add a note that the
5145 deprecation message is affected by the -fmessage-length
5146 option, and that control characters will be escaped.
5147 (#pragma GCC error): Document this pragma.
5148 (#pragma GCC warning): Likewise.
5149 * doc/invoke.texi (-fmessage-length): Document this option's
5150 effect on the #warning and #error preprocessor directives and
5151 the deprecated attribute.
5153 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
5155 * tree.c (decl_value_expr_lookup): Revert latest change.
5156 (decl_value_expr_insert): Likewise.
5158 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
5160 * gimplify.c (nonlocal_vlas): Delete.
5161 (nonlocal_vla_vars): Likewise.
5162 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
5164 (gimplify_body): Do not create and destroy nonlocal_vlas.
5165 * tree-nested.c: Include diagnostic.h.
5166 (use_pointer_in_frame): Tweak.
5167 (lookup_field_for_decl): Add assertion and declare the transformation.
5168 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
5169 internal error when the reference is in a wrong context. Do not
5170 create a debug decl by default.
5171 (note_nonlocal_block_vlas): Delete.
5172 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
5173 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
5174 create a debug decl by default.
5175 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
5176 call is in a wrong context.
5177 (fixup_vla_decls): New function.
5178 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
5179 debug variables were created.
5180 * tree.c (decl_value_expr_lookup): Add checking assertion.
5181 (decl_value_expr_insert): Likewise.
5183 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5186 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
5187 * tree-scalar-evolution.c (interpret_expr): Likewise.
5188 (expression_expensive_p): Likewise.
5189 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
5190 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
5191 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
5192 (ssa_defined_by_minus_one_stmt_p): New.
5194 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5197 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
5198 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
5199 * dojump.c (do_jump): Likewise.
5200 * expr.c (expand_expr_real_2): Check operand type's sign.
5201 * fold-const.c (const_unop): Handle ABSU_EXPR.
5202 (fold_abs_const): Likewise.
5203 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
5204 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
5205 (strip_sign_op_1): Likesise.
5206 * match.pd: Add new pattern to generate ABSU_EXPR.
5207 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
5208 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5209 * tree-eh.c (operation_could_trap_helper_p): Likewise.
5210 * tree-inline.c (estimate_operator_cost): Likewise.
5211 * tree-pretty-print.c (dump_generic_node): Likewise.
5212 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
5213 * tree.def (ABSU_EXPR): New.
5215 2018-06-16 Jakub Jelinek <jakub@redhat.com>
5218 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
5219 documented as preserved for backward compatibility only.
5220 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
5222 PR rtl-optimization/86108
5223 * bb-reorder.c (create_forwarder_block): Renamed to ...
5224 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
5225 jump from new landing pad to the second part.
5226 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
5229 2018-06-15 Jakub Jelinek <jakub@redhat.com>
5232 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
5233 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
5234 Only call store_expr for halves if the mode is the same.
5237 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
5238 Fix up comment formatting.
5240 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
5242 * typed-splay-tree.h (typed_splay_tree::remove): New function.
5243 (typed_splay_tree::closure,
5244 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
5245 (typed_splay_tree::typed_splay_tree,
5246 typed_splay_tree::operator =): Declared private.
5247 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
5248 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
5249 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
5250 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
5251 typed_splay_tree::splay_tree_splay,
5252 typed_splay_tree::splay_tree_foreach_helper,
5253 typed_splay_tree::splay_tree_insert,
5254 typed_splay_tree::splay_tree_remove,
5255 typed_splay_tree::splay_tree_lookup,
5256 typed_splay_tree::splay_tree_predecessor,
5257 typed_splay_tree::splay_tree_successor,
5258 typed_splay_tree::splay_tree_min,
5259 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
5260 (typed_splay_tree::root, typed_splay_tree::comp,
5261 typed_splay_tree::delete_key,
5262 typed_splay_tree::delete_value): New data members.
5263 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
5264 typed_splay_tree::remove.
5266 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
5268 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
5269 -mginv and -mno-ginv to the assembler.
5270 * config/mips/mips.opt (-mcrc): New option.
5272 * doc/invoke.text (-mcrc): Document.
5275 2018-06-15 Nick Clifton <nickc@redhat.com>
5278 * tree.c (escaped_string): New class. Converts an unescaped
5279 string into its escaped equivalent.
5280 (warn_deprecated_use): Use the new class to convert the
5281 deprecation message, if present.
5282 (test_escaped_strings): New self test.
5283 (test_c_tests): Add test_escaped_strings.
5284 * doc/extend.texi (deprecated): Add a note that the
5285 deprecation message is affected by the -fmessage-length
5286 option, and that control characters will be escaped.
5287 (#pragma GCC error): Document this pragma.
5288 (#pragma GCC warning): Likewise.
5289 * doc/invoke.texi (-fmessage-length): Document this option's
5290 effect on the #warning and #error preprocessor directives and
5291 the deprecated attribute.
5293 2018-06-15 Richard Biener <rguenther@suse.de>
5295 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
5296 here, also noting vector size used.
5297 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
5298 size used in MSG_OPTIMIZED_LOCATIONS dump.
5299 (pass_slp_vectorize::execute): Adjust.
5301 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
5304 * config/arc/arc.c (arc_return_address_register): Fix
5307 2018-06-15 Richard Biener <rguenther@suse.de>
5310 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
5311 leave useless conversion stripping to force_gimple_operand_gsi.
5312 (gimplify_build2): Likewise.
5313 (gimplify_build1): Likewise.
5315 2018-06-15 Richard Biener <rguenther@suse.de>
5318 * tree-cfg.c (move_stmt_op): unshare invariant addresses
5319 before adjusting their block.
5321 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
5323 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
5324 multilibs for *-*-rtems*.
5325 * config/riscv/t-rtems: New file.
5327 2018-06-14 Jakub Jelinek <jakub@redhat.com>
5330 * match.pd ((A +- CST1) +- CST2): Punt if last resort
5331 unsigned_type_for returns NULL.
5334 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
5335 subregs of multi-word pseudos unless the float mode has word size.
5337 2018-06-14 Richard Biener <rguenther@suse.de>
5340 * tree-vect-generic.c (build_word_mode_vector_type): Remove
5341 duplicate and harmful type_hash_canon.
5342 * tree.c (type_hash_canon): Assert we didn't find ourselves.
5344 2018-06-14 Richard Biener <rguenther@suse.de>
5347 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
5350 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
5352 * config/rtems.h (STDINT_LONG32): Define.
5354 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
5355 Prachi Godbole <prachi.godbole@imgtec.com>
5357 * config/mips/mips-cpus.def: Define P6600.
5358 * config/mips/mips-tables.opt: Regenerate.
5359 * config/mips/mips.c (mips_ucbranch_type): New enum.
5360 (mips_rtx_cost_data): Add support for P6600.
5361 (mips_issue_rate): Likewise.
5362 (mips_multipass_dfa_lookahead): Likewise.
5363 (mips_avoid_hazard): Likewise.
5364 (mips_reorg_process_insns): Likewise.
5365 (mips_classify_branch_p6600): New function.
5366 * config/mips/mips.h (TUNE_P6600): New define.
5367 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
5368 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
5369 * config/mips/mips.md: Include p6600.md.
5370 (processor): Add p6600.
5371 * config/mips/p6600.md: New file.
5372 * doc/invoke.texi: Add p6600 to supported architectures.
5374 2018-06-13 Martin Sebor <msebor@redhat.com>
5376 PR tree-optimization/86114
5377 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
5379 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
5381 2018-06-13 Richard Biener <rguenther@suse.de>
5383 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
5384 Properly set vector type of the intermediate stmt.
5385 * tree-vect-stmts.c (vectorizable_operation): The destination
5386 var always has vectype_out type.
5388 2018-06-13 Jeff Law <law@redhat.com>
5390 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
5391 integer 0 for argument to print_rtl_with_bb.
5392 (rl78_reorg): Likewise.
5394 2018-06-13 David Malcolm <dmalcolm@redhat.com>
5396 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
5397 from rtx to rtx_insn *.
5398 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
5400 (add_sched_insns_for_speculation): Likewise for local "target",
5401 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5402 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
5403 from rtx_insn ** to rtx_code_label **.
5404 (reorg_emit_nops): Likewise.
5405 (c6x_reorg): Likewise for local "call_labels".
5406 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
5408 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
5409 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
5410 the loops over LABEL_REFS.
5411 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
5413 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5414 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
5415 (split_branches): Strengthen local "olabel" from rtx to
5416 rtx_insn *, adding a safe_as_a cast.
5417 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
5419 (add_insn_after): Likewise for first two params.
5420 (add_insn_before): Likewise.
5421 (remove_insn): Likewise for param.
5422 (emit_pattern_before_noloc): Likewise for second and third params.
5423 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
5424 (emit_call_insn_before_noloc): Likewise.
5425 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
5427 (emit_barrier_before): Likewise.
5428 (emit_label_before): Strengthen "label" param from "rtx" to
5429 "rtx_code_label *". Strengthen "before" param from "rtx" to
5431 (emit_insn_after_1): Strengthen "after" param from "rtx" to
5433 (emit_pattern_after_noloc): Likewise.
5434 (emit_insn_after_noloc): Likewise.
5435 (emit_jump_insn_after_noloc): Likewise.
5436 (emit_call_insn_after_noloc): Likewise.
5437 (emit_debug_insn_after_noloc): Likewise.
5438 (emit_barrier_after): Likewise.
5439 (emit_label_after): Likewise for both params.
5440 (emit_pattern_after_setloc): Likewise for "after" param. Convert
5441 "loc" param from "int" to "location_t".
5442 (emit_insn_after_setloc): Likewise.
5443 (emit_jump_insn_after_setloc): Likewise.
5444 (emit_call_insn_after_setloc): Likewise.
5445 (emit_debug_insn_after_setloc): Likewise.
5446 (emit_pattern_before_setloc): Likewise for "before" param. Convert
5447 "loc" param from "int" to "location_t".
5448 (emit_pattern_before): Convert NULL_RTX to NULL.
5449 (emit_insn_before_setloc): Convert "loc" param from "int" to
5451 (emit_jump_insn_before_setloc): Likewise.
5452 (emit_call_insn_before_setloc): Likewise.
5453 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
5454 rtx_insn *. Convert "loc" param from "int" to "location_t".
5455 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
5456 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
5457 Convert 3rd param from "int" to "location_t".
5458 (emit_barrier_before, emit_barrier_after, next_real_insn):
5459 Strengthen param from rtx to rtx_insn *.
5460 (emit_label_before): Strengthen 1st param from "rtx" to
5461 "rtx_code_label *". Strengthen 2nd param from "rtx" to
5463 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
5464 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
5465 Strengthen 2nd param from "rtx" to "rtx_insn *".
5466 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
5467 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
5468 Likewise. Convert 3rd param from "int" to "location_t".
5469 (emit_label_after): Strengthen 1st param from "rtx" to
5471 (next_real_insn, remove_insn): Strengthen param from "rtx" to
5473 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
5474 from "rtx" to "rtx_insn *".
5476 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
5478 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
5479 bodies streamed in with -Q.
5480 * dumpfile.c (dump_files): Add lto-stream-out dump file.
5481 * dumpfile.h (tree_dump_index): Add lto_stream_out.
5482 * gimple-streamer-out.c: Include gimple-pretty-print.h
5483 (output_bb): Dump stmts streamed.
5484 * lto-section-out.c: Include print-tree.h
5485 (lto_begin_section): Dump sections created.
5486 (lto_output_decl_index): Dump decl encoded.
5487 * lto-streamer-out.c: Include print-tree.h
5488 (create_output_block): Dump output block created.
5489 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
5490 (output_function): Dump function output.
5491 (output_constructor): Dump constructor streamed.
5492 (write_global_stream): Output indexes encoded.
5493 (produce_asm_for_decls): Dump streams encoded.
5494 * lto-streamer.c (streamer_dump_file): New global var.
5495 * lto-streamer.h (streamer_dump_file): Declare.
5496 * passes.c (ipa_write_summaries): Initialize streamer dump.
5497 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
5500 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
5503 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
5504 offsets for register save directives. Emit a second batch of save
5505 directives, if need be, when the function accesses prior frames.
5507 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5509 * config/arc/fpu.md (fmasf4): Force operand to register.
5510 (fnmasf4): Likewise.
5512 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5514 * config/arc/arc-protos.h (arc_pad_return): Remove.
5515 * config/arc/arc.c (machine_function): Remove force_short_suffix
5517 (arc_print_operand): Adjust printing of '&'.
5518 (arc_verify_short): Remove conditional printing of short suffix.
5519 (arc_final_prescan_insn): Remove reference to size_reason.
5520 (pad_return): New function.
5521 (arc_reorg): Call pad_return.
5522 (arc_pad_return): Remove.
5523 (arc_init_machine_status): Remove reference to force_short_suffix.
5524 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
5525 (attr length): When attribute iscompact is true force to 2
5526 regardless; in the case of maybe check if we want to force the
5527 instruction to have 4 bytes length.
5528 (nopv): Change it to generate 4 byte long nop as well.
5529 (blockage): New pattern.
5530 (simple_return): Remove call to arc_pad_return.
5531 (p_return_i): Likewise.
5533 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5535 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
5537 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5539 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
5542 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5544 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
5545 for ARC700 and ARCv2.
5547 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
5550 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
5551 operands[2] instead of operands[1].
5554 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5556 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
5557 case, check whether the outer register overlaps an unallocatable
5558 register, not just whether it fits the required class.
5560 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5562 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5564 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
5565 (memrefs_conflict_p): Likewise.
5566 (init_alias_analysis): Likewise.
5567 * cfgexpand.c (expand_debug_expr): Likewise.
5568 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
5569 * cse.c (fold_rtx): Likewise.
5570 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
5571 * expr.c (emit_block_move_hints): Likewise.
5572 (clear_storage_hints, push_block, emit_push_insn): Likewise.
5573 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
5574 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
5575 (emit_group_store): Likewise.
5576 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
5577 to read the PRE/POST_MODIFY increment.
5578 * calls.c (store_one_arg): Use strip_offset.
5579 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
5581 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
5583 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
5584 (simplify_binary_operation_1): Extend CONST_INT handling to
5586 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
5587 than a HOST_WIDE_INT.
5588 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
5590 (adjust_mems, add_stores): Update accodingly.
5591 (vt_canonicalize_addr): Track polynomial offsets.
5592 (emit_note_insn_var_location): Likewise.
5593 (vt_add_function_parameter): Likewise.
5594 (vt_initialize): Likewise.
5596 2018-06-12 Jeff Law <law@redhat.com>
5598 * config.gcc (alpha*-*-freebsd*): Remove.
5599 * config/alpha/freebsd.h: Remove.
5601 2018-06-12 David Malcolm <dmalcolm@redhat.com>
5604 * spellcheck-tree.c (levenshtein_distance): Rename to...
5605 (get_edit_distance): ...this, and update for underlying renaming.
5606 * spellcheck-tree.h (levenshtein_distance): Rename to...
5607 (get_edit_distance): ...this.
5608 * spellcheck.c (levenshtein_distance): Rename to...
5609 (get_edit_distance): ...this. Convert from Levenshtein distance
5610 to Damerau-Levenshtein distance by supporting transpositions of
5611 adjacent characters. Rename "v1" to "v_next" and "v0" to
5613 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
5614 (selftest::test_edit_distance_unit_test_oneway): ...this, and
5615 update for underlying renaming.
5616 (selftest::levenshtein_distance_unit_test): Rename to...
5617 (selftest::test_get_edit_distance_unit): ...this, and update for
5618 underlying renaming.
5619 (selftest::test_find_closest_string): Add example from PR 69968
5620 where transposition helps
5621 (selftest::test_metric_conditions): Update for renaming.
5622 (selftest::test_metric_conditions): Likewise.
5623 (selftest::spellcheck_c_tests): Likewise.
5624 * spellcheck.h (levenshtein_distance): Rename both overloads to...
5625 (get_edit_distance): ...this.
5626 (best_match::consider): Update for renaming.
5628 2018-06-12 Martin Sebor <msebor@redhat.com>
5630 PR tree-optimization/85259
5631 * builtins.c (compute_objsize): Handle constant offsets.
5632 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
5633 true iff a warning has been issued.
5634 * gimple.h (gimple_nonartificial_location): New function.
5635 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
5636 gimple_nonartificial_location and handle -Wno-system-headers.
5637 (handle_builtin_stxncpy): Same.
5639 2018-06-12 Martin Sebor <msebor@redhat.com>
5642 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
5644 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
5646 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5647 BUILTIN_VEC_XST entries for pointer to double and long long.
5649 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
5652 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5654 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5657 2018-06-12 Martin Liska <mliska@suse.cz>
5659 * doc/options.texi: Document IntegerRange.
5661 2018-06-12 Martin Liska <mliska@suse.cz>
5663 * config/i386/i386.opt: Make MPX-related options as Deprecated.
5664 * opt-functions.awk: Handle Deprecated flag.
5665 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
5667 (read_cmdline_option): Report warning for a deprecated option.
5668 * opts.h (struct cl_option): Add new field cl_deprecated.
5669 (CL_ERR_DEPRECATED): New.
5671 2018-06-12 Martin Liska <mliska@suse.cz>
5673 * doc/options.texi: Document Deprecated option flag.
5675 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5677 * config/arc/arc-arch.h (arc_extras): New enum.
5678 (arc_cpu_t):Add field extra.
5679 (arc_cpu_types): Consider the extras.
5680 * config/arc/arc-cpus.def: Add extras info.
5681 * config/arc/arc-opts.h (processor_type): Consider extra field.
5682 * config/arc/arc.c (arc_override_options): Handle extra field.
5684 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5686 * config/arc/arc-arch.h: Update ARC_OPTX macro.
5687 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
5689 * config/arc/arc.c (arc_init): Update pic warning.
5690 (irq_range): Update irq range parsing warnings.
5691 (arc_override_options): Update various warning messages.
5692 (arc_handle_aux_attribute): Likewise.
5694 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
5696 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
5698 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5700 * doc/sourcebuild.texi: Document usage of line number 0 in verify
5701 compiler messages directives.
5703 2018-06-12 Matthew Fortune <mfortune@gmail.com>
5705 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
5706 * config/mips/mips-tables.opt: Regenerate.
5707 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
5709 * doc/invoke.texi: Document -march=i6500.
5711 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
5713 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
5714 (i6400_gpmul): Add cpu_unit.
5715 (i6400_gpdiv): Likewise.
5716 (i6400_msa_add_d): Update reservations.
5717 (i6400_msa_int_add) Likewise.
5718 (i6400_msa_short_logic3) Likewise.
5719 (i6400_msa_short_logic2) Likewise.
5720 (i6400_msa_short_logic) Likewise.
5721 (i6400_msa_move) Likewise.
5722 (i6400_msa_cmp) Likewise.
5723 (i6400_msa_short_float2) Likewise.
5724 (i6400_msa_div_d) Likewise.
5725 (i6400_msa_long_logic1) Likewise.
5726 (i6400_msa_long_logic2) Likewise.
5727 (i6400_msa_mult) Likewise.
5728 (i6400_msa_long_float2) Likewise.
5729 (i6400_msa_long_float4) Likewise.
5730 (i6400_msa_long_float5) Likewise.
5731 (i6400_msa_long_float8) Likewise.
5732 (i6400_fpu_fadd): Include frint type.
5733 (i6400_fpu_store): New define_insn_reservation.
5734 (i6400_fpu_load): Likewise.
5735 (i6400_fpu_move): Likewise.
5736 (i6400_fpu_fcmp): Likewise.
5737 (i6400_fpu_fmadd): Likewise.
5738 (i6400_int_mult): Include imul3nc type and update reservation.
5739 (i6400_int_div): Include idiv3 type and update reservation.
5740 (i6400_int_load): Update to check type not move_type.
5741 (i6400_int_store): Likewise.
5742 (i6400_int_prefetch): Set zero latency.
5744 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
5746 * gcc.c: Document new %@{...} sequence.
5747 (LINK_COMMAND_SPEC): Use it for the -L switches.
5748 (cpp_unique_options): Use it for the -I switches.
5749 (at_file_argbuf): New global variable.
5750 (in_at_file): Likewise.
5751 (alloc_args): Create at_file_argbuf.
5752 (clear_args): Truncate at_file_argbuf.
5753 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
5754 (open_at_file): New function.
5755 (close_at_file): Likewise.
5756 (create_at_file): Delete.
5757 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
5760 (validate_switches_from_spec): Deal with %@{...} sequence.
5761 (validate_switches): Likewise.
5762 (driver::finalize): Call clear_args.
5764 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5766 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5768 2018-06-11 Martin Sebor <msebor@redhat.com>
5770 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5771 (Wno-class-memaccess): Add @opindex.
5772 (Wno-templates, Wno-multiple-inheritance): Same.
5773 (Wno-virtual-inheritance, Wno-namespaces): Same.
5774 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5775 (Wno-format-overflow, Wno-format-truncation): Same.
5776 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5777 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5778 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5779 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5780 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5781 (Wno-misspelled-isr): Same.
5783 2018-06-11 Martin Sebor <msebor@redhat.com>
5785 * PR tree-optimization/86083
5786 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5788 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5790 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5792 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5795 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5796 on the correct operand.
5797 (*movdi_internal64): Ditto.
5799 2018-06-11 Martin Liska <mliska@suse.cz>
5801 PR tree-optimization/86089
5802 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5804 2018-06-11 Julia Koval <julia.koval@intel.com>
5806 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5807 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5808 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5810 2018-06-11 Olivier Hainque <hainque@adacore.com>
5812 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5813 for Ada with strict dwarf2.
5815 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5818 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5821 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5823 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5825 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5827 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5830 2018-06-08 Martin Liska <mliska@suse.cz>
5832 * tree-cfg.h (debug_function): Fix argument type to match
5835 2018-06-08 Martin Liska <mliska@suse.cz>
5837 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5838 Remove usage of MPX-related (and removed) fields.
5839 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5841 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5843 * cfg.c (debug): Use TDF_NONE rather than 0.
5844 * cfghooks.c (debug): Likewise.
5845 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5846 (struct dump_option_value_info): Convert to...
5847 (struct kv_pair): ...this template type.
5848 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5850 (optinfo_verbosity_options): Likewise.
5851 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5853 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5854 than int for "optgroup_flags" param.
5855 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5857 (dump_dec): Likewise.
5858 (dump_finish): Use TDF_NONE rather than 0.
5859 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5860 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5861 than 0. Update for change to option_ptr.
5862 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5863 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5864 0. Update for changes to optinfo_verbosity_options and
5866 (opt_info_switch_p): Convert optgroup_flags from int to
5868 (dump_basic_block): Use dump_flags_t rather than int
5869 for "dump_kind" param.
5870 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5871 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5872 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5873 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5874 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5875 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5876 TDF_NONE): Convert from macros to...
5877 (enum dump_flag): ...this new enum.
5878 (dump_flags_t): Update to use enum.
5879 (operator|, operator&, operator~, operator|=, operator&=):
5880 Implement for dump_flags_t.
5881 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5882 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5883 Convert from macros to...
5884 (enum optgroup_flag): ...this new enum.
5885 (optgroup_flags_t): New typedef.
5886 (operator|, operator|=): Implement for optgroup_flags_t.
5887 (struct dump_file_info): Convert field "alt_flags" to
5888 dump_flags_t. Convert field "optgroup_flags" to
5890 (dump_basic_block): Use dump_flags_t rather than int for param.
5891 (dump_generic_expr_loc): Likewise.
5892 (dump_dec): Likewise.
5893 (dump_register): Convert param "optgroup_flags" to
5895 (opt_info_enable_passes): Likewise.
5896 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5898 * gimple-pretty-print.c (debug): Likewise.
5899 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5900 (merged_store_group::apply_stores): Likewise.
5901 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5902 * gimple.c (verify_gimple_pp): Likewise.
5903 * graphite-poly.c (print_pbb_body): Likewise.
5904 * passes.c (pass_manager::register_one_dump_file): Convert
5905 local "optgroup_flags" to optgroup_flags_t.
5906 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5908 (debug_body): Likewise.
5909 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5910 to optgroup_flags_t.
5911 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5913 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5914 (convert_mult_to_fma): Likewise.
5915 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5916 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5917 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5918 "dump_kind" to dump_flags_t.
5920 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5922 * config/rs6000/rs6000.c (min, max): Delete.
5924 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5926 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5927 -mabi=spe and -mabi=no-spe.
5929 2018-06-08 Martin Liska <mliska@suse.cz>
5931 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5932 where we expect an existing summary.
5934 2018-06-08 Martin Liska <mliska@suse.cz>
5936 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5937 * ipa-inline.h (estimate_edge_growth): Likewise.
5939 2018-06-08 Martin Liska <mliska@suse.cz>
5941 * cgraph.c (function_version_hasher::hash): Use
5942 cgraph_node::get_uid ().
5943 (function_version_hasher::equal):
5944 * cgraph.h (cgraph_node::get_uid): New method.
5945 * ipa-inline.c (update_caller_keys): Use
5946 cgraph_node::get_uid ().
5947 (update_callee_keys): Likewise.
5948 * ipa-utils.c (searchc): Likewise.
5949 (ipa_reduced_postorder): Likewise.
5950 * lto-cgraph.c (input_node): Likewise.
5951 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5952 * symbol-summary.h (symtab_insertion): Likewise.
5953 (symtab_removal): Likewise.
5954 (symtab_duplication): Likewise.
5955 * tree-pretty-print.c (dump_function_header): Likewise.
5956 * tree-sra.c (convert_callers_for_node): Likewise.
5958 2018-06-08 Martin Liska <mliska@suse.cz>
5960 * cgraph.c (symbol_table::create_edge): Always assign a new
5962 (symbol_table::free_edge): Do not recycle numbers.
5963 * cgraph.h (cgraph_edge::get): New method.
5964 * symbol-summary.h (symtab_removal): Use it.
5965 (symtab_duplication): Likewise.
5966 (call_summary::hashable_uid): Remove.
5968 2018-06-08 Martin Liska <mliska@suse.cz>
5970 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5971 (initialize_growth_caches): Remove.
5972 (free_growth_caches): Likewise.
5973 (do_estimate_edge_time): Use edge_growth_cache.
5974 (do_estimate_edge_size): Likewise.
5975 (do_estimate_edge_hints): Likewise.
5976 * ipa-inline.c (reset_edge_caches): Likewise.
5977 (recursive_inlining): Likewise.
5978 (inline_small_functions): Likewise.
5979 * ipa-inline.h (initialize_growth_caches): Remove.
5980 (estimate_edge_size): Likewise.
5981 (estimate_edge_time): Likewise.
5982 (estimate_edge_hints): Likewise.
5983 (reset_edge_growth_cache): Likewise.
5984 * symbol-summary.h (call_summary::remove): New method.
5986 2018-06-08 Martin Liska <mliska@suse.cz>
5988 * ipa-cp.c (class edge_clone_summary): New summary.
5989 (grow_edge_clone_vectors): Remove.
5990 (ipcp_edge_duplication_hook): Remove.
5991 (class edge_clone_summary_t): New call_summary class.
5992 (ipcp_edge_removal_hook): Remove.
5993 (edge_clone_summary_t::duplicate): New function.
5994 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5995 (create_specialized_node): Likewise.
5996 (ipcp_driver): Initialize edge_clone_summaries and do not
5999 2018-06-08 Martin Liska <mliska@suse.cz>
6001 * symbol-summary.h (get): New function.
6002 (call_summary::m_initialize_when_cloning): New class member.
6004 2018-06-08 Martin Liska <mliska@suse.cz>
6006 * cgraph.c (cgraph_node::remove): Do not recycle uid.
6007 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
6008 (symbol_table::allocate_cgraph_symbol): Do not set uid.
6009 * passes.c (uid_hash_t): Record removed_nodes by their uids.
6010 (remove_cgraph_node_from_order): Use the removed_nodes set.
6011 (do_per_function_toporder): Likwise.
6012 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
6013 instead of summary_uid.
6014 (symtab_removal): Likewise.
6015 (symtab_duplication): Likewise.
6017 2018-06-08 Martin Liska <mliska@suse.cz>
6019 * ipa-cp.c (ipcp_store_bits_results): Use
6020 ipcp_transformation_sum.
6021 (ipcp_store_vr_results): Likewise.
6022 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
6024 (ipcp_transformation_initialize): ... this.
6025 (ipa_set_node_agg_value_chain):
6026 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
6027 (write_ipcp_transformation_info): Likewise.
6028 (read_ipcp_transformation_info): Likewise.
6029 (ipcp_update_bits): Likewise.
6030 (ipcp_update_vr): Likewise.
6031 (ipcp_transform_function): Likewise.
6032 * ipa-prop.h: Rename ipcp_transformation_summary to
6033 ipcp_transformation.
6034 (class ipcp_transformation_t): New function summary.
6035 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
6036 (ipa_get_agg_replacements_for_node): Likewise.
6038 2018-06-08 Martin Liska <mliska@suse.cz>
6040 * ipa-pure-const.c (struct funct_state_d): Do it class instead
6042 (class funct_state_summary_t): New function_summary class.
6043 (has_function_state): Remove.
6044 (get_function_state): Likewise.
6045 (set_function_state): Likewise.
6046 (add_new_function): Likewise.
6047 (funct_state_summary_t::insert): New function.
6048 (duplicate_node_data): Remove.
6049 (remove_node_data): Remove.
6050 (funct_state_summary_t::duplicate): New function.
6051 (register_hooks): Create new funct_state_summaries.
6052 (pure_const_generate_summary): Use it.
6053 (pure_const_write_summary): Likewise.
6054 (pure_const_read_summary): Likewise.
6055 (propagate_pure_const): Likewise.
6056 (propagate_nothrow): Likewise.
6057 (dump_malloc_lattice): Likewise.
6058 (propagate_malloc): Likewise.
6059 (execute): Do not register hooks, just remove summary
6061 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
6064 2018-06-08 Martin Liska <mliska@suse.cz>
6066 * ipa-reference.c (remove_node_data): Remove.
6067 (duplicate_node_data): Likewise.
6068 (class ipa_ref_var_info_summary_t): New class.
6069 (class ipa_ref_opt_summary_t): Likewise.
6070 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
6071 (get_reference_optimization_summary): Use
6072 ipa_ref_opt_sum_summaries.
6073 (set_reference_vars_info): Remove.
6074 (set_reference_optimization_summary): Likewise.
6075 (ipa_init): Create summaries.
6076 (init_function_info): Use function summary.
6077 (ipa_ref_opt_summary_t::duplicate): New function.
6078 (ipa_ref_opt_summary_t::remove): New function.
6079 (get_read_write_all_from_node): Fix GNU coding style.
6080 (propagate): Use function summary.
6081 (write_node_summary_p): Fix GNU coding style.
6082 (stream_out_bitmap): Likewise.
6083 (ipa_reference_read_optimization_summary): Use function summary.
6084 (ipa_reference_c_finalize): Do not release hooks.
6086 2018-06-08 Martin Liska <mliska@suse.cz>
6088 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
6089 (analyze_function_body): Extract multiple calls of get_create.
6090 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6091 * ipa-inline.c (recursive_inlining): Use ::get method.
6092 * ipa-inline.h (estimate_edge_growth): Likewise.
6094 2018-06-08 Martin Liska <mliska@suse.cz>
6096 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
6098 (hsa_function_summary::hsa_function_summary): Use the new enum
6100 (hsa_gpu_implementation_p): Use hsa_summaries::get.
6101 * hsa-gen.c (hsa_get_host_function): Likewise.
6102 (get_brig_function_name): Likewise.
6103 * ipa-hsa.c (process_hsa_functions): Likewise.
6104 (ipa_hsa_write_summary): Likewise.
6105 * symbol-summary.h (symtab_duplication): Use ::get function/
6106 (get): New function.
6108 2018-06-08 Martin Liska <mliska@suse.cz>
6110 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
6112 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
6113 (hsa_register_kernel): Likewise.
6114 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
6115 * hsa-gen.c (hsa_get_host_function): Likewise.
6116 (get_brig_function_name): Likewise.
6117 (generate_hsa): Likewise.
6118 (pass_gen_hsail::execute): Likewise.
6119 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
6120 (devirtualization_time_bonus): Likewise.
6121 (ipcp_propagate_stage): Likewise.
6122 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6123 (edge_set_predicate): Likewise.
6124 (evaluate_conditions_for_known_args): Likewise.
6125 (evaluate_properties_for_edge): Likewise.
6126 (ipa_fn_summary::reset): Likewise.
6127 (ipa_fn_summary_t::duplicate): Likewise.
6128 (dump_ipa_call_summary): Likewise.
6129 (ipa_dump_fn_summary): Likewise.
6130 (analyze_function_body): Likewise.
6131 (compute_fn_summary): Likewise.
6132 (estimate_edge_devirt_benefit): Likewise.
6133 (estimate_edge_size_and_time): Likewise.
6134 (estimate_calls_size_and_time): Likewise.
6135 (estimate_node_size_and_time): Likewise.
6136 (inline_update_callee_summaries): Likewise.
6137 (remap_edge_change_prob): Likewise.
6138 (remap_edge_summaries): Likewise.
6139 (ipa_merge_fn_summary_after_inlining): Likewise.
6140 (ipa_update_overall_fn_summary): Likewise.
6141 (read_ipa_call_summary): Likewise.
6142 (inline_read_section): Likewise.
6143 (write_ipa_call_summary): Likewise.
6144 (ipa_fn_summary_write): Likewise.
6145 (ipa_free_fn_summary): Likewise.
6146 * ipa-hsa.c (process_hsa_functions): Likewise.
6147 (ipa_hsa_write_summary): Likewise.
6148 (ipa_hsa_read_section): Likewise.
6149 * ipa-icf.c (sem_function::merge): Likewise.
6150 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6151 (do_estimate_edge_time): Likewise.
6152 (estimate_size_after_inlining): Likewise.
6153 (estimate_growth): Likewise.
6154 (growth_likely_positive): Likewise.
6155 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6156 (inline_call): Likewise.
6157 * ipa-inline.c (caller_growth_limits): Likewise.
6158 (can_inline_edge_p): Likewise.
6159 (can_inline_edge_by_limits_p): Likewise.
6160 (compute_uninlined_call_time): Likewise.
6161 (compute_inlined_call_time): Likewise.
6162 (want_inline_small_function_p): Likewise.
6163 (edge_badness): Likewise.
6164 (update_caller_keys): Likewise.
6165 (update_callee_keys): Likewise.
6166 (recursive_inlining): Likewise.
6167 (inline_small_functions): Likewise.
6168 (inline_to_all_callers_1): Likewise.
6169 (dump_overall_stats): Likewise.
6170 (early_inline_small_functions): Likewise.
6171 (early_inliner): Likewise.
6172 * ipa-inline.h (estimate_edge_growth): Likewise.
6173 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6174 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6175 * ipa-prop.h (IPA_NODE_REF): Likewise.
6176 (IPA_EDGE_REF): Likewise.
6177 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6178 (propagate_malloc): Likewise.
6179 * ipa-split.c (execute_split_functions): Likewise.
6180 * symbol-summary.h: Rename get to get_create.
6182 (get_create): Likewise.
6183 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6185 2018-06-08 Martin Liska <mliska@suse.cz>
6187 * symbol-summary.h (release): Move definition out of class
6189 (symtab_removal): Likewise.
6190 (symtab_duplication): Likewise.
6192 2018-06-08 Martin Liska <mliska@suse.cz>
6194 * symbol-summary.h (function_summary): Move constructor
6195 implementation out of class declaration.
6196 (release): Likewise.
6197 (symtab_insertion): Likewise.
6198 (symtab_removal): Likewise.
6199 (symtab_duplication): Likewise.
6202 2018-06-08 Martin Liska <mliska@suse.cz>
6204 * Makefile.in: Remove support for MPX (macros, related functions,
6205 fields in cgraph_node, ...).
6206 * builtin-types.def (BT_BND): Likewise.
6207 (BT_FN_BND_CONST_PTR): Likewise.
6208 (BT_FN_CONST_PTR_BND): Likewise.
6209 (BT_FN_VOID_PTR_BND): Likewise.
6210 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
6211 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
6212 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
6213 (expand_builtin_mempcpy_with_bounds): Likewise.
6214 (expand_builtin_memset_with_bounds): Likewise.
6215 (expand_builtin_memset_args): Likewise.
6216 (std_expand_builtin_va_start): Likewise.
6217 (expand_builtin): Likewise.
6218 (expand_builtin_with_bounds): Likewise.
6219 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
6220 (DEF_LIB_BUILTIN_CHKP): Likewise.
6221 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
6222 (DEF_CHKP_BUILTIN): Likewise.
6223 (BUILT_IN_MEMCPY): Likewise.
6224 (BUILT_IN_MEMMOVE): Likewise.
6225 (BUILT_IN_MEMPCPY): Likewise.
6226 (BUILT_IN_MEMSET): Likewise.
6227 (BUILT_IN_STPCPY): Likewise.
6228 (BUILT_IN_STRCAT): Likewise.
6229 (BUILT_IN_STRCHR): Likewise.
6230 (BUILT_IN_STRCPY): Likewise.
6231 (BUILT_IN_STRLEN): Likewise.
6232 (BUILT_IN_MEMCPY_CHK): Likewise.
6233 (BUILT_IN_MEMMOVE_CHK): Likewise.
6234 (BUILT_IN_MEMPCPY_CHK): Likewise.
6235 (BUILT_IN_MEMSET_CHK): Likewise.
6236 (BUILT_IN_STPCPY_CHK): Likewise.
6237 (BUILT_IN_STRCAT_CHK): Likewise.
6238 (BUILT_IN_STRCPY_CHK): Likewise.
6239 * calls.c (store_bounds): Likewise.
6240 (emit_call_1): Likewise.
6241 (special_function_p): Likewise.
6242 (maybe_warn_nonstring_arg): Likewise.
6243 (initialize_argument_information): Likewise.
6244 (finalize_must_preallocate): Likewise.
6245 (compute_argument_addresses): Likewise.
6246 (expand_call): Likewise.
6247 * cfgexpand.c (expand_call_stmt): Likewise.
6248 (expand_return): Likewise.
6249 (expand_gimple_stmt_1): Likewise.
6250 (pass_expand::execute): Likewise.
6251 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
6252 (cgraph_node::remove): Likewise.
6253 (cgraph_node::dump): Likewise.
6254 (cgraph_node::verify_node): Likewise.
6255 * cgraph.h (chkp_function_instrumented_p): Likewise.
6256 (symtab_node::get_alias_target): Likewise.
6257 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
6258 (cgraph_local_p): Likewise.
6259 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
6260 (cgraph_edge::rebuild_references): Likewise.
6261 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
6262 (walk_polymorphic_call_targets): Likewise.
6263 (cgraph_node::expand_thunk): Likewise.
6264 (symbol_table::output_weakrefs): Likewise.
6265 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
6266 (ix86_handle_option): Likewise.
6267 * config/i386/constraints.md: Likewise.
6268 * config/i386/i386-builtin-types.def (BND): Likewise.
6272 * config/i386/i386-builtin.def (BDESC_END): Likewise.
6273 (BDESC_FIRST): Likewise.
6275 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
6276 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
6277 * config/i386/i386.c (enum reg_class): Likewise.
6278 (ix86_target_string): Likewise.
6279 (ix86_option_override_internal): Likewise.
6280 (ix86_conditional_register_usage): Likewise.
6281 (ix86_valid_target_attribute_inner_p): Likewise.
6282 (ix86_set_indirect_branch_type): Likewise.
6283 (ix86_set_current_function): Likewise.
6284 (ix86_function_arg_regno_p): Likewise.
6285 (init_cumulative_args): Likewise.
6286 (ix86_function_arg_advance): Likewise.
6287 (ix86_function_arg): Likewise.
6288 (ix86_pass_by_reference): Likewise.
6289 (ix86_function_value_regno_p): Likewise.
6290 (ix86_function_value_1): Likewise.
6291 (ix86_function_value_bounds): Likewise.
6292 (ix86_return_in_memory): Likewise.
6293 (ix86_setup_incoming_vararg_bounds): Likewise.
6294 (ix86_va_start): Likewise.
6295 (indirect_thunk_need_prefix): Likewise.
6296 (print_reg): Likewise.
6297 (ix86_print_operand): Likewise.
6298 (ix86_expand_call): Likewise.
6299 (ix86_output_function_return): Likewise.
6300 (reg_encoded_number): Likewise.
6301 (BDESC_VERIFYS): Likewise.
6302 (ix86_init_mpx_builtins): Likewise.
6303 (ix86_init_builtins): Likewise.
6304 (ix86_emit_cmove): Likewise.
6305 (ix86_emit_move_max): Likewise.
6306 (ix86_expand_builtin): Likewise.
6307 (ix86_builtin_mpx_function): Likewise.
6308 (ix86_get_arg_address_for_bt): Likewise.
6309 (ix86_load_bounds): Likewise.
6310 (ix86_store_bounds): Likewise.
6311 (ix86_load_returned_bounds): Likewise.
6312 (ix86_store_returned_bounds): Likewise.
6313 (ix86_class_likely_spilled_p): Likewise.
6314 (ix86_hard_regno_mode_ok): Likewise.
6315 (x86_order_regs_for_local_alloc): Likewise.
6316 (ix86_mitigate_rop): Likewise.
6317 (ix86_bnd_prefixed_insn_p): Likewise.
6318 (ix86_mpx_bound_mode): Likewise.
6319 (ix86_make_bounds_constant): Likewise.
6320 (ix86_initialize_bounds): Likewise.
6321 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
6322 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
6323 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
6324 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
6325 (TARGET_CHKP_BOUND_MODE): Likewise.
6326 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
6327 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
6328 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
6329 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
6330 * config/i386/i386.h (TARGET_MPX): Likewise.
6331 (TARGET_MPX_P): Likewise.
6332 (VALID_BND_REG_MODE): Likewise.
6333 (FIRST_BND_REG): Likewise.
6334 (LAST_BND_REG): Likewise.
6335 (enum reg_class): Likewise.
6336 (BND_REG_P): Likewise.
6337 (BND_REGNO_P): Likewise.
6338 (BNDmode): Likewise.
6339 (ADJUST_INSN_LENGTH): Likewise.
6340 * config/i386/i386.md: Likewise.
6341 * config/i386/i386.opt: Likewise.
6342 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
6343 (defined): Likewise.
6344 (LINK_MPX): Likewise.
6345 (MPX_SPEC): Likewise.
6346 (LIBMPX_SPEC): Likewise.
6347 (LIBMPXWRAPPERS_SPEC): Likewise.
6348 (CHKP_SPEC): Likewise.
6349 * config/i386/predicates.md: Likewise.
6350 * dbxout.c (dbxout_type): Likewise.
6351 * doc/extend.texi: Likewise.
6352 * doc/invoke.texi: Likewise.
6353 * doc/md.texi: Likewise.
6354 * doc/tm.texi: Likewise.
6355 * doc/tm.texi.in: Likewise.
6356 * dwarf2out.c (is_base_type): Likewise.
6357 (gen_formal_types_die): Likewise.
6358 (gen_subprogram_die): Likewise.
6359 (gen_type_die_with_usage): Likewise.
6360 (gen_decl_die): Likewise.
6361 (dwarf2out_late_global_decl): Likewise.
6362 * expr.c (expand_assignment): Likewise.
6363 (emit_storent_insn): Likewise.
6364 (store_expr_with_bounds): Likewise.
6365 (store_expr): Likewise.
6366 (expand_expr_real_1): Likewise.
6367 * expr.h (store_expr_with_bounds): Likewise.
6368 * function.c (use_register_for_decl): Likewise.
6369 (struct bounds_parm_data): Likewise.
6370 (assign_parms_augmented_arg_list): Likewise.
6371 (assign_parm_find_entry_rtl): Likewise.
6372 (assign_parm_is_stack_parm): Likewise.
6373 (assign_parm_load_bounds): Likewise.
6374 (assign_bounds): Likewise.
6375 (assign_parms): Likewise.
6376 (expand_function_start): Likewise.
6377 * gcc.c (CHKP_SPEC): Likewise.
6378 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
6379 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
6380 (wrestrict_dom_walker::check_call): Likewise.
6381 * gimple.c (gimple_build_call_from_tree): Likewise.
6382 * gimple.h (enum gf_mask): Likewise.
6383 (gimple_call_with_bounds_p): Likewise.
6384 (gimple_call_set_with_bounds): Likewise.
6385 * gimplify.c (gimplify_init_constructor): Likewise.
6386 * ipa-cp.c (initialize_node_lattices): Likewise.
6387 (propagate_constants_across_call): Likewise.
6388 (find_more_scalar_values_for_callers_subset): Likewise.
6389 * ipa-hsa.c (process_hsa_functions): Likewise.
6390 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
6391 * ipa-icf.c (sem_function::merge): Likewise.
6392 * ipa-inline.c (early_inliner): Likewise.
6393 * ipa-pure-const.c (warn_function_noreturn): Likewise.
6394 (warn_function_cold): Likewise.
6395 (propagate_pure_const): Likewise.
6396 * ipa-ref.h (enum GTY): Likewise.
6397 * ipa-split.c (find_retbnd): Likewise.
6398 (consider_split): Likewise.
6399 (split_function): Likewise.
6400 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6401 * ipa.c (walk_polymorphic_call_targets): Likewise.
6402 (symbol_table::remove_unreachable_nodes): Likewise.
6403 (process_references): Likewise.
6404 (cgraph_build_static_cdtor_1): Likewise.
6405 * lto-cgraph.c (lto_output_node): Likewise.
6406 (output_refs): Likewise.
6407 (compute_ltrans_boundary): Likewise.
6408 (input_overwrite_node): Likewise.
6409 (input_node): Likewise.
6410 (input_cgraph_1): Likewise.
6411 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
6412 * passes.c (pass_manager::execute_early_local_passes): Likewise.
6413 (class pass_chkp_instrumentation_passes): Likewise.
6414 (make_pass_chkp_instrumentation_passes): Likewise.
6415 * passes.def: Likewise.
6416 * rtl.h (struct GTY): Likewise.
6417 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
6418 * stor-layout.c (layout_type): Likewise.
6419 * symtab.c: Likewise.
6420 * target.def: Likewise.
6421 * targhooks.c (default_chkp_bound_type): Likewise.
6422 (default_chkp_bound_mode): Likewise.
6423 (default_builtin_chkp_function): Likewise.
6424 (default_chkp_function_value_bounds): Likewise.
6425 (default_chkp_make_bounds_constant): Likewise.
6426 (default_chkp_initialize_bounds): Likewise.
6427 * targhooks.h (default_chkp_bound_type): Likewise.
6428 (default_chkp_bound_mode): Likewise.
6429 (default_builtin_chkp_function): Likewise.
6430 (default_chkp_function_value_bounds): Likewise.
6431 (default_chkp_make_bounds_constant): Likewise.
6432 (default_chkp_initialize_bounds): Likewise.
6433 * toplev.c (compile_file): Likewise.
6434 (process_options): Likewise.
6435 * tree-core.h (DEF_BUILTIN): Likewise.
6436 (DEF_BUILTIN_CHKP): Likewise.
6437 * tree-inline.c (declare_return_variable): Likewise.
6438 (remap_gimple_stmt): Likewise.
6439 (copy_bb): Likewise.
6440 (initialize_inlined_parameters): Likewise.
6441 (expand_call_inline): Likewise.
6442 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
6443 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
6444 (make_pass_ipa_chkp_produce_thunks): Likewise.
6445 (make_pass_chkp): Likewise.
6446 (make_pass_chkp_opt): Likewise.
6447 (make_pass_chkp_instrumentation_passes): Likewise.
6448 * tree-pretty-print.c (dump_generic_node): Likewise.
6449 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
6450 * tree-ssa-dce.c (propagate_necessity): Likewise.
6451 (eliminate_unnecessary_stmts): Likewise.
6452 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
6453 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6454 * tree-ssa-sccvn.h: Likewise.
6455 * tree-ssa-strlen.c (get_string_length): Likewise.
6456 (valid_builtin_call): Likewise.
6457 (adjust_last_stmt): Likewise.
6458 (handle_builtin_strchr): Likewise.
6459 (handle_builtin_strcpy): Likewise.
6460 (handle_builtin_stxncpy): Likewise.
6461 (handle_builtin_memcpy): Likewise.
6462 (handle_builtin_strcat): Likewise.
6463 (strlen_check_and_optimize_stmt): Likewise.
6464 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
6465 * tree-streamer-in.c: Likewise.
6466 * tree-streamer.c (record_common_node): Likewise.
6467 * tree.c (tree_code_size): Likewise.
6468 (wide_int_to_tree_1): Likewise.
6469 (type_contains_placeholder_1): Likewise.
6470 (build_common_tree_nodes): Likewise.
6471 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
6472 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
6473 (POINTER_BOUNDS_P): Likewise.
6474 (BOUNDED_TYPE_P): Likewise.
6475 (BOUNDED_P): Likewise.
6476 (CALL_WITH_BOUNDS_P): Likewise.
6477 (pointer_bounds_type_node): Likewise.
6478 * value-prof.c (gimple_ic): Likewise.
6479 * var-tracking.c (vt_add_function_parameters): Likewise.
6480 * varasm.c (make_decl_rtl): Likewise.
6481 (assemble_start_function): Likewise.
6482 (output_constant): Likewise.
6483 (maybe_assemble_visibility): Likewise.
6484 * varpool.c (ctor_for_folding): Likewise.
6485 * chkp-builtins.def: Remove.
6486 * ipa-chkp.c: Remove.
6487 * ipa-chkp.h: Remove.
6488 * rtl-chkp.c: Remove.
6489 * rtl-chkp.h: Remove.
6490 * tree-chkp-opt.c: Remove.
6491 * tree-chkp.c: Remove.
6492 * tree-chkp.h: Remove.
6494 2018-06-07 Carl Love <cel@us.ibm.com>
6496 * config/rs6000/vsx.md (vextract_fp_from_shorth,
6497 vextract_fp_from_shortl): Add BE support.
6499 2018-06-07 Paul Koning <ni1d@arrl.net>
6501 * compare-elim.c (try_merge_compare): Don't merge compare if
6502 address contains a side effect.
6503 (try_eliminate_compare): Likewise.
6505 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
6507 * config.gcc: Support "tremont".
6508 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
6509 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6511 * config/i386/i386.c (m_TREMONT): Define.
6512 (processor_target_table): Add "tremont".
6513 (PTA_TREMONT): Define.
6514 (ix86_lea_outperforms): Add TARGET_TREMONT.
6515 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
6516 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
6517 and M_INTEL_GOLDMONT_PLUS.
6518 (fold_builtin_cpu): Add "tremont".
6519 (ix86_add_stmt_cost): Add TARGET_TREMONT.
6520 (ix86_option_override_internal): Add "tremont".
6521 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
6522 (processor_type): Add PROCESSOR_TREMONT.
6523 * config/i386/x86-tune.def: Add m_TREMONT.
6524 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
6526 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6528 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
6529 symbol defined for msp430i* devices to be lower case.
6531 2018-06-07 Richard Biener <rguenther@suse.de>
6533 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
6534 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
6535 Properly wrap signed arithmetic if overflow wraps.
6537 2018-06-07 Jakub Jelinek <jakub@redhat.com>
6539 PR tree-optimization/69615
6540 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
6541 of a cast from a same precision integral SSA_NAME in a bb dominated
6542 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
6543 cast to utype if rhs2 has already a compatible type.
6545 2018-06-07 Richard Biener <rguenther@suse.de>
6547 PR tree-optimization/85935
6548 * graphite-scop-detection.c (find_params_in_bb): Analyze
6549 condition operands with respect to the correct loop. Assert
6550 the analysis doesn't fail.
6552 2018-06-04 Carl Love <cel@us.ibm.com>
6554 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
6555 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
6556 as it is slightly cheaper.
6557 (first_match_or_eos_index_<mode>):
6558 Calculate index using natural element order.
6559 (first_match_index_<mode>):
6560 Calculate index using natural element order.
6561 (first_match_or_eos_index_<mode>):
6562 Calculate index using natural order.
6563 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
6564 for BE and LE modes.
6565 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
6566 P9V_BUILTIN_VCLZLSBB_V16QI.
6567 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
6570 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6572 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
6573 indentation and line wrap for many prototypes. Add missing
6574 @smallexample directives around block of prototypes for vec_xl and
6577 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
6579 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
6580 track if we pass or return IEEE 128-bit floating point.
6581 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6582 C++ mangling that is compatible with GCC 8.1.
6583 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
6584 (init_cumulative_args): Note if we pass or return IEEE 128-bit
6585 floating point types.
6586 (rs6000_function_arg_advance_1): Likewise.
6587 (rs6000_mangle_type): Optionally generate mangled names that match
6588 what GCC 8.1 generated for IEEE 128-bit floating point types.
6589 (rs6000_globalize_decl_name): If we have an external function that
6590 passes or returns IEEE 128-bit types, generate a weak reference
6591 from the mangled name used in GCC 8.1 to the current mangled
6593 (rs6000_init_builtins): Make __ibm128 use the long double type if
6594 long double is IBM extended double. Make __float128 use the long
6595 double type if long double is IEEE 128-bit.
6598 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
6599 macro for __ibm128 built-in functions.
6600 (PACK_IF): Add __ibm128 pack/unpack functions.
6601 (UNPACK_IF): Likewise.
6602 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
6603 enable long double built-in functions if long double is IEEE
6604 128-bit floating point.
6605 (rs6000_invalid_builtin): Update long double built-in function
6607 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
6608 functions, adjust the built-in function to use the long double
6609 built-in function if __ibm128 and long double are the same type.
6610 * doc/extend.texi (PowerPC builtins): Update documention for
6611 __builtin_{,un}pack_longdouble. Add documentation for
6612 __builtin_{,un}pack_ibm128.
6614 2018-06-06 Jim Wilson <jimw@sifive.com>
6616 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
6617 (struct machine_function): New field interrupt_mode.
6618 (riscv_handle_type_attribute): New function. Add forward declaration.
6619 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
6620 (riscv_expand_epilogue): Check interrupt_mode field.
6621 (riscv_set_current_function): Check interrupt attribute args and
6622 set interrupt_mode field.
6623 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
6624 (riscv_sret, riscv_uret): New.
6625 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
6626 new arguments to interrupt attribute.
6628 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
6631 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
6632 Don't handle -mcpu=power8 if -mpower9-vector is also used.
6634 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6636 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
6637 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
6638 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6639 several redundant entries.
6641 2018-06-06 David Malcolm <dmalcolm@redhat.com>
6643 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
6644 type from "rtx" to "rtx_insn *".
6645 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
6646 for local "call_insn", removing cast.
6647 (ix86_expand_call): Likewise, introducing a "call_insn" local.
6649 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
6651 PR tree-optimization/86066
6652 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
6653 for BIT_INSERT_EXPR stores.
6655 2018-06-06 Richard Biener <rguenther@suse.de>
6657 PR tree-optimization/86062
6658 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6659 component refs ontop
6660 of to be offsetted base.
6662 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6664 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
6665 to be static and remove check on interrupt attribute name.
6667 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
6669 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6670 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
6672 2018-06-05 Steve Ellcey <sellcey@cavium.com>
6675 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
6677 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
6678 Remove second argument, change how error is called.
6679 (aarch64_layout_arg): Remove second argument from
6680 aarch64_err_no_fpadvsimd call.
6681 (aarch64_init_cumulative_args): Ditto.
6682 (aarch64_gimplify_va_arg_expr): Ditto.
6683 * config/aarch64/aarch64.md (mov<mode>): Ditto.
6685 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
6687 * config/i386/i386.md (simple_return_indirect_internal): New expander.
6688 (*simple_return_indirect_internal<mode>): Rename from
6689 simple_return_indirect_internal. Use W mode iterator.
6690 (rstorssp): New expander.
6691 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
6692 (clrssbsy): New expander.
6693 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
6695 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
6697 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
6699 (cmse_check_pointed_object): Likewise.
6701 2018-06-05 Martin Liska <mliska@suse.cz>
6703 PR gcov-profile/47618
6704 * doc/invoke.texi: Document how -fprofile-dir format
6707 2018-06-05 Richard Biener <rguenther@suse.de>
6709 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
6710 removal pretend DOM info isn't available so we do not update
6711 it and only remove edges, not dominated blocks. Actually free
6712 DOM info in case we removed something. Remove unreachable blocks.
6713 (mfb_keep_latches): Work with either DOM info or marked backedges.
6714 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
6715 first. Mark backedges if DOM info isn't available.
6716 (Re-)compute DOM info after cleanup_control_flow_pre.
6718 2018-06-05 Richard Biener <rguenther@suse.de>
6720 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
6721 (locus_discrim_hasher::hash): Adjust.
6722 (locus_discrim_hasher::equal): Likewise.
6723 (next_discriminator_for_locus): Work on line directly.
6724 (same_line_p): Pass in expanded locus1 as well.
6725 (assign_discriminators): Avoid redundant location expansions.
6727 2018-06-05 Richard Biener <rguenther@suse.de>
6729 PR tree-optimization/86046
6730 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
6731 if required after clearing TREE_ADDRESSABLE.
6733 2018-06-05 Richard Biener <rguenther@suse.de>
6735 PR tree-optimization/86047
6736 * tree-ssa-loop.c (for_each_index): Glob handling of all
6737 decls and constants and really handle all of them.
6739 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6742 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6743 qualifier_void_pointer and qualifier_const_void_pointer.
6744 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
6745 (arm_init_builtins): Handle the above.
6746 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
6747 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
6748 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
6751 2018-06-05 Martin Liska <mliska@suse.cz>
6753 * auto-profile.c (read_autofdo_file): Do not use
6754 gcov_ctr_summary struct.
6755 (afdo_callsite_hot_enough_for_early_inline): Likewise.
6756 * coverage.c (struct counts_entry): Likewise.
6757 (read_counts_file): Read just single summary entry.
6758 (get_coverage_counts): Use gcov_summary struct.
6759 * coverage.h (get_coverage_counts): Likewise.
6760 * gcov-dump.c (dump_working_sets): Likewise.
6761 (tag_summary): Dump just single summary.
6762 * gcov-io.c (gcov_write_summary): Write just histogram
6764 (gcov_read_summary): Read just single summary.
6765 (compute_working_sets): Use gcov_summary struct.
6766 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6767 of GCOV_COUNTERS_SUMMABLE.
6768 (GCOV_COUNTERS_SUMMABLE): Remove.
6769 (GCOV_FIRST_VALUE_COUNTER): Replace with
6770 GCOV_COUNTER_V_INTERVAL.
6771 (struct gcov_ctr_summary): Remove.
6772 (struct gcov_summary): Directly use fields of former
6774 (compute_working_sets): Use gcov_summary struct.
6775 * gcov.c (read_count_file): Do not use ctrs fields.
6776 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6778 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6780 * profile.c: Likewise.
6781 * profile.h: Likewise.
6783 2018-06-05 Martin Liska <mliska@suse.cz>
6785 PR gcov-profile/84846
6786 * gcov.c (output_lines): Print working directory only
6787 in intermediate format.
6789 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6791 * config/s390/s390-builtin-types.def: Add void function type.
6792 * config/s390/s390-builtins.def: Use the function type for the
6795 2018-06-04 Jim Wilson <jimw@sifive.com>
6797 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6799 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6800 and maybe_eh_return. Change regno to unsigned int. Use new args to
6801 handle EH_RETURN_DATA_REGNO registers properly.
6802 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6803 (riscv_expand_epilogue): Update comment. Change argument name and
6804 type. Update code to use new name and type. Pass new args to
6805 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6807 * config/riscv/riscv.md (NORMAL_RETURN): New.
6808 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6809 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6810 (eh_return): Call gen_eh_return_internal and emit barrier.
6811 (eh_return_internal): Call riscv_expand_epilogue.
6813 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6815 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6816 bit_insertion field and declare can_be_merged_into method.
6817 (merged_store_group::can_be_merged_into): New method.
6818 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6819 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6820 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6822 2018-06-04 Richard Biener <rguenther@suse.de>
6824 PR tree-optimization/85955
6825 * builtins.c (fold_builtin_sincos): Convert pointers to
6826 destination to appropriate type before dereferencing.
6828 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6830 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6832 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6834 * expr.c (expand_expr_real_1): Force the operand into memory if
6835 its TYPE_MODE is BLKmode and if there is no integer mode for
6836 the number of bits being extracted.
6838 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6842 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6843 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6845 2018-06-04 Richard Biener <rguenther@suse.de>
6847 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6848 (cleanup_tree_cfg_noloop): ... single caller. Do
6849 start_recording_case_labels later.
6851 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6853 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6854 to _IMMINTRIN_H_INCLUDED.
6855 * config/i386/pconfigintrin.h: Ditto.
6856 * config/i386/waitpkgintrin.h: Ditto.
6857 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6858 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6859 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6860 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6861 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6862 waitpkgintrin.h and cldemoteintrin.h.
6864 2018-06-04 Richard Biener <rguenther@suse.de>
6866 PR tree-optimization/86038
6867 * tracer.c (find_best_successor): Check probability for
6868 being initialized, bail out if not.
6870 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6873 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6874 of bits to ignore when comparing architectures.
6876 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6878 PR tree-optimization/69615
6879 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6880 maximum or minimum of the type, try to merge it also as if
6881 range1 is + [-, x - 1] or + [x + 1, -].
6884 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6886 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6888 PR tree-optimization/86034
6889 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6890 the unsigned bitfield type in a bit insertion sequence if it does not
6891 have a larger precision than the bitfield size.
6892 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6894 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6896 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6898 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6900 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6901 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6902 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6903 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6905 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6907 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6908 Disable -fdelete-null-pointer-checks for ELF toolchain.
6910 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6911 Kito Cheng <kito.cheng@gmail.com>
6913 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6914 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6915 (nds32*-*-*): Add glibc and uclibc conditions.
6916 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6917 (TARGET_EXCEPT_UNWIND_INFO): Define.
6918 * config/nds32/elf.h: New file.
6919 * config/nds32/linux.h: New file.
6920 * config/nds32/nds32-elf.opt: New file.
6921 * config/nds32/nds32-linux.opt: New file.
6922 * config/nds32/nds32-fp-as-gp.c
6923 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6924 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6926 (nds32_asm_file_end): Ditto.
6927 (nds32_print_operand): Ditto.
6928 (nds32_insert_attributes): Ditto.
6929 (nds32_init_libfuncs): New function.
6930 (TARGET_HAVE_TLS): Define.
6931 (TARGET_INIT_LIBFUNCS): Define.
6932 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6934 (TARGET_ELF): Apply different mcmodel setting.
6935 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6936 been migrated into elf.h and linux.h files.
6937 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6938 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6939 (mcmodel): The content has been migrated into nds32-elf.opt and
6940 nds32-linux.opt files.
6941 * config/nds32/t-elf: New file.
6942 * config/nds32/t-linux: New file.
6944 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6945 Shiva Chen <shiva0217@gmail.com>
6947 * config/nds32/constants.md (unspec_volatile_element): Add
6948 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6949 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6951 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6952 (make_pass_nds32_fp_as_gp): Declare.
6953 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6955 (nds32_asm_function_end_prologue): Remove unused asm output.
6956 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6957 (nds32_asm_file_start): Output necessary fp_as_gp information.
6958 (nds32_option_override): Adjust register usage.
6959 (nds32_expand_prologue): Consider fp_as_gp situation.
6960 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6961 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6964 (simple_return): Ditto.
6965 (omit_fp_begin): Output special directive for fp_as_gp.
6966 (omit_fp_end): Output special directive for fp_as_gp.
6967 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6968 mforbid-fp-as-gp): New options.
6970 2018-06-01 Mark Wielaard <mark@klomp.org>
6972 * dwarf2out.c (dwarf2out_finish): Remove generation of
6973 DW_AT_loclists_base.
6975 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6977 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6978 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6979 (struct merged_store_group): Add bit_insertion field.
6980 (dump_char_array): Use standard hexadecimal format.
6981 (merged_store_group::merged_store_group): Set bit_insertion to false.
6982 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6983 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6984 also print the mask in the dump file.
6985 (pass_store_merging::gate): Minor tweak.
6986 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6987 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6988 stores with INTEGER_CST stores.
6989 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6990 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6991 and use it throughout. Generate bit insertion sequences if need be.
6992 (pass_store_merging::process_store): Remove redundant condition.
6993 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6995 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6997 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6998 the 128-bit floating point types. Fix function comment.
7000 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7002 * config/aarch64/aarch64-simd.md
7003 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
7005 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
7008 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
7010 PR tree-optimization/85989
7011 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
7013 (backprop::intersect_uses): Check it when deciding whether this
7014 is a backedge reference.
7015 (backprop::process_block): Add each phi to m_visited_phis
7016 after visiting it, then clear it at the end.
7018 2018-06-01 Richard Biener <rguenther@suse.de>
7020 * tree-vectorizer.h (vect_dr_stmt): New function.
7021 (vect_get_load_cost): Adjust.
7022 (vect_get_store_cost): Likewise.
7023 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
7024 Use vect_dr_stmt instead of DR_SMTT.
7025 (vect_record_base_alignments): Likewise.
7026 (vect_calculate_target_alignment): Likewise.
7027 (vect_compute_data_ref_alignment): Likewise and make static.
7028 (vect_update_misalignment_for_peel): Likewise.
7029 (vect_verify_datarefs_alignment): Likewise.
7030 (vector_alignment_reachable_p): Likewise.
7031 (vect_get_data_access_cost): Likewise. Pass down
7032 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
7033 (vect_get_peeling_costs_all_drs): Likewise.
7034 (vect_peeling_hash_get_lowest_cost): Likewise.
7035 (vect_enhance_data_refs_alignment): Likewise.
7036 (vect_find_same_alignment_drs): Likewise.
7037 (vect_analyze_data_refs_alignment): Likewise.
7038 (vect_analyze_group_access_1): Likewise.
7039 (vect_analyze_group_access): Likewise.
7040 (vect_analyze_data_ref_access): Likewise.
7041 (vect_analyze_data_ref_accesses): Likewise.
7042 (vect_vfa_segment_size): Likewise.
7043 (vect_small_gap_p): Likewise.
7044 (vectorizable_with_step_bound_p): Likewise.
7045 (vect_prune_runtime_alias_test_list): Likewise.
7046 (vect_analyze_data_refs): Likewise.
7047 (vect_supportable_dr_alignment): Likewise.
7048 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7049 (vect_gen_prolog_loop_niters): Likewise.
7050 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7051 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
7053 (vect_recog_mask_conversion_pattern): Likewise.
7054 (vect_try_gather_scatter_pattern): Likewise.
7055 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
7056 to vect_get_store_cost.
7057 (vect_get_store_cost): Get stmt_info instead of DR.
7058 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
7059 (vect_get_load_cost): Get stmt_info instead of DR.
7061 2018-06-01 Richard Biener <rguenther@suse.de>
7064 * gimple-fold.c (var_decl_component_p): Also allow offsetted
7065 vars wrapped in MEM_REFs.
7067 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
7069 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7070 Fix subreg tests so that we only return a choice between
7071 GENERAL_REGS and FP_REGS if the original classes included both.
7073 2018-06-01 Richard Biener <rguenther@suse.de>
7076 * tree-ssa-structalias.c (get_function_part_constraint):
7077 Handle NULL fi->decl.
7078 (find_func_aliases_for_call): Properly handle indirect
7079 fi from direct call.
7080 (find_func_clobbers): Likewise.
7081 (ipa_pta_execute): Likewise.
7082 (create_variable_info_for): For functions that are ifunc_resolver
7083 resolve to a varinfo that contains the result of the resolver call.
7084 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
7087 2018-05-31 Michael Collison <michael.collison@arm.com>
7089 * config/aarch64/aarch64.md:
7090 (*fix_to_zero_extenddfdi2): New pattern.
7091 * gcc.target/aarch64/fix_extend1.c: New testcase.
7093 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
7097 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
7098 and BUILT_IN_STRNCMP_EQ.
7099 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
7100 BUILT_IN_STRNCMP_EQ.
7101 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
7102 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7103 (gimple_fold_builtin): Likewise.
7104 * tree-ssa-strlen.c (compute_string_length): New function.
7105 (determine_min_obsize): New function.
7106 (handle_builtin_string_cmp): New function to handle calls to
7107 string compare functions.
7108 (strlen_optimize_stmt): Add handling to builtin string compare
7110 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
7111 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7112 * tree.c (build_common_builtin_nodes): Add new defines of
7113 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
7115 2018-05-31 Jakub Jelinek <jakub@redhat.com>
7118 * bb-reorder.c (pass_partition_blocks::gate): Return false for
7119 functions with naked attribute.
7121 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
7123 * config/i386/sse.md (avx_vec_concat<mode>):
7124 Substitute concat_tg_mode mode attribute with xtg_mode.
7125 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
7126 (concat_tg_mode): Remove mode attribute.
7128 2018-05-31 Martin Sebor <msebor@redhat.com>
7131 * calls.c (alloc_max_size): Correct a logic error/typo.
7132 Treat excessive arguments as infinite. Warn for invalid arguments.
7133 * doc/invoke.texi (-Walloc-size-larger-than): Update.
7135 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
7138 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
7139 and movx for Haswell.
7141 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
7142 Cesar Philippidis <cesar@codesourcery.com>
7145 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
7146 when emitting error on private/firstprivate reductions.
7147 * omp-low.c (lower_omp_target): Avoid reference-type processing
7148 on pointers for firstprivate clause.
7150 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7152 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
7155 * config/aarch64/aarch64-simd.md
7156 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
7157 (aarch64_ld1_x3_<mode>): Likewise
7158 (aarch64_st1x2<VALLDIF:mode>): Likewise
7159 (aarch64_st1_x2_<mode>): Likewise
7160 (aarch64_st1x3<VALLDIF:mode>): Likewise
7161 (aarch64_st1_x3_<mode>): Likewise
7162 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
7163 (vld1_s8_x3): Likewise.
7164 (vld1_u16_x3): Likewise.
7165 (vld1_s16_x3): Likewise.
7166 (vld1_u32_x3): Likewise.
7167 (vld1_s32_x3): Likewise.
7168 (vld1_u64_x3): Likewise.
7169 (vld1_s64_x3): Likewise.
7170 (vld1_f16_x3): Likewise.
7171 (vld1_f32_x3): Likewise.
7172 (vld1_f64_x3): Likewise.
7173 (vld1_p8_x3): Likewise.
7174 (vld1_p16_x3): Likewise.
7175 (vld1_p64_x3): Likewise.
7176 (vld1q_u8_x3): Likewise.
7177 (vld1q_s8_x3): Likewise.
7178 (vld1q_u16_x3): Likewise.
7179 (vld1q_s16_x3): Likewise.
7180 (vld1q_u32_x3): Likewise.
7181 (vld1q_s32_x3): Likewise.
7182 (vld1q_u64_x3): Likewise.
7183 (vld1q_s64_x3): Likewise.
7184 (vld1q_f16_x3): Likewise.
7185 (vld1q_f32_x3): Likewise.
7186 (vld1q_f64_x3): Likewise.
7187 (vld1q_p8_x3): Likewise.
7188 (vld1q_p16_x3): Likewise.
7189 (vld1q_p64_x3): Likewise.
7190 (vst1_s64_x2): Likewise.
7191 (vst1_u64_x2): Likewise.
7192 (vst1_f64_x2): Likewise.
7193 (vst1_s8_x2): Likewise.
7194 (vst1_p8_x2): Likewise.
7195 (vst1_s16_x2): Likewise.
7196 (vst1_p16_x2): Likewise.
7197 (vst1_s32_x2): Likewise.
7198 (vst1_u8_x2): Likewise.
7199 (vst1_u16_x2): Likewise.
7200 (vst1_u32_x2): Likewise.
7201 (vst1_f16_x2): Likewise.
7202 (vst1_f32_x2): Likewise.
7203 (vst1_p64_x2): Likewise.
7204 (vst1q_s8_x2): Likewise.
7205 (vst1q_p8_x2): Likewise.
7206 (vst1q_s16_x2): Likewise.
7207 (vst1q_p16_x2): Likewise.
7208 (vst1q_s32_x2): Likewise.
7209 (vst1q_s64_x2): Likewise.
7210 (vst1q_u8_x2): Likewise.
7211 (vst1q_u16_x2): Likewise.
7212 (vst1q_u32_x2): Likewise.
7213 (vst1q_u64_x2): Likewise.
7214 (vst1q_f16_x2): Likewise.
7215 (vst1q_f32_x2): Likewise.
7216 (vst1q_f64_x2): Likewise.
7217 (vst1q_p64_x2): Likewise.
7218 (vst1_s64_x3): Likewise.
7219 (vst1_u64_x3): Likewise.
7220 (vst1_f64_x3): Likewise.
7221 (vst1_s8_x3): Likewise.
7222 (vst1_p8_x3): Likewise.
7223 (vst1_s16_x3): Likewise.
7224 (vst1_p16_x3): Likewise.
7225 (vst1_s32_x3): Likewise.
7226 (vst1_u8_x3): Likewise.
7227 (vst1_u16_x3): Likewise.
7228 (vst1_u32_x3): Likewise.
7229 (vst1_f16_x3): Likewise.
7230 (vst1_f32_x3): Likewise.
7231 (vst1_p64_x3): Likewise.
7232 (vst1q_s8_x3): Likewise.
7233 (vst1q_p8_x3): Likewise.
7234 (vst1q_s16_x3): Likewise.
7235 (vst1q_p16_x3): Likewise.
7236 (vst1q_s32_x3): Likewise.
7237 (vst1q_s64_x3): Likewise.
7238 (vst1q_u8_x3): Likewise.
7239 (vst1q_u16_x3): Likewise.
7240 (vst1q_u32_x3): Likewise.
7241 (vst1q_u64_x3): Likewise.
7242 (vst1q_f16_x3): Likewise.
7243 (vst1q_f32_x3): Likewise.
7244 (vst1q_f64_x3): Likewise.
7245 (vst1q_p64_x3): Likewise.
7247 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7249 * config/msp430/msp430.c (msp430_output_labelref): Prepend
7250 user_label_prefix to name.
7252 * tree-core.h: Update comment about the format of NAME string
7253 passed to handler in attribute_spec.
7255 * config/msp430/msp430.md: Remove erroneous subreg expression from
7256 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
7257 zero_extend{q,h}isi2.
7259 2018-05-30 Borislav Petkov <bp@suse.de>
7261 * doc/extend.texi: Document some architecture specific
7262 constraints and sort entries.
7264 2018-05-30 Martin Sebor <msebor@redhat.com>
7267 * builtins.c (expand_builtin_stpcpy_1): New function.
7268 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
7269 only if the former succeeds.
7271 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7273 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
7276 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7278 * doc/invoke.texi (-flinker-output): Document
7280 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7282 * passes.c (ipa_write_summaries): Only modify statements if body
7284 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
7285 incrementally linking.
7286 (ipa_passes): Likewise.
7287 * lto-cgraph.c (lto_output_node): When incrementally linking do not
7288 pass down resolution info.
7289 * common.opt (flag_incremental_link): Update info.
7290 * gcc.c (plugin specs): Turn flinker-output=* to
7291 -plugin-opt=-linker-output-known
7292 * toplev.c (compile_file): Also cut compilation when doing incremental
7294 * flag-types. (enum lto_partition_model): Add
7295 LTO_LINKER_OUTPUT_NOLTOREL.
7296 (invoke.texi): Add -flinker-output docs.
7297 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
7298 link same way as WPA; do not stream in dead initializers.
7300 * dwarf2out.c (dwarf2out_die_ref_for_decl,
7301 darf2out_register_external_decl): Support incremental link.
7303 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7305 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
7307 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7309 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
7310 it down to simple_object_copy_lto_debug_sections.
7311 (run_gcc): Determine incremental LTO link time and configure
7312 lto1 into non-wpa mode, disable renaming of debug sections.
7314 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
7316 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7317 descriptions of various incorrectly documented functions.
7319 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
7322 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7323 address check not strict.
7325 2018-05-30 Richard Biener <rguenther@suse.de>
7327 PR tree-optimization/85964
7328 * tracer.c (better_p): Drop initialized count check, we only
7329 call the function with initialized counts now.
7330 (find_best_successor): Do find a best edge if one
7331 has uninitialized count.
7332 (find_best_predecessor): Likewise. Do BB frequency check only
7333 if count is initialized.
7335 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
7337 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
7338 (aarch64_ldrstr_offset_compare): New.
7339 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
7340 load/store orderings.
7341 (aarch64_gen_adjusted_ldpstp): Likewise.
7343 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
7345 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7346 Check for subset of GENERAL_REGS and FP_REGS.
7347 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
7350 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
7352 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
7353 and wi::to_poly_offset. Add the current offset and then check
7354 whether the sum fits, rather than using an unchecked addition of
7355 a checked term. Check for a shwi rather than a uhwi.
7356 * expr.c (get_bit_range): Use tree_to_poly_uint64.
7357 (store_constructor): Use poly_int_tree_p.
7358 (expand_expr_real_1): Likewise.
7359 * function.c (assign_temp): Likewise.
7360 * fold-const.c (const_binop): Use poly_int_tree_p and
7362 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
7364 * ipa-icf-gimple.c (func_checker::compare_operand): Use
7365 to_poly_offset for MEM offsets.
7366 * ipa-icf.c (sem_variable::equals): Likewise.
7367 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
7368 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
7369 wi::to_poly_offset for BIT_FIELD_REF offsets.
7370 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
7372 * var-tracking.c (emit_note_insn_var_location): Use
7373 tree_to_poly_uint64.
7375 2018-05-29 Jim Wilson <jimw@sifive.com>
7377 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
7379 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
7382 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
7383 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
7385 (sse4_1_round<mode>2): Use nonimmediate_operand
7386 for operand 1 predicate.
7388 2018-05-29 Martin Sebor <msebor@redhat.com>
7389 Richard Biener <rguenther@suse.de>
7392 * calls.c (get_size_range): Call determine_value_range instead
7393 of get_value_range..
7394 * tree-vrp.h (determine_value_range): Declared new function.
7395 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
7397 2018-05-29 Richard Biener <rguenther@suse.de>
7399 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
7400 sure to use non-pattern stmts for get_earlier_stmt arguments.
7401 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
7402 called on pattern stmts.
7403 (get_later_stmt): Likewise.
7405 2018-05-29 Martin Liska <mliska@suse.cz>
7407 PR gcov-profile/85759
7408 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
7411 2018-05-29 Jakub Jelinek <jakub@redhat.com>
7413 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
7415 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
7419 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7420 VEC_PACK_FLOAT_EXPR): New tree codes.
7421 * tree-pretty-print.c (op_code_prio): Handle
7422 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7423 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7424 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7425 * tree-inline.c (estimate_operator_cost): Likewise.
7426 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
7427 * fold-const.c (const_binop): Likewise.
7428 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
7429 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7430 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7431 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
7432 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7433 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7434 * expr.c (expand_expr_real_2): Likewise.
7435 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
7436 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
7437 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
7439 * optabs.c (expand_widen_pattern_expr): For
7440 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
7441 sign from result type rather than operand's type.
7442 (expand_binop_directly): For vec_packu_float_optab and
7443 vec_packs_float_optab allow result type to be different from operand's
7445 * optabs-tree.c (optab_for_tree_code): Handle
7446 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7447 VEC_PACK_FLOAT_EXPR. Formatting fixes.
7448 * tree-vect-generic.c (expand_vector_operations_1): Handle
7449 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7450 VEC_PACK_FLOAT_EXPR.
7451 * tree-vect-stmts.c (supportable_widening_operation): Handle
7453 (supportable_narrowing_operation): Handle FLOAT_EXPR.
7454 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
7455 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
7456 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
7457 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
7459 (vec_pack<floatprefix>_float_<mode>): New expander.
7460 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
7462 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
7463 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
7464 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
7465 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
7466 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
7468 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
7469 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
7470 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7471 VEC_PACK_FLOAT_EXPR): Document.
7473 2018-05-29 Richard Biener <rguenther@suse.de>
7475 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
7477 (stmt_vec_info_vec): Make pointer.
7478 (init_stmt_vec_info_vec): Remove.
7479 (free_stmt_vec_info_vec): Likewise.
7480 (set_stmt_vec_info_vec): New function.
7481 (free_stmt_vec_infos): Likewise.
7482 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
7483 (set_vinfo_for_stmt): Likewise.
7484 (get_earlier_stmt): Likewise.
7485 (get_later_stmt): Likewise.
7486 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
7487 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
7488 (vec_info::~vec_info): Free stmt_vec_infos.
7489 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
7490 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
7491 (pass_slp_vectorize::execute): Likewise.
7492 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
7493 (free_stmt_vec_info_vec): Likewise.
7494 (set_stmt_vec_info_vec): New function.
7495 (free_stmt_vec_infos): Likewise.
7496 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
7497 the global stmt_vec_info_vec.
7498 * tree-parloops.c (gather_scalar_reductions): Use
7499 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
7502 2018-05-29 Richard Biener <rguenther@suse.de>
7504 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
7506 2018-05-29 Martin Liska <mliska@suse.cz>
7507 David Malcolm <dmalcolm@redhat.com>
7509 * vec.c (test_reverse): New.
7510 (vec_c_tests): Add new test.
7511 * vec.h (vl_ptr>::reverse): New function.
7513 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
7515 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
7517 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
7520 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
7522 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
7524 2018-05-28 Richard Biener <rguenther@suse.de>
7526 PR tree-optimization/85933
7527 * tree-vect-data-refs.c (vect_record_base_alignments): Only
7528 look at stmts marked as vectorizable.
7530 2018-05-28 Richard Biener <rguenther@suse.de>
7532 PR tree-optimization/85934
7533 * tree-vect-generic.c (expand_vector_operations_1): Hoist
7534 vector boolean check before scalar optimization.
7536 2018-05-28 Jakub Jelinek <jakub@redhat.com>
7538 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
7541 2018-05-28 Mark Wielaard <mark@klomp.org>
7543 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
7544 if it is an expression containing a minus sign.
7546 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
7548 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
7550 2018-05-27 Paul Koning <ni1d@arrl.net>
7552 * config/pdp11/pdp11.md (truncsihi2): Remove.
7554 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
7555 Chung-Ju Wu <jasonwucj@gmail.com>
7557 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
7559 (unaligned_store_dw): Ditto.
7560 * config/nds32/nds32-memory-manipulation.c
7561 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
7562 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
7563 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
7564 (emit_setmem_word_loop): Rename to ...
7565 (emit_setmem_doubleword_loop): ... this.
7566 (nds32_gen_dup_4_byte_to_word_value): New function.
7567 (nds32_gen_dup_8_byte_to_double_word_value): New function.
7568 (nds32_expand_setmem_loop): Refine implementation.
7569 (nds32_expand_setmem_loop_v3m): Ditto.
7570 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
7573 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7575 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
7577 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7579 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
7580 (nds32_init_machine_status): Initialize machine->attr_naked_p and
7581 machine->attr_no_prologue_p.
7582 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7583 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
7584 (nds32_expand_epilogue): Consider attr_naked_p.
7585 (nds32_expand_epilogue_v3pop): Likewise.
7586 (nds32_can_use_return_insn): Likewise.
7587 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
7588 attr_no_prologue_p fields.
7589 * config/nds32/nds32.opt (mret-in-naked-func): New option.
7591 2018-05-27 Jakub Jelinek <jakub@redhat.com>
7594 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
7596 * config/i386/sse.md
7597 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
7599 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
7601 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
7603 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
7605 (*<floatsuffix>floatv2div2sf2): Rename to ...
7606 (*float<floatunssuffix>v2div2sf2): ... this.
7607 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
7608 (float<floatunssuffix>v2div2sf2_mask): ... this.
7609 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
7610 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
7611 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
7613 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
7615 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7617 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7619 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7621 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7623 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
7624 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
7625 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
7626 gen_ufix_truncv8dfv8si2.
7627 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
7628 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
7629 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
7630 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
7631 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
7632 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
7633 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
7634 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
7636 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7640 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
7642 2018-05-25 Jim Wilson <jimw@sifive.com>
7644 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
7645 * config/riscv/riscv.c (struct machine_function): Add
7646 interrupt_handler_p and attribute_checked_p fields.
7647 (riscv_attribute_table): Add interrupt.
7648 (riscv_interrupt_type_p): New.
7649 (riscv_save_reg_p): Save extra regs for interrupt handler.
7650 (riscv_use_save_libcall): Return false for interrupt handler.
7651 (riscv_first_stack_step): Add forward declaration.
7652 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
7653 for interrupt handler with large frame. Use it for saved reg list.
7654 (riscv_expand_prologue): Move flag_stack_usage_info support to
7655 eliminate duplication.
7656 (riscv_expand_epilogue): Generate mret for interrupt handler.
7657 (riscv_epilogue_uses): New.
7658 (riscv_can_use_return_insn): Return false for interrupt handler.
7659 (riscv_function_ok_for_sibcall): Likewise.
7660 (riscv_set_current_function): Add interrupt handler support.
7661 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
7662 * config/riscv/riscv.md (UNSPECV_MRET): New.
7664 (riscv_frflags, riscv_fsflags): Use tab after opcode.
7666 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
7668 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
7670 PR tree-optimization/85712
7671 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
7672 this candidate has already been replaced in-situ by a copy.
7674 2018-05-25 Jason Merrill <jason@redhat.com>
7676 PR c++/80485 - inline function non-zero address.
7677 * symtab.c (nonzero_address): Check DECL_COMDAT.
7679 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
7682 * config/alpha/alpha.md (ashlsi3): New insn pattern.
7683 (*ashlsi_se): Rename from *ashldi_se. Define as sign
7684 extension of SImode operation. Use const123_operand predicate.
7685 (*saddsi_1): Remove.
7686 (*saddl_se_1): Ditto.
7688 (*ssubl_se_1): Ditto.
7689 * config/alpha/predicates.md (const123_operand): New predicate.
7690 * config/alpha/constraints.md (P): Use IN_RANGE.
7692 2018-05-25 Richard Biener <rguenther@suse.de>
7694 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
7696 (ref_maybe_used_by_stmt_p): Likewise.
7697 (stmt_may_clobber_ref_p): Likewise.
7698 (stmt_may_clobber_ref_p_1): Likewise.
7699 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
7701 (ref_maybe_used_by_stmt_p): Likewise.
7702 (stmt_may_clobber_ref_p): Likewise.
7703 (stmt_may_clobber_ref_p_1): Likewise.
7704 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
7705 the alias oracle to disambiguate DRs with stmts DR analysis
7707 (vect_analyze_data_refs): Do not give up on not analyzable
7708 DRs for BB vectorization. Remove code truncating the dataref
7711 2018-05-25 Jakub Jelinek <jakub@redhat.com>
7714 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7715 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
7716 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
7718 2018-05-25 Richard Biener <rguenther@suse.de>
7720 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
7721 function, combining stmt data ref gathering and fatal analysis
7723 (vect_analyze_data_refs): Remove now redudnant code and simplify.
7724 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
7725 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
7726 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
7727 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
7729 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7731 PR tree-optimization/85720
7732 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
7733 SCC if all partitions are builtins.
7734 (version_loop_by_alias_check): New parameter. Generate cancelable
7735 runtime alias check if all partitions are builtins.
7736 (distribute_loop): Update call to above function.
7738 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7740 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
7741 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
7742 (parm_default_def_partition_arg): Ditto.
7743 (set_parm_default_def_partition): Ditto.
7744 (get_parm_default_def_partitions): Ditto and make it static.
7745 (get_undefined_value_partitions): Ditto and make it static.
7746 (remove_ssa_form): Refactor call to init_var_map here.
7747 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
7748 computation for loop region.
7749 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
7750 (register_default_def): Delete.
7751 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
7752 (parm_default_def_partition_arg): Ditto.
7753 (set_parm_default_def_partition): Ditto.
7754 (get_parm_default_def_partitions): Ditto and make it static.
7755 (get_undefined_value_partitions): Ditto and make it static.
7756 (coalesce_with_default, coalesce_with_default): Update comment.
7757 (create_coalesce_list_for_region): New func factored out from
7758 create_outofssa_var_map.
7759 (populate_coalesce_list_for_outofssa): New func factored out from
7760 create_outofssa_var_map and coalesce_ssa_name.
7761 (create_outofssa_var_map): Delete.
7762 (coalesce_ssa_name): Refactor to support live range computation.
7763 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
7764 (get_parm_default_def_partitions): Delete.
7765 (get_undefined_value_partitions): Ditto.
7766 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7767 computation for loop region.
7768 (new_tree_live_info, loe_visit_block): Ditto.
7769 (live_worklist, set_var_live_on_entry): Ditto.
7770 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7771 * tree-ssa-live.h (struct _var_map): New fields.
7772 (init_var_map): Change decl.
7773 (region_contains_p): New.
7775 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7777 * tree-ssa-live.h (live_merge_and_clear): Delete.
7779 2018-05-25 Richard Biener <rguenther@suse.de>
7782 * tree-dump.c (dequeue_and_dump): Remove access to removed
7783 operand 2 of a SWITCH_EXPR.
7785 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7787 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7788 multiplication and division.
7789 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7790 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7791 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7792 (cond_udiv_optab, cond_umod_optab): New optabs.
7793 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7794 (IFN_COND_RDIV): New internal functions.
7795 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7796 TRUNC_MOD_EXPR and RDIV_EXPR.
7797 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7798 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7800 (SVE_INT_BINARY): Include mult.
7801 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7802 (optab, sve_int_op): Handle mult.
7803 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7805 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7806 for SVE_INT_BINARY_SD.
7808 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7810 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7811 (optab, sve_int_op): Handle div and udiv.
7812 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7813 for SVE_INT_BINARY_SD.
7814 (*<optab><mode>3): New insn for the same.
7816 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7818 * tree-vect-patterns.c: Include predict.h.
7819 (vect_recog_divmod_pattern): Restrict check for division support
7820 to when optimizing for size.
7822 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7824 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7825 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7826 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7827 (gimple_match_op::set_op): Likewise.
7828 (gimple_resimplify4): Declare.
7829 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7830 (expr::gen_transform): Likewise.
7831 (decision_tree::gen): Generate a simplification routine for 4 operands.
7832 * gimple-match-head.c (gimple_simplify): Add an overload for
7833 4 operands. In the top-level function, handle up to 4 call
7834 arguments and call gimple_resimplify4.
7835 (gimple_resimplify4): New function.
7836 (build_call_internal): Pass a fourth operand.
7837 (maybe_push_to_seq): Likewise.
7838 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7839 Fold VEC_COND_EXPRs of an operation and a default value into
7840 an IFN_COND_* function if possible.
7841 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7843 (SVE_COND_FP_BINARY): Include them.
7844 (optab, sve_fp_op): Handle them.
7845 (SVE_INT_BINARY_REV): New code iterator.
7846 (SVE_COND_FP_BINARY_REV): New int iterator.
7847 (commutative): New int attribute.
7848 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7850 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7852 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7853 (*cond_<optab><mode>): New patterns for reversed operands.
7855 2018-05-25 Richard Biener <rguenther@suse.de>
7857 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7858 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7859 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7860 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7861 * tree-vect-data-refs.c (everywhere): Adjust users.
7862 * tree-vect-loop.c (everywhere): Likewise.
7863 * tree-vect-slp.c (everywhere): Likewise.
7864 * tree-vect-stmts.c (everywhere): Likewise.
7865 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7867 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7869 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7871 (gcc_cv_as_section_exclude): ... this.
7872 Try Solaris as #exclude syntax.
7873 * configure: Regenerate.
7874 * config.in: Regenerate.
7875 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7877 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7878 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7880 * varasm.c (default_elf_asm_named_section): Don't check if
7881 HAVE_GAS_SECTION_EXCLUDE is defined.
7883 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7885 * doc/md.texi: Update the documentation of the cond_* optabs
7886 to mention the new final operand. Fix GET_MODE_NUNITS call.
7887 Describe the scalar case too.
7888 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7889 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7891 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7892 (get_conditional_internal_fn): Update comment.
7893 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7894 accumulator value as a final argument to conditional functions.
7895 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7896 a define_expand and add an "else" operand. Assert for now that
7897 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7898 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7899 (*cond_<optab><mode>): New patterns.
7900 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7901 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7902 (UNSPEC_COND_EOR): Delete.
7903 (optab): Remove associated mappings.
7904 (SVE_INT_BINARY): New code iterator.
7905 (sve_int_op): Remove int attribute and add "minus" to the code
7907 (SVE_COND_INT_OP): Delete.
7908 (SVE_COND_FP_OP): Rename to...
7909 (SVE_COND_FP_BINARY): ...this.
7911 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7913 * optabs.c (can_reuse_operands_p): New function.
7914 (maybe_legitimize_operands): Try to reuse the results for
7917 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7919 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7920 Add {q} suffix to insn mnemonic.
7922 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7924 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7925 (msp430_warn_func_return): New.
7927 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7929 * fold-const.c (tree_nonzero_bits): New function.
7930 * fold-const.h (tree_nonzero_bits): Likewise.
7931 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7932 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7934 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7938 * varasm.c (assemble_alias): Check ifunc_resolver only on
7941 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7944 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7945 when memory input operand is handled.
7947 2018-05-24 Luis Machado <luis.machado@linaro.org>
7949 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7951 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7953 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7955 * match.pd: Delay FMA folds until after vectorization.
7957 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7960 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7961 address check not strict.
7963 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7965 * gimple-match.h (gimple_match_op): New class.
7966 (mprts_hook): Replace parameters with a gimple_match_op *.
7967 (maybe_build_generic_op): Likewise.
7968 (gimple_simplified_result_is_gimple_val): Replace parameters with
7969 a const gimple_match_op *.
7970 (gimple_simplify): Replace code_helper * and tree * parameters with
7971 a gimple_match_op * parameter.
7972 (gimple_resimplify1): Replace code_helper *, tree and tree *
7973 parameters with a gimple_match_op * parameter.
7974 (gimple_resimplify2): Likewise.
7975 (gimple_resimplify3): Likewise.
7976 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7977 parameters with a gimple_match_op * parameter.
7978 * gimple-match-head.c (gimple_simplify): Change prototypes of
7979 auto-generated functions to take a gimple_match_op * instead of
7980 separate code_helper * and tree * parameters. Make the same
7981 change in the top-level overload and update calls to the
7982 gimple_resimplify routines. Update calls to the auto-generated
7983 functions and to maybe_push_res_to_seq in the publicly-facing
7984 operation-specific gimple_simplify overloads.
7985 (gimple_match_op::MAX_NUM_OPS): Define.
7986 (gimple_resimplify1): Replace rcode and ops with a single res_op
7987 parameter. Update call to gimple_simplify.
7988 (gimple_resimplify2): Likewise.
7989 (gimple_resimplify3): Likewise.
7990 (mprts_hook): Replace parameters with a gimple_match_op *.
7991 (maybe_build_generic_op): Likewise.
7992 (build_call_internal): Replace type, nargs and ops with
7993 a gimple_match_op *.
7994 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7995 with a single gimple_match_op *. Update calls to mprts_hook,
7996 build_call_internal and gimple_simplified_result_is_gimple_val.
7997 Factor out code that is common to the tree_code and combined_fn cases.
7998 * genmatch.c (expr::gen_transform): Replace tem_code and
7999 tem_ops with a gimple_match_op called tem_op. Update calls
8000 to the gimple_resimplify functions and maybe_push_res_to_seq.
8001 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
8002 res_ops. Update call to the gimple_resimplify functions.
8003 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
8004 (decision_tree::gen): Make the functions take a gimple_match_op *
8005 called res_op instead of separate res_code and res_ops parameters.
8006 Update call accordingly.
8007 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
8008 and ops with a single res_op parameter. Update calls to
8009 maybe_build_generic_op and maybe_push_res_to_seq.
8010 (fold_stmt_1): Update calls to gimple_simplify and
8011 replace_stmt_with_simplification.
8012 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
8013 and gimple_simplified_result_is_gimple_val.
8014 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
8016 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
8017 with a gimple_match_op *.
8018 (vn_nary_build_or_lookup): Likewise. Update call to
8019 vn_nary_build_or_lookup_1.
8020 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
8021 gimple_match_op *. Update calls to the gimple_resimplify routines
8022 and to gimple_simplified_result_is_gimple_val.
8023 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
8024 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
8025 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
8026 (visit_nary_op): Likewise.
8027 (visit_reference_op_load): Likewise.
8029 2018-05-23 Luis Machado <luis.machado@linaro.org>
8031 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
8032 modifier for printing the step amount.
8034 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
8037 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
8040 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
8042 * doc/sourcebuild.texi (Endianness): New subsubsection.
8044 2018-05-23 Luis Machado <luis.machado@linaro.org>
8046 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8047 <prefetch_dynamic_strides>: New const bool field.
8048 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8049 prefetch_dynamic_strides.
8050 (exynosm1_prefetch_tune): Likewise.
8051 (thunderxt88_prefetch_tune): Likewise.
8052 (thunderx_prefetch_tune): Likewise.
8053 (thunderx2t99_prefetch_tune): Likewise.
8054 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
8056 (aarch64_override_options_internal): Update to set
8057 PARAM_PREFETCH_DYNAMIC_STRIDES.
8058 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8059 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8060 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8061 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8062 prefetch-dynamic-strides setting.
8064 2018-05-23 Luis Machado <luis.machado@linaro.org>
8066 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8067 <minimum_stride>: New const int field.
8068 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8069 minimum_stride field defaulting to -1.
8070 (exynosm1_prefetch_tune): Likewise.
8071 (thunderxt88_prefetch_tune): Likewise.
8072 (thunderx_prefetch_tune): Likewise.
8073 (thunderx2t99_prefetch_tune): Likewise.
8074 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
8075 <default_opt_level>: Set to 3.
8076 (aarch64_override_options_internal): Update to set
8077 PARAM_PREFETCH_MINIMUM_STRIDE.
8078 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8079 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8080 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8081 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8082 stride is constant and is below the minimum stride threshold.
8084 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8086 * config/arm/arm-cpus.in (mode26): Delete.
8087 (armv4): Delete mode26 reference.
8088 * config/arm/arm.c (arm_configure_build_target): Delete use of
8091 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8093 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
8095 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
8096 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
8098 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
8099 (floatunsdidf2): Ditto.
8101 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8103 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
8104 (fixuns_trunc<mode>si2_avx512f): Ditto.
8105 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
8106 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
8107 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
8109 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8111 PR rtl-optimization/79985
8112 * df-scan.c (df_insn_refs_collect): Remove special case for
8113 global registers and asm statements.
8115 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8117 * extend.texi (Global Register Variables): Rewrite the bullet list.
8118 Note that the register is available for allocation. Note that access
8119 via inline asm must use constraints. Add note about async-signal
8120 handlers. Remove paragraph about automagic register selection.
8122 2018-05-23 Richard Biener <rguenther@suse.de>
8124 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
8125 of fixed offset from memset VN.
8127 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
8129 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
8131 (alloc_cand_and_find_basis): Initialize first_interp field.
8132 (slsr_process_mul): Modify first_interp field.
8133 (slsr_process_add): Likewise.
8134 (slsr_process_cast): Modify first_interp field for each new
8136 (slsr_process_copy): Likewise.
8137 (dump_candidate): Dump first_interp field.
8138 (replace_mult_candidate): Process all interpretations, not just
8140 (replace_rhs_if_not_dup): Likewise.
8141 (replace_one_candidate): Likewise.
8143 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
8145 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
8147 (aarch64_needs_frame_chain): New function.
8148 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
8150 2018-05-23 Sudakshina Das <sudi.das@arm.com>
8153 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
8154 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
8155 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
8156 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
8157 as true for strict-align.
8158 (aarch64_can_inline_p): Perform checks even when callee has no
8159 attributes to check for strict alignment.
8160 * doc/extend.texi (AArch64 Function Attributes): Document
8162 * doc/invoke.texi: (AArch64 Options): Likewise.
8164 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
8166 PR tree-optimization/85853
8167 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
8168 the handling of the root of the node to...
8169 (vect_slp_analyze_node_operations_1): ...this new function,
8170 and run the whole thing with the child nodes' def types
8171 set according to their SLP node's def type.
8173 2018-05-23 Richard Biener <rguenther@suse.de>
8176 * tree-data-ref.c (create_runtime_alias_checks): Defer
8177 and ignore overflow warnings.
8179 2018-05-23 Yury Gribov <tetra2005@gmail.com>
8181 PR tree-optimization/85822
8182 * tree-vrp.c (is_masked_range_test): Fix handling of negative
8185 2018-05-23 Richard Biener <rguenther@suse.de>
8187 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8188 memset constants via native_interpret_expr.
8190 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
8193 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
8195 (cgraph_node::create_alias): Likewise.
8196 (cgraph_node::get_availability): Check ifunc_resolver instead
8197 of looking up ifunc attribute.
8198 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8199 * varasm.c (do_assemble_alias): Likewise.
8200 (assemble_alias): Likewise.
8201 (default_binds_local_p_3): Likewise.
8202 * cgraph.h (cgraph_node): Add ifunc_resolver.
8203 (cgraph_node::only_called_directly_or_aliased_p): Return false
8205 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
8207 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
8208 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
8209 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
8210 instead of looking up ifunc attribute.
8212 2018-05-22 Luis Machado <luis.machado@linaro.org>
8214 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
8216 2018-05-22 Martin Sebor <msebor@redhat.com>
8219 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
8220 only when expasion succeeds.
8221 (expand_builtin_strcmp): Same.
8222 (expand_builtin_strncmp): Same.
8224 2018-05-22 Martin Sebor <msebor@redhat.com>
8226 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
8228 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8229 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8231 * config/aarch64/aarch64-ldpstp.md: Replace uses of
8232 aarch64_mem_pair_operand with memory_operand and delete operand swapping
8234 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8235 Add check for legitimate_address.
8236 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
8237 (aarch64_swap_ldrstr_operands): New.
8238 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
8241 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8242 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8244 * config/aarch64/aarch64.md: New patterns to generate stp
8246 (store_pair_sw, store_pair_dw): New patterns to generate stp for
8247 single words and double words.
8248 (load_pair_sw, load_pair_dw): Likewise.
8249 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
8251 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
8253 * config/aarch64/aarch64-ldpstp.md: Modify peephole
8254 for different mode ldpstp and add peephole for merged zero stores.
8256 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8258 (aarch64_gen_store_pair): Rename calls to match new patterns.
8259 (aarch64_gen_load_pair): Rename calls to match new patterns.
8260 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
8261 (load_pair<DREG:mode><DREG2:mode>): ... This.
8262 (store_pair<mode>): Rename to...
8263 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
8264 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
8266 (V_INT_EQUIV): Handle SImode.
8267 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
8270 2018-05-22 Martin Sebor <msebor@redhat.com>
8273 * calls.c (maybe_warn_nonstring_arg): Use string length to set
8274 or ajust the presumed bound on an operation to avoid unnecessary
8277 2018-05-22 Martin Sebor <msebor@redhat.com>
8279 PR tree-optimization/85826
8280 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
8281 assuming that a DECL necesarily has a constant size.
8283 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
8286 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
8288 2018-05-22 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/85834
8291 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
8292 non-constant and non-zero memset arguments.
8294 2018-05-22 Martin Liska <mliska@suse.cz>
8297 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
8299 2018-05-22 Richard Biener <rguenther@suse.de>
8301 PR tree-optimization/85863
8302 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
8303 comparisons when vectype is specified.
8304 (vectorizable_condition): Do not specify vectype for
8305 vect_is_simple_cond when SLP vectorizing.
8307 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
8310 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
8311 define __ibm128 as long double.
8312 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
8313 as a distinct type when IEEE 128-bit support is enabled.
8314 (init_float128_ieee): Fix up conversions between IFmode and IEEE
8315 128-bit types to use the correct functions.
8316 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
8317 convert between 128-bit floating point types that have different
8318 modes but the same representation, instead of using gen_lowpart to
8320 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
8322 (IFKF_reg): New attributes to give the register constraints for
8324 (extend<mode>tf2_internal): New insns to mark an explicit
8325 conversion between 128-bit floating point types that have a
8326 different mode but share the same representation.
8328 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
8330 PR tree-optimization/85814
8331 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
8332 a null return from get_strinfo when unsharing the next
8333 strinfo in the chain.
8335 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8338 * varasm.c (weak_finish): Clean up weak_decls.
8340 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8342 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
8343 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
8344 UNSPEC_UADALP values.
8345 * config/aarch64/iterators.md (ABAL): New int iterator.
8348 (sur): Add mappings for the above.
8349 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
8351 (aarch64_<sur>abal<mode>_4): Likewise.
8352 (aarch64_<sur>adalp<mode>_3): Likewise.
8353 (<sur>sadv16qi): New define_expand.
8355 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
8357 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
8358 (*movdf_internal): Ditto.
8359 (*rcpsf2_sse): Ditto.
8360 (*rsqrtsf2_sse): Ditto.
8361 (*sqrt<mode>2_sse): Ditto.
8363 2018-05-21 Tamar Christina <tamar.christina@arm.com>
8365 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
8367 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
8368 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
8369 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
8370 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
8372 * config/aarch64/arm_neon.h: Likewise.
8373 * config/aarch64/iterators.md (VQ_I): New.
8375 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
8377 * config.gcc: Add arc/t-multilib-linux to tmake_file for
8379 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
8382 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8384 * config/nds32/constraints.md (S): New constraint.
8385 * config/nds32/nds32.md (call_internal): Use constraint S.
8386 (call_value_internal): Likewise.
8387 (sibcall_internal): Likewise.
8388 (sibcall_value_internal): Likewise.
8390 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8391 Chung-Ju Wu <jasonwucj@gmail.com>
8393 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
8396 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8397 Chung-Ju Wu <jasonwucj@gmail.com>
8399 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
8400 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
8401 (nds32_rtx_costs_impl): Simplify.
8402 (nds32_address_cost_impl): Simplify.
8403 (nds32_init_rtx_costs): New function.
8404 (nds32_rtx_costs_speed_prefer): Likewise.
8405 (nds32_rtx_costs_size_prefer): Likewise.
8406 (nds32_address_cost_speed_prefer): Likewise.
8407 (nds32_address_cost_speed_fwprop): Likewise.
8408 (nds32_address_cost_size_prefer): Likewise.
8409 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
8410 * config/nds32/nds32.c (nds32_option_override): Use
8411 nds32_init_rtx_costs function.
8413 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8415 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
8416 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
8417 (TARGET_PIPELINE_N8): Likewise.
8418 (TARGET_PIPELINE_N10): Likewise.
8419 (TARGET_PIPELINE_N13): Likewise.
8420 (TARGET_PIPELINE_GRAYWOLF): Likewise.
8422 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
8424 * config/nds32/nds32-fpu.md: Update copyright year.
8426 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8428 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
8430 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8432 * config/nds32/nds32.c
8433 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
8434 * config/nds32/nds32.opt (minline-asm-r15): New option.
8436 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8438 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
8440 * config/nds32/nds32.md (abssi2): New pattern.
8442 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
8444 * config/i386/i386.md (rex64namesuffix): New mode attribute.
8445 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
8446 Merge insn pattern from sse_cvtsi2ss<round_name> and
8447 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
8448 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
8449 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
8450 using SWI48 mode iterator.
8451 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
8452 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
8453 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
8454 pattern from sse_cvttss2si<round_saeonly_name>
8455 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
8456 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
8457 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
8458 using SWI48 mode iterator.
8459 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
8460 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
8461 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
8462 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
8463 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
8464 using SWI48 mode iterator.
8465 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
8466 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
8467 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
8468 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
8469 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
8470 SWI48 mode iterator.
8471 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
8472 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
8473 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
8474 pattern from sse_cvttsd2si<round_saeonly_name>
8475 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
8477 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8479 * config/nds32/nds32-md-auxiliary.c
8480 (nds32_valid_smw_lwm_base_p): Refine.
8481 (nds32_output_smw_single_word): Refine.
8482 (nds32_output_smw_double_word): New.
8483 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
8485 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8487 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
8488 (nds32_output_stack_pop): Refine.
8489 (nds32_expand_unaligned_load): Refine.
8490 (nds32_expand_unaligned_store): Refine.
8492 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8493 Chung-Ju Wu <jasonwucj@gmail.com>
8495 * config/nds32/constants.md: Add TP_REGNUM constant.
8496 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
8497 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
8499 * config/nds32/nds32-doubleword.md: Consider flag_pic.
8500 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
8501 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
8502 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
8503 and PIC code generation.
8504 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
8506 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
8508 * config/nds32/nds32.md: Support TLS and PIC.
8509 * config/nds32/nds32.c: Support TLS and PIC.
8510 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
8511 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
8514 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8516 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
8517 mode with E_ prefix.
8519 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8520 Chung-Ju Wu <jasonwucj@gmail.com>
8522 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
8523 * config/nds32/nds32-md-auxiliary.c
8524 (symbolic_reference_mentioned_p): New.
8525 (nds32_legitimize_ict_address): New.
8526 (nds32_expand_ict_move): New.
8527 (nds32_indirect_call_referenced_p): New.
8528 (nds32_symbol_binds_local_p): Delete.
8529 (nds32_long_call_p): Modify.
8530 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
8531 * config/nds32/nds32-protos.h
8532 (symbolic_reference_mentioned_p): Declare.
8533 (nds32_legitimize_ict_address): Declare.
8534 (nds32_expand_ict_move): Declare.
8535 (nds32_indirect_call_referenced_p): Declare.
8536 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
8537 (nds32_relax_group): Use nds32_ict_const_p as condition.
8538 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
8539 (nds32_asm_file_start): Output ict_model directive in asm code.
8540 (nds32_legitimate_address_p): Consider indirect call.
8541 (nds32_print_operand): Consider indirect call.
8542 (nds32_print_operand_address): Consider indirect call.
8543 (nds32_insert_attributes): Handle "indirect_call" attribute.
8544 (TARGET_LEGITIMATE_ADDRESS_P): Define.
8545 (TARGET_LEGITIMATE_CONSTANT_P): Define.
8546 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
8547 (TARGET_DELEGITIMIZE_ADDRESS): Define.
8548 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
8549 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
8550 (TARGET_ICT_MODEL_SMALL): Define.
8551 (TARGET_ICT_MODEL_LARGE): Define.
8552 * config/nds32/nds32.md (movsi): Consider ict model.
8553 (call, call_value): Consider ict model.
8554 (sibcall, sibcall_value): Consider ict model.
8555 * config/nds32/nds32.opt (mict-model): New option.
8556 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
8559 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
8560 Monk Chiang <sh.chiang04@gmail.com>
8561 Jim Wilson <jimw@sifive.com>
8563 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
8564 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
8565 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
8566 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
8567 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
8568 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
8569 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
8570 compute save_libcall_adjustment properly.
8571 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
8572 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
8573 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
8574 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
8575 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
8576 (ABI_SPEC): Handle mabi=ilp32e.
8577 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
8578 (RVE): Add RVE mask.
8579 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
8580 <-march>: Add rv32e as an example.
8582 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
8585 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
8586 (intra_create_variable_infos): Handle C++ constructors.
8588 2018-05-18 Martin Liska <mliska@suse.cz>
8590 * passes.def: Remove a redundant pass.
8592 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8595 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
8597 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8599 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
8601 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
8603 (armv2, armv2a, armv3, armv3m): Delete architectures.
8604 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
8605 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
8606 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
8608 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
8609 (*mulsidi3adddi): Likewise.
8610 (mulsidi3): Likewise.
8611 (*mulsidi3_nov6): Likewise.
8612 (umulsidi3): Likewise.
8613 (umulsidi3_nov6): Likewise.
8614 (umaddsidi4): Likewise.
8615 (*umulsidi3adddi): Likewise.
8616 (smulsi3_highpart): Likewise.
8617 (*smulsi3_highpart_nov6): Likewise.
8618 (umulsi3_highpart): Likewise.
8619 (*umulsi3_highpart_nov6): Likewise.
8620 * config/arm/arm.h (arm_arch3m): Delete.
8621 * config/arm/arm.c (arm_arch3m): Delete.
8622 (arm_option_override_internal): Update armv3-related comment.
8623 (arm_configure_build_target): Delete use of isa_bit_mode32.
8624 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
8625 (arm_rtx_costs_internal): Delete check of arm_arch3m.
8626 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
8628 (mulusa3): Likewise.
8629 * config/arm/arm-protos.h (arm_arch3m): Delete.
8630 * config/arm/arm-tables.opt: Regenerate.
8631 * config/arm/arm-tune.md: Likewise.
8632 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
8633 deleted architectures.
8635 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8637 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
8638 (armv5t, armv5te): New features.
8639 (ARMv5, ARMv5e): Delete fgroups.
8640 (ARMv5t, ARMv5te): Adjust for above changes.
8642 (armv5, armv5e): Delete arches.
8643 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
8645 (*call_reg_arm): Likewise.
8646 (*call_value_reg_armv5): Likewise.
8647 (*call_value_reg_arm): Likewise.
8648 (*call_symbol): Likewise.
8649 (*call_value_symbol): Likewise.
8650 (*sibcall_insn): Likewise.
8651 (*sibcall_value_insn): Likewise.
8653 (prefetch): Likewise.
8654 (define_split and define_peephole2 dependent on arm_arch5):
8656 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
8658 (TARGET_ARM_QBIT): Likewise.
8659 (TARGET_DSP_MULTIPLY): Likewise.
8660 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
8661 (arm_arch5, arm_arch5e): Delete.
8662 (arm_arch5t, arm_arch5te): Declare.
8663 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
8664 (arm_arch5t): Declare.
8665 (arm_option_reconfigure_globals): Update for the above.
8666 (arm_options_perform_arch_sanity_checks): Update comment, replace
8667 use of arm_arch5 with arm_arch5t.
8668 (use_return_insn): Likewise.
8669 (arm_emit_call_insn): Likewise.
8670 (output_return_instruction): Likewise.
8671 (arm_final_prescan_insn): Likewise.
8672 (arm_coproc_builtin_available): Likewise.
8673 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
8674 arm_arch5e with arm_arch5t and arm_arch5te.
8675 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
8676 (arm_arch5t, arm_arch5te): Declare.
8677 * config/arm/arm-tables.opt: Regenerate.
8678 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
8679 * config/arm/t-multilib: Likewise.
8680 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
8681 instead of arm_arch5.
8682 (*call_reg_thumb1): Likewise.
8683 (*call_value_reg_thumb1_v5): Likewise.
8684 (*call_value_reg_thumb1): Likewise.
8685 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
8687 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
8689 2018-05-18 Martin Liska <mliska@suse.cz>
8691 PR gcov-profile/84846
8692 * doc/gcov.texi: Document -t option of gcov tool.
8694 2018-05-18 Martin Liska <mliska@suse.cz>
8696 PR gcov-profile/84846
8697 * gcov.c (print_usage): Add new -t option.
8698 (process_args): Handle the option.
8699 (generate_results): Use stdout as output when requested by
8702 2018-05-18 Martin Liska <mliska@suse.cz>
8704 PR gcov-profile/84846
8705 * coverage.c (coverage_init): Write PWD to .gcno file.
8706 * doc/gcov.texi: Document how working directory is printed.
8707 * gcov-dump.c (dump_gcov_file): Print PWD.
8708 * gcov.c (output_intermediate_file): Likewise.
8709 (read_graph_file): Read PWD string.
8710 (output_lines): Print PWD.
8712 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8715 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
8716 for retval and return false if all args to phi are zero.
8718 2018-05-18 Richard Biener <rguenther@suse.de>
8720 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
8722 (evrp_dom_walker::before_dom_children): Call it.
8724 2018-05-18 Richard Biener <rguenther@suse.de>
8726 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
8727 results when processing array refs with variable index.
8729 2018-05-18 Toon Moene <toon@moene.org>
8731 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
8732 directly after that of -floop-interchange. Indicate that both
8733 options are enabled by default when specifying -O3.
8735 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8737 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
8738 iterator. Delete separate integer-mode vec_set<mode> expander.
8739 (aarch64_simd_vec_setv2di): Delete.
8740 (vec_setv2di): Delete.
8741 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
8742 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
8743 the "w, r" alternative.
8745 2018-05-18 Martin Liska <mliska@suse.cz>
8747 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
8748 * tree-pass.h (make_pass_lower_switch_O0): New function.
8749 * tree-switch-conversion.c (node_has_low_bound): Remove.
8750 (node_has_high_bound): Likewise.
8751 (node_is_bounded): Likewise.
8752 (class pass_lower_switch): Make it a template type and create
8754 (pass_lower_switch::execute): Add template argument.
8755 (make_pass_lower_switch): New function.
8756 (make_pass_lower_switch_O0): New function.
8757 (do_jump_if_equal): Remove.
8758 (emit_case_nodes): Simplify to just handle all 3 cases and leave
8759 all the hard work to tree optimization passes.
8761 2018-05-18 Martin Liska <mliska@suse.cz>
8763 * dbgcnt.c (limit_low): Renamed from limit.
8764 (limit_high): New variable.
8765 (dbg_cnt_is_enabled): Check for upper limit.
8766 (dbg_cnt): Adjust dumping.
8767 (dbg_cnt_set_limit_by_index): Add new argument for high
8769 (dbg_cnt_set_limit_by_name): Likewise.
8770 (dbg_cnt_process_single_pair): Parse new format.
8771 (dbg_cnt_process_opt): Use strtok.
8772 (dbg_cnt_list_all_counters): Remove 'value' and add
8774 * doc/invoke.texi: Document changes.
8776 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8778 * doc/sourcebuild.texi (scalar_all_fma): Document.
8779 * tree.def (FMA_EXPR): Delete.
8780 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8781 * internal-fn.c (ternary_direct): New macro.
8782 (expand_ternary_optab_fn): Likewise.
8783 (direct_ternary_optab_supported_p): Likewise.
8784 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8785 * builtins.c (fold_builtin_fma): Delete.
8786 (fold_builtin_3): Don't call it.
8787 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8788 * expr.c (expand_expr_real_2): Likewise.
8789 * fold-const.c (operand_equal_p): Likewise.
8790 (fold_ternary_loc): Likewise.
8791 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8792 * gimple.c (DEFTREECODE): Likewise.
8793 * gimplify.c (gimplify_expr): Likewise.
8794 * optabs-tree.c (optab_for_tree_code): Likewise.
8795 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8796 * tree-eh.c (operation_could_trap_p): Likewise.
8797 (stmt_could_throw_1_p): Likewise.
8798 * tree-inline.c (estimate_operator_cost): Likewise.
8799 * tree-pretty-print.c (dump_generic_node): Likewise.
8800 (op_code_prio): Likewise.
8801 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8802 * tree-ssa-operands.c (get_expr_operands): Likewise.
8803 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8804 * fold-const-call.h (fold_fma): Delete.
8805 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8806 CFN_FNMA and CFN_FNMS.
8808 * genmatch.c (combined_fn): New enum.
8809 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8810 (commutative_op): New function.
8811 (commutate): Use it. Handle more than 2 operands.
8812 (dt_operand::gen_gimple_expr): Use commutative_op.
8813 (parser::parse_expr): Allow :c to be used with non-binary
8814 operators if the commutative operand is known.
8815 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8816 CFN_FMS, CFN_FNMA and CFN_FNMS.
8817 (backprop::process_assign_use): Remove FMA_EXPR handling.
8818 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8819 (gen_hsa_fma): New function.
8820 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8821 IFN_FNMA and IFN_FNMS.
8822 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8823 * gimple-fold.h (follow_all_ssa_edges): Declare.
8824 * gimple-fold.c (follow_all_ssa_edges): New function.
8825 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8826 gimple_build interface and use follow_all_ssa_edges to fold the result.
8827 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8828 instead of checking for optabs directly.
8829 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8830 rather than FMA_EXPRs.
8831 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8832 call to IFN_FMA instead of an FMA_EXPR.
8834 2018-05-17 Jim Wilson <jimw@sifive.com>
8836 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8837 sign extended promoted subreg, and the range does not have the sign bit
8838 set, then do a sign extend.
8840 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8841 test, check for sign extended subreg and/or constant operands, and
8842 do a sign extend in that case.
8844 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8846 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8847 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8849 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8850 Change logics_shift_reg to logics_shift_imm.
8851 (thunderx2t99_fp_loadpair_basic): Delete.
8852 (thunderx2t99_fp_storepair_basic): Delete.
8853 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8854 (thunderx2t99_asimd_polynomial): Delete.
8855 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8856 and neon_fp_mul_d_scalar_q.
8857 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8858 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8859 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8860 (thunderx2t99_asimd_lut): Add missing tbl types.
8861 (thunderx2t99_asimd_ext): Delete.
8862 (thunderx2t99_asimd_load1_1_mult): Delete.
8863 (thunderx2t99_asimd_load1_2_mult): Delete.
8864 (thunderx2t99_asimd_load1_ldp): New.
8865 (thunderx2t99_asimd_load1): New.
8866 (thunderx2t99_asimd_load2): Add missing *load2* types.
8867 (thunderx2t99_asimd_load3): New.
8868 (thunderx2t99_asimd_load4): New.
8869 (thunderx2t99_asimd_store1_1_mult): Delete.
8870 (thunderx2t99_asimd_store1_2_mult): Delete.
8871 (thunderx2t99_asimd_store2_mult): Delete.
8872 (thunderx2t99_asimd_store2_onelane): Delete.
8873 (thunderx2t99_asimd_store_stp): New.
8874 (thunderx2t99_asimd_store1): New.
8875 (thunderx2t99_asimd_store2): New.
8876 (thunderx2t99_asimd_store3): New.
8877 (thunderx2t99_asimd_store4): New.
8879 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8881 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8882 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8884 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8885 Segher Boessenkool <segher@kernel.crashing.org>
8888 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8891 2018-05-17 Richard Biener <rguenther@suse.de>
8893 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8894 for pruning loop and prune defs feeding only already visited PHIs.
8896 2018-05-17 Richard Biener <rguenther@suse.de>
8898 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8900 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8901 Richard Biener <rguenther@suse.de>
8903 PR tree-optimization/85793
8904 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8905 for VMAT_ELEMENTWISE.
8907 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8909 * internal-fn.h (lookup_internal_fn): Declare
8910 * internal-fn.c (lookup_internal_fn): New function.
8911 * gimple.c (gimple_build_call_from_tree): Handle calls to
8913 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8914 internal function names.
8915 * tree-pretty-print.c (dump_generic_node): Likewise.
8916 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8918 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8920 * gimple-fold.h (gimple_build): Make the function forms take
8921 combined_fn rather than built_in_function.
8922 (gimple_simplify): Likewise.
8923 * gimple-match-head.c (gimple_simplify): Likewise.
8924 * gimple-fold.c (gimple_build): Likewise.
8925 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8926 rather than gimple_build_call_internal.
8927 (get_initial_defs_for_reduction): Likewise.
8928 (vect_create_epilog_for_reduction): Likewise.
8929 (vectorizable_live_operation): Likewise.
8931 2018-05-17 Martin Liska <mliska@suse.cz>
8933 * gimple-ssa-sprintf.c (format_directive): Do not use
8934 space in between 'G_' and '('.
8936 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8939 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8940 even if the mask is not all ones.
8943 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8945 (ix86_gimple_fold_builtin): Likewise.
8948 * config/i386/i386.c: Include tree-vector-builder.h.
8949 (ix86_vector_shift_count): New function.
8950 (ix86_fold_builtin): Fold shift builtins by scalar count.
8951 (ix86_gimple_fold_builtin): Likewise.
8953 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8954 _mm512_setzero): New intrinsics.
8956 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8957 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8959 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8960 code generation for cases where splatting a value is not useful.
8961 * simplify-rtx.c (simplify_ternary_operation): Simplify
8962 vec_merge across a vec_duplicate and a paradoxical subreg forming
8963 a vector mode to a vec_concat.
8965 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8967 * config.gcc: Support "goldmont-plus".
8968 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8970 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8971 PROCESSOR_GOLDMONT_PLUS.
8972 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8973 (processor_target_table): Add "goldmont-plus".
8974 (PTA_GOLDMONT_PLUS): Define.
8975 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8976 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8977 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8978 (fold_builtin_cpu): Add "goldmont-plus".
8979 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8980 (ix86_option_override_internal): Add "goldmont-plus".
8981 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8982 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8983 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8984 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8986 2018-05-17 Richard Biener <rguenther@suse.de>
8988 PR tree-optimization/85757
8989 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8990 remove defs that only feed that PHI from further processing.
8992 2018-05-16 Jim Wilson <jimw@sifive.com>
8994 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8996 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8998 2018-05-16 Mark Wielaard <mark@klomp.org>
9000 * dwarf2out.c (count_index_strings): New function.
9001 (output_indirect_strings): Call count_index_strings and generate
9002 header for dwarf_version >= 5.
9004 2018-05-16 Mark Wielaard <mark@klomp.org>
9006 * dwarf2out.c (dwarf_FORM): New function.
9007 (set_indirect_string): Use dwarf_FORM.
9008 (reset_indirect_string): Likewise.
9009 (size_of_die): Likewise.
9010 (value_format): Likewise.
9011 (output_die): Likewise.
9012 (add_skeleton_AT_string): Likewise.
9013 (output_macinfo_op): Likewise.
9014 (index_string): Likewise.
9015 (output_index_string_offset): Likewise.
9016 (output_index_string): Likewise.
9017 (count_index_strings): Likewise.
9019 2018-05-16 Carl Love <cel@us.ibm.com>
9021 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9022 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
9024 2018-05-16 Martin Jambor <mjambor@suse.cz>
9026 * ipa-prop.c (ipa_free_all_edge_args): Remove.
9027 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
9029 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
9031 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
9032 (fnma<mode>4): Likewise.
9033 (fms<mode>4): Likewise.
9034 (fnms<mode>4): Likewise.
9035 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
9036 (aarch64_fnma<mode>4): Likewise.
9037 (aarch64_fms<mode>4): Likewise.
9038 (aarch64_fnms<mode>4): Likewise.
9039 (aarch64_fnmadd<mode>4): Likewise.
9041 2018-05-16 Jason Merrill <jason@redhat.com>
9043 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
9045 2018-05-16 Richard Biener <rguenther@suse.de>
9047 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
9048 (dump_stmt_cost): Declare.
9049 (add_stmt_cost): Dump cost we add.
9050 (add_stmt_costs): New function.
9051 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
9053 (vect_analyze_stmt): Adjust prototype.
9054 (vectorizable_condition): Likewise.
9055 (vectorizable_live_operation): Likewise.
9056 (vectorizable_reduction): Likewise.
9057 (vectorizable_induction): Likewise.
9058 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
9059 cost vector to pass to vectorizable_ and record afterwards.
9060 (vect_model_reduction_cost): Take cost vector argument and adjust.
9061 (vect_model_induction_cost): Likewise.
9062 (vectorizable_reduction): Likewise.
9063 (vectorizable_induction): Likewise.
9064 (vectorizable_live_operation): Likewise.
9065 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
9066 SLP_TREE_NUMBER_OF_VEC_STMTS.
9067 (vect_analyze_slp_cost_1): Remove.
9068 (vect_analyze_slp_cost): Likewise.
9069 (vect_slp_analyze_node_operations): Take visited args and
9070 a target cost vector. Avoid processing already visited stmt sets.
9071 (vect_slp_analyze_operations): Use a local cost vector to gather
9072 costs and register those of non-discarded instances.
9073 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
9074 (vect_schedule_slp_instance): Remove copying of
9075 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
9077 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
9078 adding cost. Record cost entry location.
9079 (vect_prologue_cost_for_slp_op): Function to compute cost of
9080 a constant or invariant generated for SLP vect in the prologue,
9081 split out from vect_analyze_slp_cost_1.
9082 (vect_model_simple_cost): Make static. Adjust for SLP costing.
9083 (vect_model_promotion_demotion_cost): Likewise.
9084 (vect_model_store_cost): Likewise, make static.
9085 (vect_model_load_cost): Likewise.
9086 (vectorizable_bswap): Add cost vector arg and adjust.
9087 (vectorizable_call): Likewise.
9088 (vectorizable_simd_clone_call): Likewise.
9089 (vectorizable_conversion): Likewise.
9090 (vectorizable_assignment): Likewise.
9091 (vectorizable_shift): Likewise.
9092 (vectorizable_operation): Likewise.
9093 (vectorizable_store): Likewise.
9094 (vectorizable_load): Likewise.
9095 (vectorizable_condition): Likewise.
9096 (vectorizable_comparison): Likewise.
9097 (can_vectorize_live_stmts): Likewise.
9098 (vect_analyze_stmt): Likewise.
9099 (vect_transform_stmt): Adjust calls to vectorizable_*.
9100 * tree-vectorizer.c: Include gimple-pretty-print.h.
9101 (dump_stmt_cost): New function.
9103 2018-05-16 Richard Biener <rguenther@suse.de>
9105 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
9106 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
9107 * tree-ssa-dse.c: Include tree-ssa-loop.h.
9108 (check_name): New callback.
9109 (dse_classify_store): Track cycles via a visited bitmap of PHI
9110 defs and simplify handling of in-loop and across loop dead stores
9111 and properly fail for loop-variant refs. Handle byte-tracking with
9112 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
9115 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
9117 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
9118 (vect_get_mask_type_for_stmt): Likewise.
9119 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
9121 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
9122 to determine the statement's vector type and the vector type that
9123 should be used for calculating nunits. Deal with cases in which
9124 the type has to be deferred.
9125 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
9126 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
9127 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
9128 (vect_determine_vf_for_stmt): New functions, split out from...
9129 (vect_determine_vectorization_factor): ...here.
9130 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
9131 (vect_get_mask_type_for_stmt): New functions, split out from
9132 vect_determine_vectorization_factor.
9134 2018-05-16 Richard Biener <rguenther@suse.de>
9136 * tree-cfg.c (verify_gimple_assign_ternary): Properly
9137 verify the [VEC_]COND_EXPR embedded comparison.
9139 2018-05-15 Martin Sebor <msebor@redhat.com>
9141 PR tree-optimization/85753
9142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
9143 RECORD_TYPE in addition to ARRAY_TYPE.
9145 2018-05-15 Martin Sebor <msebor@redhat.com>
9148 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
9150 2018-05-15 Richard Biener <rguenther@suse.de>
9152 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
9153 add by_clobber_p one. Change algorithm to collect all defs
9154 representing uses we need to walk and try reducing them to
9155 a single one before failing.
9156 (dse_dom_walker::dse_optimize_stmt): Adjust.
9158 2018-05-13 Mark Wielaard <mark@klomp.org>
9160 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
9161 (size_of_loc_descr): Likewise.
9162 (output_loc_operands): Likewise.
9163 (output_loc_operands_raw): Likewise.
9164 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
9165 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
9166 (hash_loc_operands): Likewise.
9167 (compare_loc_operands): Likewise.
9169 2018-05-14 Mark Wielaard <mark@klomp.org>
9171 * dwarf2out.c (count_index_addrs): New function.
9172 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
9174 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9176 PR tree-optimization/83648
9177 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
9178 return value as malloc candidate.
9180 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9183 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
9184 param as true in call to suggest_attribute.
9186 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
9188 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
9189 -mreadonly-in-sdata.
9191 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9193 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
9195 (aarch64_crypto_aesd_fused): Likewise.
9197 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
9199 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
9200 (movsi_aarch64): Likewise.
9201 (load_pairsi): Likewise.
9202 (load_pairdi): Likewise.
9203 (store_pairsi): Likewise.
9204 (store_pairdi): Likewise.
9205 (load_pairsf): Likewise.
9206 (load_pairdf): Likewise.
9207 (store_pairsf): Likewise.
9208 (store_pairdf): Likewise.
9209 (zero_extend): Likewise.
9210 (trunc): Swap alternatives.
9211 (fcvt_target): Add '?' to prefer w over r.
9213 2018-05-14 Jakub Jelinek <jakub@redhat.com>
9216 * config/i386/i386.md: Disallow non-commutative arithmetics in
9217 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
9218 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
9219 in the peephole2 before it.
9221 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
9223 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
9224 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
9225 (ix86_handle_option): Handle -mcldemote.
9226 * config.gcc: New header.
9227 * config/i386/cldemoteintrin.h: New file.
9228 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
9229 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
9231 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9232 OPTION_MASK_ISA_CLDEMOTE.
9233 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
9234 (ix86_valid_target_attribute_inner_p): Ditto.
9235 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
9236 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
9237 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
9238 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
9239 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
9241 * config/i386/i386.opt: Add -mcldemote.
9242 * config/i386/x86intrin.h: New header.
9243 * doc/invoke.texi: Add -mcldemote.
9245 2018-05-14 Richard Biener <rguenther@suse.de>
9247 * doc/match-and-simplify.texi: Adjust :s documentation.
9249 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
9251 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
9252 intended memcpy size.
9253 (REORDER_45): Likewise.
9255 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
9257 * sort.cc: New file.
9258 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
9259 * vec.c (qsort_chk): Use gcc_qsort.
9260 * Makefile.in (OBJS-libcommon): Add sort.o.
9261 (build/sort.o): New target. Use it...
9262 (BUILD_RTL): ... here, and...
9263 (build/gencfn-macros): ... here, and...
9264 (build/genmatch): ... here.
9266 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9267 Chung-Ju Wu <jasonwucj@gmail.com>
9269 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
9270 * config/nds32/nds32-graywolf.md: New file.
9271 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
9272 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
9274 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
9275 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
9276 * config/nds32/nds32.md (pipeline_model): Add graywolf.
9277 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
9278 * config/nds32/pipelines.md: Include n15 settings.
9280 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9281 Chung-Ju Wu <jasonwucj@gmail.com>
9283 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
9284 * config/nds32/nds32-n13.md: New file.
9285 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
9286 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
9288 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
9289 * config/nds32/nds32.md (pipeline_model): Add n13.
9290 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
9291 * config/nds32/pipelines.md: Include n13 settings.
9293 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9294 Chung-Ju Wu <jasonwucj@gmail.com>
9296 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
9297 * config/nds32/nds32-n10.md: New file.
9298 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
9299 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
9301 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
9302 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
9303 * config/nds32/nds32.md (pipeline_model): Add n10.
9304 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
9305 * config/nds32/pipelines.md: Include n10 settings.
9307 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
9308 Kito Cheng <kito.cheng@gmail.com>
9309 Chung-Ju Wu <jasonwucj@gmail.com>
9311 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
9312 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9313 Add enum values for DSP extension instructions.
9314 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
9316 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
9317 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
9319 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
9320 * config/nds32/nds32-dspext.md: New file for DSP implementation.
9321 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
9322 * config/nds32/nds32-intrinsic.md: Likewise.
9323 * config/nds32/nds32_intrinsic.h: Likewise.
9324 * config/nds32/nds32-md-auxiliary.c: Likewise.
9325 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
9326 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
9327 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
9328 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
9329 * config/nds32/nds32-protos.h: New declarations for DSP extension.
9330 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
9331 TYPE_DMAC in switch statement.
9332 * config/nds32/nds32.c: New checking and implementation for DSP
9333 extension instructions.
9334 * config/nds32/nds32.h: Likewise.
9335 * config/nds32/nds32.md: Likewise.
9336 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
9337 * config/nds32/predicates.md: Implement new predicates for DSP
9340 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9342 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
9343 Reformat alternatives and attributes so it is easier to identify
9344 which constraints/attributes go with which instruction.
9345 (mov<mode>_hardfloat32, FMOVE64): Likewise.
9346 (mov<mode>_softfloat32, FMOVE64): Likewise.
9347 (mov<mode>_hardfloat64, FMOVE64): Likewise.
9348 (mov<mode>_softfloat64, FMOVE64): Likewise.
9350 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
9352 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9354 (Basic PowerPC Built-in Functions): The new name of the
9355 subsection previously known as "PowerPC Built-in Functions".
9356 (Basic PowerPC Built-in Functions Available on all Configurations):
9358 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
9359 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
9360 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
9361 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
9363 2018-05-11 Martin Jambor <mjambor@suse.cz>
9366 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
9369 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9372 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
9374 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
9376 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
9377 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
9378 (ix86_handle_option): Handle -mwaitpkg.
9379 * config.gcc: New header.
9380 * config/i386/cpuid.h (bit_WAITPKG): New bit.
9381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
9382 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
9384 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9385 OPTION_MASK_ISA_WAITPKG.
9386 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
9387 (ix86_option_override_internal): Add PTA_WAITPKG.
9388 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
9389 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
9390 IX86_BUILTIN_TPAUSE.
9391 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
9392 __builtin_ia32_umwait and __builtin_ia32_tpause.
9393 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
9394 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
9395 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
9396 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
9397 UNSPECV_TPAUSE): New.
9398 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
9399 * config/i386/i386.opt: Add -mwaitpkg.
9400 * config/i386/waitpkgintrin.h: New file.
9401 * config/i386/x86intrin.h: New header.
9402 * doc/invoke.texi: Add -mwaitpkg.
9404 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9407 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
9409 (cortex-m0): Use armv6s-m isa.
9410 (cortex-m0plus): Likewise.
9411 (cortex-m1): Likewise.
9412 (cortex-m0.small-multiply): Likewise.
9413 (cortex-m0plus.small-multiply): Likewise.
9414 (cortex-m1.small-multiply): Likewise.
9416 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
9417 Jakub Jelinek <jakub@redhat.com>
9419 PR tree-optimization/85692
9420 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
9421 source permute as well.
9423 2018-05-11 Martin Liska <mliska@suse.cz>
9426 * doc/extend.texi: Document LLVM style format for no_sanitize
9429 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9431 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
9432 mode_supports_vsx_dform_quad to mode_supports_dq_form.
9433 (mode_supports_vsx_dform_quad): Likewise.
9434 (mode_supports_vmx_dform): Move these functions to be next to the
9435 other mode_supports functions.
9436 (mode_supports_dq_form): Likewise.
9437 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
9438 mode_supports_dq_form.
9439 (reg_offset_addressing_ok_p): Likewise.
9440 (offsettable_ok_by_alignment): Likewise.
9441 (rs6000_legitimate_offset_address_p): Likewise.
9442 (legitimate_lo_sum_address_p): Likewise.
9443 (rs6000_legitimize_address): Likewise.
9444 (rs6000_legitimize_reload_address): Likewise.
9445 (rs6000_secondary_reload_inner): Likewise.
9446 (rs6000_preferred_reload_class): Likewise.
9447 (rs6000_output_move_128bit): Likewise.
9449 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9451 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
9452 Generate SImode target register for null target.
9453 <case IX86_BUILTIN_XGETBV>: Ditto.
9454 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
9455 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
9457 2018-05-10 Carl Love <cel@us.ibm.com>
9459 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9460 dcbtt and dcbtstt if operands[2] is 0.
9462 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9465 * config/i386/sse.md (usadv64qi): New expander.
9467 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
9469 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
9470 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
9471 -maltivec=be support.
9472 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
9473 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
9474 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
9475 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
9476 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
9477 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
9478 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
9479 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
9480 altivec_vsumsws): Adjust.
9481 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
9482 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
9483 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
9485 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
9486 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
9487 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
9488 (altivec_lve<VI_char>x): Delete expand.
9489 (*altivec_lve<VI_char>x_internal): Rename to...
9490 (altivec_lve<VI_char>x): ... this.
9491 (altivec_lvxl_<mode>): Delete expand.
9492 (*altivec_lvxl_<mode>_internal): Rename to ...
9493 (altivec_lvxl_<mode>): ... this.
9494 (altivec_stvxl_<mode>): Delete expand.
9495 (*altivec_stvxl_<mode>_internal): Rename to ...
9496 (altivec_stvxl_<mode>): ... this.
9497 (altivec_stve<VI_char>x): Delete expand.
9498 (*altivec_stve<VI_char>x_internal): Rename to ...
9499 (altivec_stve<VI_char>x): ... this.
9500 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
9501 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
9502 reduc_plus_scal_<mode>): Adjust.
9503 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
9505 (rs6000_cpu_cpp_builtins): Adjust.
9506 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
9507 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
9508 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
9509 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
9510 -maltivec=be support.
9511 (rs6000_split_vec_extract_var): Adjust.
9512 (rs6000_split_v4si_init): Adjust.
9513 (swap_selector_for_mode): Delete.
9514 (altivec_expand_lvx_be, altivec_expand_stvx_be,
9515 altivec_expand_stvex_be): Delete.
9516 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
9517 -maltivec=be support.
9518 (rs6000_gimple_fold_builtin): Ditto.
9519 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
9521 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
9522 (TARGET_DIRECT_MOVE_64BIT): Adjust.
9523 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
9524 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
9525 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
9526 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
9527 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
9528 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
9529 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
9530 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
9531 anonymous split): Adjust.
9532 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
9533 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
9535 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
9537 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
9538 when --with-gxx-include-dir is also specified.
9539 * configure: Regenerate.
9541 2018-05-09 Jim Wilson <jimw@sifive.com>
9544 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
9545 * config/riscv/t-withmultilib: New.
9546 * config/riscv/withmultilib.h: New.
9547 * doc/install.texi: Document RISC-V --with-multilib-list support.
9549 2018-05-09 Richard Biener <rguenther@suse.de>
9551 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
9553 (vect_bb_vectorization_profitable_p): Adjust. Compute
9554 actual scalar cost using the cost vector and the add_stmt_cost
9557 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9559 PR rtl-optimization/85645
9560 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
9561 in the REG_CFA_REGISTER note for LR, don't leave it empty.
9563 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9565 PR rtl-optimization/85645
9566 * shrink-wrap.c (spread_components): Return a boolean saying if
9567 anything was changed.
9568 (try_shrink_wrapping_separate): Iterate spread_components until
9569 nothing changes anymore.
9571 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9573 PR rtl-optimization/85645
9574 * regrename.c (build_def_use): Also kill the chains that include the
9575 destination of a REG_CFA_REGISTER note.
9577 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9579 PR rtl-optimization/85645
9580 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
9581 insn that has a REG_CFA_REGISTER note.
9583 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
9585 * cfgexpand.c (expand_clobber): New function.
9586 (expand_gimple_stmt_1): Use it.
9587 * tree-vect-stmts.c (vect_clobber_variable): New function,
9589 (vectorizable_simd_clone_call): ...here.
9590 (vectorizable_store): Emit a clobber either side of an
9591 IFN_STORE_LANES sequence.
9592 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
9594 2018-05-09 Tom de Vries <tom@codesourcery.com>
9597 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
9598 (define_insn "trap_if_false"): Add exit after trap.
9600 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
9602 PR rtl-optimization/85638
9603 * bb-reorder.c: Include common/common-target.h.
9604 (create_forwarder_block): New function extracted from...
9605 (fix_up_crossing_landing_pad): ...here. Rename into...
9606 (dw2_fix_up_crossing_landing_pad): ...this.
9607 (sjlj_fix_up_crossing_landing_pad): New function.
9608 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
9609 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
9610 from both partitions and exit the loop after one iteration.
9612 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9615 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9617 (Basic PowerPC Built-in Functions): The new name of the
9618 subsection previously known as "PowerPC Built-in Functions".
9619 (Basic PowerPC Built-in Functions Available on all Configurations):
9621 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9623 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9625 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9627 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9630 2018-05-08 Jim Wilson <jimw@sifive.com>
9632 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
9633 (LD_EMUL_SUFFIX): New.
9634 (LINK_SPEC): Use it.
9636 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9638 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9640 (Basic PowerPC Built-in Functions): The new name of the
9641 subsection previously known as "PowerPC Built-in Functions".
9642 (Basic PowerPC Built-in Functions Available on all Configurations):
9644 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9646 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9648 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9650 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9653 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9656 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
9657 after cmpelim optimization.
9659 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
9661 * config.gcc: Support "goldmont".
9662 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
9663 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9665 * config/i386/i386.c (m_GOLDMONT): Define.
9666 (processor_target_table): Add "goldmont".
9667 (PTA_GOLDMONT): Define.
9668 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
9669 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
9670 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
9671 (fold_builtin_cpu): Add "goldmont".
9672 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
9673 (ix86_option_override_internal): Add "goldmont".
9674 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
9675 (processor_type): Add PROCESSOR_GOLDMONT.
9676 * config/i386/i386.md: Add CPU "glm".
9677 * config/i386/glm.md: New file.
9678 * config/i386/x86-tune.def: Add m_GOLDMONT.
9679 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
9681 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9684 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
9686 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
9687 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
9688 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
9691 * config/i386/i386.c (ix86_fold_builtin): Handle
9692 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
9695 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
9696 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
9698 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
9701 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
9702 (check_arch): Likewise.
9703 (check_fpu): Return the result rather than printing it.
9704 (end arch): Fix operator precedence.
9705 (end cpu): Likewise.
9706 (END): Print the result from check_fpu.
9708 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9709 Alan Hayward <alan.hayward@arm.com>
9710 David Sherwood <david.sherwood@arm.com>
9712 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
9713 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
9714 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
9715 (*fcmuo<mode>_and): New patterns.
9717 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9719 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
9720 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
9721 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
9722 (cmp_op, sve_imm_con): New code attributes.
9723 (SVE_COND_INT_CMP, imm_con): Delete.
9724 (cmp_op): Remove above unspecs from int attribute.
9725 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
9727 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
9728 comparison-specific unspecs.
9729 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
9730 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
9731 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
9732 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
9733 (*vec_fcm<cmp_op><mode>): Rename to...
9734 (*fcm<cmp_op><mode>): ...this and adjust likewise.
9735 (*vec_fcmuo<mode>): Rename to...
9736 (*fcmuo<mode>): ...this and adjust likewise.
9737 (*pred_fcm<cmp_op><mode>): New pattern.
9738 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
9739 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
9741 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
9743 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
9744 (aarch64_emit_sve_predicated_cond): New function.
9745 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
9746 (aarch64_emit_unspec_cond_or): Replace with...
9747 (aarch64_emit_sve_or_conds): ...this new function. Use
9748 aarch64_emit_sve_ptrue_op for the individual comparisons and
9749 aarch64_emit_binop to OR them together.
9750 (aarch64_emit_inverted_unspec_cond): Replace with...
9751 (aarch64_emit_sve_inverted_cond): ...this new function. Use
9752 aarch64_emit_sve_ptrue_op for the comparison and
9753 aarch64_emit_unop to invert the result.
9754 (aarch64_expand_sve_vec_cmp_float): Update after the above
9755 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
9757 2018-05-07 Nathan Sidwell <nathan@acm.org>
9759 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
9760 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
9761 (Backwards Compatibility): Likewise.
9763 2018-05-07 Luis Machado <luis.machado@linaro.org>
9767 2018-05-07 Luis Machado <luis.machado@linaro.org>
9769 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9770 <prefetch_dynamic_strides>: New const bool field.
9771 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9772 prefetch_dynamic_strides.
9773 (exynosm1_prefetch_tune): Likewise.
9774 (thunderxt88_prefetch_tune): Likewise.
9775 (thunderx_prefetch_tune): Likewise.
9776 (thunderx2t99_prefetch_tune): Likewise.
9777 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9779 (aarch64_override_options_internal): Update to set
9780 PARAM_PREFETCH_DYNAMIC_STRIDES.
9781 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9782 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9783 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9784 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9785 prefetch-dynamic-strides setting.
9787 2018-05-07 Luis Machado <luis.machado@linaro.org>
9789 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9790 <minimum_stride>: New const int field.
9791 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9792 minimum_stride field.
9793 (exynosm1_prefetch_tune): Likewise.
9794 (thunderxt88_prefetch_tune): Likewise.
9795 (thunderx_prefetch_tune): Likewise.
9796 (thunderx2t99_prefetch_tune): Likewise.
9797 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9798 (aarch64_override_options_internal): Update to set
9799 PARAM_PREFETCH_MINIMUM_STRIDE.
9800 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9801 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9802 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9803 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9804 stride is constant and is below the minimum stride threshold.
9806 2018-05-07 Luis Machado <luis.machado@linaro.org>
9808 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9811 2018-05-07 Luis Machado <luis.machado@linaro.org>
9813 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9814 <prefetch_dynamic_strides>: New const bool field.
9815 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9816 prefetch_dynamic_strides.
9817 (exynosm1_prefetch_tune): Likewise.
9818 (thunderxt88_prefetch_tune): Likewise.
9819 (thunderx_prefetch_tune): Likewise.
9820 (thunderx2t99_prefetch_tune): Likewise.
9821 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9823 (aarch64_override_options_internal): Update to set
9824 PARAM_PREFETCH_DYNAMIC_STRIDES.
9825 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9826 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9827 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9828 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9829 prefetch-dynamic-strides setting.
9831 2018-05-07 Luis Machado <luis.machado@linaro.org>
9833 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9834 <minimum_stride>: New const int field.
9835 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9836 minimum_stride field.
9837 (exynosm1_prefetch_tune): Likewise.
9838 (thunderxt88_prefetch_tune): Likewise.
9839 (thunderx_prefetch_tune): Likewise.
9840 (thunderx2t99_prefetch_tune): Likewise.
9841 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9842 (aarch64_override_options_internal): Update to set
9843 PARAM_PREFETCH_MINIMUM_STRIDE.
9844 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9845 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9846 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9847 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9848 stride is constant and is below the minimum stride threshold.
9850 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9853 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9854 the type is addressable. Don't force op into register if it has
9857 2018-05-05 Roland McGrath <mcgrathr@google.com>
9860 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9861 any section for which we don't know a specific type it should have,
9862 regardless of name. Previously this was done only for the exact
9863 names ".init_array", ".fini_array", and ".preinit_array".
9864 (default_elf_asm_named_section): Add comment about
9865 relationship with default_section_type_flags and SECTION_NOTYPE.
9866 (get_section): Don't consider it a type conflict if one side has
9867 SECTION_NOTYPE and the other doesn't, as long as neither has the
9868 SECTION_BSS et al used in the default_section_type_flags logic.
9870 2018-05-05 Tom de Vries <tom@codesourcery.com>
9873 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9874 (workaround_barsyncs): New function.
9875 (nvptx_reorg): Use workaround_barsyncs.
9876 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9877 (define_expand "nvptx_membar_cta"): New define_expand.
9878 (define_insn "*nvptx_membar_cta"): New insn.
9880 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9882 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9883 To improve optimization opportunities.
9884 * builtin-types.def: The new needed builtin types for the above.
9886 2018-05-04 Richard Biener <rguenther@suse.de>
9888 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9889 * gimple-ssa-store-merging.c
9890 (imm_store_chain_info::output_merged_store): Remove redundant create,
9891 release split_store vector contents on failure.
9892 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9893 scalar stmt vector on cache hit.
9895 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9897 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9899 * config.gcc (powerpc-xilinx-eabi*): Remove.
9900 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9902 (fusion_addis_mem_combo_load): Ditto.
9903 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9905 (rs6000_cpu_cpp_builtins): Ditto.
9906 * config/rs6000/rs6000-linux.c
9907 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9908 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9909 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9911 (rs6000_setup_reg_addr_masks): Ditto.
9912 (rs6000_init_hard_regno_mode_ok): Ditto.
9913 (rs6000_option_override_internal): Ditto.
9914 (legitimate_lo_sum_address_p): Ditto.
9915 (rs6000_legitimize_address): Ditto.
9916 (rs6000_legitimize_reload_address): Ditto.
9917 (rs6000_legitimate_address_p): Ditto.
9918 (abi_v4_pass_in_fpr): Ditto.
9919 (setup_incoming_varargs): Ditto.
9920 (rs6000_gimplify_va_arg): Ditto.
9921 (rs6000_split_multireg_move): Ditto.
9922 (rs6000_savres_strategy): Ditto.
9923 (rs6000_emit_prologue_components): Ditto.
9924 (rs6000_emit_epilogue_components): Ditto.
9925 (rs6000_emit_prologue): Ditto.
9926 (rs6000_emit_epilogue): Ditto.
9927 (rs6000_elf_file_end): Ditto.
9928 (rs6000_function_value): Ditto.
9929 (rs6000_libcall_value): Ditto.
9930 * config/rs6000/rs6000.h: Ditto.
9931 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9932 (TARGET_MINMAX): ... this. New.
9933 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9934 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9935 (*movsi_internal1_single): Delete.
9936 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9937 mfpu=, mxilinx-fpu): Delete.
9938 * config/rs6000/singlefp.h: Delete.
9939 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9940 * config/rs6000/t-rs6000: Ditto.
9941 * config/rs6000/t-xilinx: Delete.
9942 * config/rs6000/titan.md: Adjust for fp_type removal.
9943 * config/rs6000/vsx.md: Remove Xilinx FP support.
9944 (VStype_simple): Delete.
9945 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9946 * config/rs6000/xfpu.h: Delete.
9947 * config/rs6000/xfpu.md: Delete.
9948 * config/rs6000/xilinx.h: Delete.
9949 * config/rs6000/xilinx.opt: Delete.
9950 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9951 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9953 2018-05-04 Tom de Vries <tom@codesourcery.com>
9956 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9959 2018-05-04 Richard Biener <rguenther@suse.de>
9962 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9963 (expand_complex_div_wide): Likewise.
9964 (expand_complex_operations_1): Likewise.
9965 (expand_complex_libcall): Preserve EH info of the original stmt.
9966 (tree_lower_complex): Handle removed blocks.
9967 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9968 on complex multiplication and division libcall builtins.
9970 2018-05-04 Richard Biener <rguenther@suse.de>
9973 * fold-const.c (negate_expr_p): Restrict negation of operand
9974 zero of a division to when we know that can happen without
9976 (fold_negate_expr_1): Likewise.
9978 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9981 * real.h (real_nextafter): Declare.
9982 * real.c (real_nextafter): New function.
9983 * fold-const-call.c (fold_const_nextafter): New function.
9984 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9985 CASE_CFN_NEXTTOWARD.
9986 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9987 even when arg1_mode is different from arg0_mode.
9989 2018-05-03 Nathan Sidwell <nathan@acm.org>
9991 * doc/extend.texi (Deprecated Features): Remove
9993 (Backwards Compatibility): Likewise.
9994 * doc/invoke.texi (C++ Language Options): Likewise.
9995 (C++ Dialect Options): Likewise.
9997 2018-05-03 Jakub Jelinek <jakub@redhat.com>
10000 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
10001 _mm512_mask_mullox_epi64): New intrinsics.
10003 2018-05-03 Tom de Vries <tom@codesourcery.com>
10006 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10007 dump files): Add offload-tree.
10009 2018-05-03 Richard Biener <rguenther@suse.de>
10011 PR tree-optimization/85615
10012 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
10013 to loops not nested in BBs loop father to avoid creating multi-entry
10016 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10018 PR tree-optimization/70291
10019 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
10020 arguments. Change return type to tree. Emit libcall as a new
10021 statement rather than replacing existing one when inplace_p is true.
10022 (expand_complex_multiplication_components): New function.
10023 (expand_complex_multiplication): Expand floating-point complex
10024 multiplication using the above.
10025 (expand_complex_division): Rename inner_type parameter to type.
10026 Update expand_complex_libcall call-site.
10027 (expand_complex_operations_1): Update expand_complex_multiplication
10028 and expand_complex_division call-sites.
10030 2018-05-02 Jakub Jelinek <jakub@redhat.com>
10033 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
10034 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
10035 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
10036 the highest significant bit of the shift count mask is clear. In
10037 check whether and[sq]i3 is needed verify that all significant bits
10038 of the shift count other than the highest are set.
10040 2018-05-02 Tom de Vries <tom@codesourcery.com>
10043 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
10044 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
10045 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
10046 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
10047 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
10048 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
10049 __builtin_goacc_parlevel_size.
10051 2018-05-02 Richard Biener <rguenther@suse.de>
10053 PR tree-optimization/85597
10054 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
10055 do not use split vect_get_vec_defs call but call vect_get_slp_defs
10058 2018-05-02 Tom de Vries <tom@codesourcery.com>
10061 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10062 dump files): Add ltrans-tree.
10064 2018-05-02 Tom de Vries <tom@codesourcery.com>
10067 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10068 dump files): Add wpa-ipa.
10070 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
10072 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
10073 powerpc*-*-linux*paired* target.
10074 * config/rs6000/750cl.h: Delete.
10075 * config/rs6000/paired.h: Delete.
10076 * config/rs6000/paired.md: Delete.
10077 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
10079 * config/rs6000/rs6000-builtin.def: Remove paired float support.
10080 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
10081 comment. Remove paired float support.
10082 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
10083 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
10085 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
10086 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
10088 * config/rs6000/rs6000.c: Remove paired float support.
10089 (paired_expand_vector_init, paired_expand_vector_move,
10090 paired_emit_vector_compare, paired_emit_vector_cond_expr,
10091 (paired_expand_lv_builtin, paired_expand_stv_builtin,
10092 paired_expand_builtin, paired_expand_predicate_builtin,
10093 paired_init_builtins): Delete.
10094 * config/rs6000/rs6000.h: Remove paired float support.
10095 * config/rs6000/rs6000.md: Remove paired float support.
10096 (move_from_CR_ov_bit): Delete.
10097 * config/rs6000/rs6000.opt (mpaired): Delete.
10098 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
10099 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
10101 2018-05-02 Richard Biener <rguenther@suse.de>
10103 PR middle-end/85567
10104 * gimplify.c (gimplify_save_expr): When in SSA form allow
10105 SAVE_EXPRs to compute to SSA vars.
10107 2018-05-02 Jakub Jelinek <jakub@redhat.com>
10110 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
10111 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
10112 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
10113 clobber operands[2], instead use a new pseudo. Formatting fixes.
10115 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
10117 PR tree-optimization/85586
10118 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
10119 exit early for statements in the same group if the accesses are
10122 2018-05-02 Tom de Vries <tom@codesourcery.com>
10125 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
10128 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
10130 PR tree-optimization/85143
10131 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
10133 2018-05-01 Tom de Vries <tom@codesourcery.com>
10136 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
10137 not found" error message.
10139 2018-05-01 Tom de Vries <tom@codesourcery.com>
10142 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
10143 * vec.c (test_ordered_remove_if): New function.
10144 (vec_c_tests): Call test_ordered_remove_if.
10145 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
10146 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
10147 * tree-vect-patterns.c (vect_pattern_recog_1): Use
10148 VEC_ORDERED_REMOVE_IF.
10150 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10152 PR tree-optimization/82665
10153 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
10154 pointer subtraction where arguments come from a memchr call.
10156 2018-05-01 Jakub Jelinek <jakub@redhat.com>
10158 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
10159 --push-state --as-needed and --pop-state instead of --as-needed and
10160 --no-as-needed if ld supports it.
10161 * configure: Regenerated.
10164 * doc/install.texi2html: Replace _002d with - and _002a with * in
10165 generated html files using sed.
10167 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10170 * gcc-rich-location.c (blank_line_before_p): New function.
10171 (use_new_line): New function.
10172 (gcc_rich_location::add_fixit_insert_formatted): New function.
10173 * gcc-rich-location.h
10174 (gcc_rich_location::add_fixit_insert_formatted): New function.
10176 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10178 * selftest.c (assert_streq): Rename "expected" and "actual" to
10179 "val1" and "val2". Extend NULL-handling to cover both inputs
10180 symmetrically, while still requiring both to be non-NULL for a pass.
10181 * selftest.h (assert_streq): Rename "expected" and "actual" to
10183 (ASSERT_EQ): Likewise.
10184 (ASSERT_EQ_AT): Likewise.
10185 (ASSERT_KNOWN_EQ): Likewise.
10186 (ASSERT_KNOWN_EQ_AT): Likewise.
10187 (ASSERT_NE): Likewise.
10188 (ASSERT_MAYBE_NE): Likewise.
10189 (ASSERT_MAYBE_NE_AT): Likewise.
10190 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
10191 the assertion to pass.
10192 (ASSERT_STREQ_AT): Likewise.
10194 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10196 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
10197 interaction with -pie.
10199 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10201 * selftest.h: Fix alphabetization of per-source-file selftest
10204 2018-04-30 Jason Merrill <jason@redhat.com>
10206 PR c++/61982 - dead stores to destroyed objects.
10207 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
10210 2018-04-30 Jason Merrill <jason@redhat.com>
10212 * tree.c (build_clobber): New.
10213 * tree.h: Declare it.
10214 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
10216 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10218 * diagnostic-show-locus.c (layout::layout): Update for
10219 location_get_source_line returning a char_span.
10220 (struct char_span): Move to input.h.
10221 (struct correction): Update for fields in char_span becoming
10223 (struct source_line): Update for location_get_source_line
10224 returning a char_span.
10225 (layout::print_line): Likewise.
10226 * edit-context.c (edited_file::print_content): Likewise.
10227 (edited_file::print_diff_hunk): Likewise.
10228 (edited_file::print_run_of_changed_lines): Likewise.
10229 (edited_file::get_num_lines): Likewise.
10230 (edited_line::edited_line): Likewise.
10231 * final.c (asm_show_source): Likewise.
10232 * input.c (location_get_source_line): Convert return type
10233 from const char * to char_span, losing the final "line_len"
10235 (dump_location_info): Update for the above.
10236 (get_substring_ranges_for_loc): Likewise. Use a char_span
10237 when handling the literal within the line.
10238 (test_reading_source_line): Update for location_get_source_line
10239 returning a char_span.
10240 * input.h (class char_span): Move here from
10241 diagnostic-show-locus.c, converting from a struct to a class.
10242 Make data members private.
10243 (char_span::operator bool): New.
10244 (char_span::length): New.
10245 (char_span::get_buffer): New.
10246 (char_span::operator[]): New.
10247 (char_span::subspan): Make const.
10248 (char_span::xstrdup): New.
10249 (location_get_source_line): Convert return type from const char *
10250 to char_span, losing the final "line_size" param.
10252 2018-04-30 Jan Hubicka <jh@suse.cz>
10254 * lto-wrapper.c (ltrans_priorities): New static var.
10255 (cmp_priority): New.
10256 (run_gcc): Read priorities and if doing parallel build order
10257 the Makefile by them.
10259 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10261 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
10263 2018-04-30 Richard Biener <rguenther@suse.de>
10265 * tree-cfg.c (verify_address): Remove base argument, add
10266 flag whether to check TREE_ADDRESSABLE and do that.
10267 (verify_expr): Remove.
10268 (verify_types_in_gimple_reference): Add pieces from verify_expr.
10269 (verify_gimple_assign_single): Likewise.
10270 (verify_gimple_switch): Likewise.
10271 (verify_expr_location_1): Dereference tp once. Add (disabled)
10272 piece from verify_expr.
10273 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
10275 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10277 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
10279 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10281 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
10282 (small_data_pattern): Likewise.
10283 (arc_rewrite_small_data): Likewise.
10284 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
10285 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
10286 (get_symbol_alignment): New function.
10287 (legitimate_small_data_address_p): Likewise.
10288 (legitimate_scaled_address): Update, call
10289 legitimate_small_data_address_p.
10290 (output_sdata): New static variable.
10291 (arc_print_operand): Update how we handle small data operands.
10292 (arc_print_operand_address): Likewise.
10293 (arc_legitimate_address_p): Update, use
10294 legitimate_small_data_address_p.
10295 (arc_rewrite_small_data_p): Remove.
10296 (arc_rewrite_small_data_1): Likewise.
10297 (arc_rewrite_small_data): Likewise.
10298 (small_data_pattern): Likewise.
10299 (compact_sda_memory_operand): Update to use
10300 legitimate_small_data_address_p and get_symbol_alignment.
10301 (prepare_move_operands): Don't rewite sdata pattern.
10302 (prepare_extend_operands): Remove.
10303 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
10305 (zero_extendqisi2): Likewise.
10306 (zero_extendhisi2): Likewise.
10307 (extendqihi2): Likewise.
10308 (extendqisi2): Likewise.
10309 (extendhisi2): Likewise.
10310 (addsi3): Likewise.
10311 (subsi3): Likewise.
10312 (andsi3): Likewise.
10313 * config/arc/constraints.md (Usd): Change it to memory constraint.
10315 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10317 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
10318 as source of std instructions.
10319 * config/arc/arc.md (movsi_insn): Update pattern predicate to
10320 allow 6-bit constants as source for store instructions.
10321 (movdi_insn): Update instruction pattern to allow 6-bit constants
10322 as source for store instructions.
10324 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10326 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
10328 2018-04-30 Nathan Sidwell <nathan@acm.org>
10329 Sandra Loosemore <sandra@codesourcery.com>
10331 * dumpfile.c (dump_open): Allow '-' for stdout.
10332 * doc/invoke.texi (Developer Options): Document dump filename
10333 determination early. Document stdin/stdout selection.
10335 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
10337 Microblaze Target: PIC data text relative
10339 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
10340 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
10342 * config/microblaze/microblaze.h (microblaze_constant_address_p):
10343 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
10344 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
10345 New addressing mode for data-text relative position indepenedent code.
10346 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
10347 'ADDRESS_SYMBOLIC_TXT_REL'.
10348 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
10349 (microblaze_legitimate_pic_operand): Exclude function calls from
10350 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
10351 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
10353 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10354 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10355 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
10356 for 'address + offset'.
10357 (microblaze_expand_prologue): Add new function prologue call for
10359 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
10360 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
10361 table in case of TARGET_PIC_DATA_TEXT_REL.
10362 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
10363 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
10364 Add new macros 'UNSPEC_TEXT',
10365 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
10366 + exclude function calls from 'UNSPEC_PLT' in case of data text
10368 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10369 new target hook for generating address diff vector tables in case of
10371 * doc/tm.texi : Regenerate.
10372 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
10373 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
10374 of addr diff vector generation.
10375 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10376 target hook definition.
10377 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
10378 Add default function for generate_pic_addr_diff_vec -> flag_pic.
10379 * doc/invoke.texi (Add new pic option): Add new microblaze pic
10380 option for data text relative.
10382 2018-04-30 Richard Biener <rguenther@suse.de>
10384 * tree-chrec.h (evolution_function_is_constant_p): Remove
10386 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
10388 2018-04-30 Richard Biener <rguenther@suse.de>
10391 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
10393 2018-04-30 Richard Biener <rguenther@suse.de>
10395 PR tree-optimization/28364
10396 PR tree-optimization/85275
10397 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
10398 copying first exit test.
10400 2018-04-28 Mark Wielaard <mark@klomp.org>
10402 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
10403 dwarf_version >= 5.
10404 (dwarf_AT): Handle DW_AT_addr_base.
10405 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
10407 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
10410 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
10411 (*ashl<dwi>3_doubleword_mask_1): Ditto.
10412 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
10413 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
10415 2018-04-28 Richard Biener <rguenther@suse.de>
10417 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
10418 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
10419 to reflect use. Only add interesting stmts.
10421 2018-04-27 Martin Jambor <mjambor@suse.cz>
10424 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
10425 the jump function allows for passing through aggregate values.
10427 2018-04-27 David Malcolm <dmalcolm@redhat.com>
10429 * input.h (in_system_header_at): Convert from macro to inline
10431 (from_macro_expansion_at): Likewise.
10432 (from_macro_definition_at): Likewise.
10434 2018-04-27 Jeff Law <law@redhat.com>
10436 * config.gcc: Mark tile* targets as deprecated/obsolete.
10438 2018-04-27 Richard Biener <rguenther@suse.de>
10440 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
10443 2018-04-27 Richard Biener <rguenther@suse.de>
10445 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
10447 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
10449 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
10450 with Yd constraint. Set "preferred_for_speed" attribute from
10451 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
10452 with Yd constraint.
10453 (*movdi_internal): Ditto.
10454 (movti_interunit splitters): Remove
10455 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
10456 (movdi_interunit splitters): Ditto.
10457 * config/i386/constraints.md (Ye): Remove.
10458 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
10460 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10463 * config/aarch64/constraints.md (Usg): Limit to 31.
10464 (Usj): Limit to 63.
10466 2018-04-27 Jakub Jelinek <jakub@redhat.com>
10468 PR tree-optimization/85529
10469 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
10470 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
10471 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
10472 zero extension or masking of the MSB bit.
10473 (optimize_range_tests): Add FIRST_BB argument, pass it through
10474 to optimize_range_tests_var_bound.
10475 (maybe_optimize_range_tests, reassociate_bb): Adjust
10476 optimize_range_tests callers.
10478 2018-04-26 Richard Biener <rguenther@suse.de>
10479 Jakub Jelinek <jakub@redhat.com>
10481 * cgraph.h (symbol_table): Just declare debug method here.
10482 * symtab.c (symbol_table::debug): Define.
10484 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
10486 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
10488 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
10490 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
10491 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
10492 (*movdi_internal): Substitute Yi and Yj constraint with x
10493 and Ym and Yn constraint with y constraint. Update "isa"
10494 attribute and set "preferred_for_speed" attribute from
10495 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
10496 (*movsi_internal): Ditto.
10497 (*movdf_internal): Ditto.
10498 (*movsf_internal): Ditto.
10499 (*zero_extendsidi2): Ditto.
10500 * config/i386/sse.md (vec_set<mode>_0): Ditto.
10501 (sse2_loadld): Ditto.
10502 (*vec_extract<ssevecmodelower>_0): Ditto.
10503 (*vec_extractv4si_0_zext_sse4): Ditto.
10504 (vec_concatv2di): Ditto.
10505 (*vec_dup<mode>): Ditto.
10506 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
10507 * config/i386/constraints.md (Yi): Remove.
10512 2018-04-26 Nathan Sidwell <nathan@acm.org>
10514 * dumpfile.c (dump_open): New.
10515 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
10516 (dump_finish): Detect stdio/stderr by value not name.
10518 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
10520 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
10522 2018-04-26 Tom de Vries <tom@codesourcery.com>
10525 * config/nvptx/nvptx.c (verify_neutering_jumps)
10526 (verify_neutering_labels): New function
10527 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
10529 2018-04-26 Tom de Vries <tom@codesourcery.com>
10532 * config/nvptx/nvptx.c (needs_neutering_p): New function.
10533 (nvptx_single): Use needs_neutering_p to skip over insns that do not
10536 2018-04-26 Richard Biener <rguenther@suse.de>
10537 Tom de Vries <tom@codesourcery.com>
10540 * lto-streamer-out.c (output_function): Fixup loops if required to match
10541 discovery done in the reader.
10543 2018-04-26 Richard Biener <rguenther@suse.de>
10545 PR tree-optimization/85116
10546 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
10547 have a loop exit from the single latch predecessor. Remove
10548 case of header with just condition.
10549 (ch_base::copy_headers): Exclude infinite loops from any
10551 (pass_ch::execute): Record exits.
10553 2018-04-26 Richard Biener <rguenther@suse.de>
10555 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
10556 prologue cost vector and pass it to vect_get_load_cost.
10557 (vect_get_peeling_costs_all_drs): Likewise.
10558 (vect_peeling_hash_get_lowest_cost): Likewise.
10559 (vect_enhance_data_refs_alignment): Likewise.
10561 2018-04-26 Richard Biener <rguenther@suse.de>
10563 PR middle-end/85450
10564 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
10565 checking of integer<->pointer conversions.
10566 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
10567 sign-/zero-extending pointer types.
10568 (expand_omp_for_static_chunk): Likewise.
10570 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
10571 Jean Lee <xiaoyur347@gmail.com>
10573 * config/mips/mips.c (mips_asan_shadow_offset): New function.
10574 (TARGET_ASAN_SHADOW_OFFSET): Define.
10575 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
10576 true for -fsanitize=address.
10578 2018-04-25 Mark Wielaard <mark@klomp.org>
10580 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
10583 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10585 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
10586 than "alu", remove explicit "memory" and "imm_disp" attributes.
10587 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
10589 PR middle-end/85414
10590 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
10591 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
10592 gen_lowpart_no_emit.
10594 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
10597 * config/i386/i386.c (ix86_expand_builtin): Change memory
10598 operand to XI, extend p0 to Pmode.
10599 * config/i386/i386.md: Change unspec volatile and operand
10600 1 mode to XI, change operand 0 mode to P.
10602 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10604 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
10605 GET_MODE_MASK before any checking.
10606 (nds32_can_use_bset_p): Likewise.
10607 (nds32_can_use_btgl_p): Likewise.
10609 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10611 * config/nds32/nds32-doubleword.md: New define_split pattern for
10612 illegal register number.
10614 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10616 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10618 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10620 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
10622 2018-04-25 Richard Biener <rguenther@suse.de>
10624 * lto-streamer.h (LTO_major_version): Bump to 8.
10626 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10628 * BASE-VER: Set to 9.0.0.
10630 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
10632 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
10633 in __abskf2 and __powikf2.
10635 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10638 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10639 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
10640 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
10641 Use the above on operand 2. Reindent.
10642 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
10644 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10647 * common/config/i386/i386-common.c (ix86_handle_option): Don't
10649 * config/i386/i386.opt (mcet): Removed.
10650 * doc/install.texi: Remove -mcet documentation.
10651 * doc/invoke.texi: Likewise.
10653 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10656 * doc/install.texi: Remove -mcet from bootstrap-cet.
10658 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10661 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
10662 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
10666 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
10667 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
10668 containing a CONST_VECTOR.
10670 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
10672 * doc/install.texi: Update newlib dependency for nvptx.
10674 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10677 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
10678 instead of INTVAL when shifting x left.
10680 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
10682 PR tree-optimization/85478
10683 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
10684 vect_grouped_store_supported for single element vectors.
10686 2018-04-24 Richard Biener <rguenther@suse.de>
10689 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
10690 load cost increase to the case of non-constant step.
10692 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10695 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
10696 destination if any_malformed_asm.
10698 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
10700 PR middle-end/85496
10701 * expr.c (store_field): In the bitfield case, if the value comes from
10702 a function call and is returned in registers by means of a PARALLEL,
10703 do not change the mode of the temporary unless BLKmode and VOIDmode.
10705 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
10707 PR rtl-optimization/85423
10708 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
10709 dependencies to debug insns when the previous insn is non-debug.
10711 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10713 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
10714 enums into a single definition.
10715 (fls): Fix predicates and printing.
10718 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10720 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
10721 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
10722 and short u6 immediate.
10723 (check_if_valid_sleep_operand): Remove.
10724 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
10726 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10728 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
10729 flag_always_save_lp condition.
10730 * config/nds32/nds32.opt (malways-save-lp): New option.
10732 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10734 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
10735 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
10736 * config/nds32/nds32.h
10737 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
10738 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
10740 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10742 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
10743 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
10745 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10746 Chung-Ju Wu <jasonwucj@gmail.com>
10748 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
10750 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
10751 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
10753 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10755 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
10757 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10759 * config/nds32/nds32-protos.h (nds32_data_alignment,
10760 nds32_local_alignment): Declare.
10761 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10762 nds32_local_alignment): New functions.
10763 (TARGET_CONSTANT_ALIGNMENT): Define.
10764 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
10766 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10768 * config/nds32/nds32.c
10769 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10770 (TARGET_MODES_TIEABLE_P): Likewise.
10772 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10774 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10775 level Ofast and Og.
10777 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10778 Chung-Ju Wu <jasonwucj@gmail.com>
10780 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10781 for unaligned access.
10782 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10784 * config/nds32/nds32-intrinsic.md: Likewise.
10785 * config/nds32/nds32_intrinsic.h: Likewise.
10786 * config/nds32/nds32.h (nds32_builtins): Likewise.
10787 * config/nds32/nds32.opt (munaligned-access): New option.
10788 * config/nds32/nds32.c (nds32_asm_file_start): Display
10789 flag_unaligned_access status.
10791 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10793 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10794 -mno-relax is present.
10795 * config/riscv/linux.h (LINK_SPEC): Ditto.
10797 2018-04-20 Martin Sebor <msebor@redhat.com>
10800 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10802 (gimple_fold_builtin_stxcpy_chk): Same.
10803 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10805 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10808 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10809 __powikf2 when long double is IEEE 128-bit.
10811 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10813 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10814 step to make sure stack always aligned.
10816 2018-04-20 Carl Love <cel@us.ibm.com>
10819 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10820 size check for arg0.
10822 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10823 Tom de Vries <tom@codesourcery.com>
10826 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10827 Emit insns for calls too.
10828 (nvptx_find_par): Always look for worker-level predecessor insn.
10829 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10831 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10832 (nvptx_process_pars): Propagate frames for calls.
10834 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10837 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10839 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10840 (ix86_handle_option): Don't handle OPT_mibt.
10841 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10843 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10845 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10846 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10847 (ix86_target_macros): Define __CET__ with flag_cf_protection
10848 for -fcf-protection.
10849 * config/i386/i386.c (isa2_opts): Remove -mibt.
10850 * config/i386/i386.h (TARGET_IBT): Removed.
10851 (TARGET_IBT_P): Likewise.
10852 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10853 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10854 * config/i386/i386.opt (mcet): Update help message.
10855 (mshstk): Likewise.
10857 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10858 -mcet as an alias for -mshstk.
10860 2018-04-20 Richard Biener <rguenther@suse.de>
10862 PR middle-end/85475
10863 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10864 complexity by forcing a single use of the multiply operand.
10866 2018-04-20 Martin Jambor <mjambor@suse.cz>
10869 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10870 recursion dependency to only apply to non-clones.
10872 2018-04-20 Martin Jambor <mjambor@suse.cz>
10875 * ipa-cp.c (create_specialized_node): Check that clones of
10876 self-recursive edges exist during IPA-CP.
10878 2018-04-19 Toon Moene <toon@moene.org>
10880 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10883 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10885 PR tree-optimization/85467
10886 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10887 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10888 VECTOR_CST element to type.
10890 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10893 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10894 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10895 (builtin_longjmp): Likewise.
10896 (save_stack_nonlocal): New pattern.
10897 (restore_stack_nonlocal): Likewise.
10899 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10902 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10903 Replace ASM_OUTPUT_LABEL with fprintf.
10905 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10908 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10909 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10910 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10911 define __IBT__ and __SHSTK__ for -fcf-protection.
10912 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10914 (ix86_trampoline_init): Likewise.
10915 (x86_output_mi_thunk): Likewise.
10916 (ix86_notrack_prefixed_insn_p): Likewise.
10917 (ix86_option_override_internal): Don't disallow -fcf-protection.
10918 * config/i386/i386.md (rdssp<mode>): Also enable for
10920 (incssp<mode>): Likewise.
10921 (nop_endbr): Likewise.
10922 * config/i386/i386.opt (mcet): Change help message to built-in
10925 (mshstk): Likewise.
10926 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10927 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10928 enable CET built-in functions.
10930 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10932 * common/config/i386/i386-common.c
10933 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10934 OPTION_MASK_ISA_MOVDIRI_UNSET,
10935 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10936 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10937 * config.gcc (movdirintrin.h): New header.
10938 * config/i386/cpuid.h (bit_MOVDIRI,
10939 bit_MOVDIR64B): New bits.
10940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10942 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10943 (VOID, PVOID, PCVOID)): New function types.
10944 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10945 __builtin_ia32_directstoreu_u64,
10946 __builtin_ia32_movdir64b): New builtins.
10947 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10948 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10950 (ix86_valid_target_attribute_inner_p): Ditto.
10951 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10952 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10953 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10954 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10955 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10956 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10957 (movdiri<mode>, movdir64b_<mode>): New.
10958 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10959 * config/i386/immintrin.h: Include movdirintrin.h.
10960 * config/i386/movdirintrin.h: New file.
10961 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10963 2018-04-19 Richard Biener <rguenther@suse.de>
10965 PR middle-end/85455
10966 * cfg.c (clear_bb_flags): When loop state says we have
10967 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10969 2018-04-19 Richard Biener <rguenther@suse.de>
10971 PR tree-optimization/84737
10972 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10973 copying restrict info.
10974 (vect_setup_realignment): Use it.
10975 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10976 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10977 the first DR to all generated stores.
10978 (vectorizable_load): Likewise for loads.
10980 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10982 PR tree-optimization/85446
10983 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10984 the integral and pointer types to have the same precision.
10986 * doc/install.texi: Document --disable-cet being the default and
10989 2018-04-18 Martin Liska <mliska@suse.cz>
10991 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10994 2018-04-18 Martin Liska <mliska@suse.cz>
10997 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11000 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11001 arguments if they are comparable.
11003 2018-04-18 Martin Liska <mliska@suse.cz>
11006 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11009 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11012 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
11015 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
11016 ENDBR after calling __morestack.
11018 2018-04-18 David Malcolm <dmalcolm@redhat.com>
11021 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
11022 by using gcc_base_ver to generate a gcc_driver_version, and use
11023 it when generating GCC_DRIVER_NAME.
11024 * configure: Regenerate.
11026 2018-04-18 Jakub Jelinek <jakub@redhat.com>
11029 * config.gcc: Obsolete powerpc*-*-*spe*.
11031 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11034 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
11035 (stabstr_D): Change type of unum from unsigned int to
11036 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
11039 2018-04-17 Jim Wilson <jimw@sifive.com>
11042 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
11043 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
11044 Set arg_pointer_offset after using pretend_args_size.
11046 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11048 PR rtl-optimization/85431
11049 * dse.c (record_store): Ignore zero width stores.
11052 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
11053 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
11054 __builtin_stack_restore rather than after it.
11055 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
11056 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
11057 argument instead of virtual_dynamic_stack_rtx.
11059 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
11061 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
11063 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
11064 Add note to error message to explain internal mapping of overloaded
11065 built-in function name to non-overloaded built-in function name.
11066 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
11069 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11072 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
11073 where the inputs overlap with the output.
11075 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11077 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
11078 (=v, v) alternative and explicit "memory" attribute.
11079 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
11080 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11082 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
11083 "sselog1" type instead of "sselog".
11084 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
11085 "sselog". Remove explicit "memory" attribute.
11086 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
11087 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11089 (vec_extract_hi_v32hi): Merge all alternatives into one, use
11090 "sselog1" type instead of "sselog". Remove explicit "memory"
11092 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
11093 use "sselog1" type instead of "sselog". Remove explicit "memory"
11095 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
11096 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11098 (vec_extract_hi_v64qi): Merge all alternatives into one, use
11099 "sselog1" type instead of "sselog". Remove explicit "memory"
11101 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
11102 use "sselog1" type instead of "sselog". Remove explicit "memory"
11106 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
11108 PR middle-end/85414
11109 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
11112 2018-04-17 Martin Jambor <mjambor@suse.cz>
11115 * ipa-cp.c (create_specialized_node): Call
11116 expand_all_artificial_thunks if necessary.
11118 2018-04-17 Martin Liska <mliska@suse.cz>
11121 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
11122 in message, remote space in between '_G' and '('.
11124 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11127 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
11128 avx512f_vmcmp<mode>3<round_saeonly_name>,
11129 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
11130 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
11131 avx512f_rndscale<mode><round_saeonly_name>,
11132 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
11133 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
11134 Use %<iptr>2 instead of %2 for -masm=intel.
11135 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
11136 avx512f_vcvttss2usi<round_saeonly_name>,
11137 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
11139 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
11140 avx512f_vcvttsd2usi<round_saeonly_name>,
11141 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
11142 Use %q1 instead of %1 for -masm=intel.
11143 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
11144 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
11145 of %3 for -masm=intel.
11146 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
11148 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
11150 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
11152 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
11153 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
11155 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
11157 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
11158 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
11159 %g1 and one with %0 and %1.
11160 (avx512er_vmrcp28<mode><round_saeonly_name>,
11161 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
11162 %1 for -masm=intel.
11163 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
11164 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
11165 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
11166 of %0 and %{%4%} for -masm=intel.
11167 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
11168 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
11169 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
11170 order of %0 and %{%5%}%{z%} for -masm=intel.
11172 2018-04-17 Jan Hubicka <jh@suse.cz>
11175 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
11177 2018-04-17 Martin Liska <mliska@suse.cz>
11180 * multiple_target.c (create_dispatcher_calls): Set apostrophes
11181 for target_clone error message. Make default implementation
11182 clone to be a local declaration.
11183 (separate_attrs): Add new argument and check for an empty
11185 (expand_target_clones): Handle it.
11186 (ipa_target_clone): Make redirection just for target_clones
11189 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
11190 Tom de Vries <tom@codesourcery.com>
11192 PR middle-end/84955
11193 * omp-expand.c (expand_oacc_for): Add dummy false branch for
11194 tiled basic blocks without omp continue statements.
11196 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
11199 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
11200 vec_extract expression as having side effects to make sure it gets
11203 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
11206 * config/i386/i386.c (get_builtin_code_for_version): Check
11209 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
11212 * config.gcc: Support "skylake".
11213 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11215 * config/i386/i386.c (m_SKYLAKE): Define.
11216 (processor_target_table): Add "skylake".
11217 (ix86_option_override_internal): Add "skylake".
11218 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
11219 PROCESSOR_CANNONLAKE.
11220 (get_builtin_code_for_version): Fix priority for
11221 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
11222 PROCESSOR_SKYLAKE-AVX512.
11223 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
11224 (processor_type): Add PROCESSOR_SKYLAKE.
11226 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
11227 Jason Merrill <jason@redhat.com>
11230 * convert.c (convert_to_integer_1): Use direct recursion for
11231 enumeral types and types with a precision less than the number
11232 of bits in their mode.
11234 2018-04-16 Julia Koval <julia.koval@intel.com>
11237 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
11238 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
11240 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
11243 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
11244 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
11245 and -mno-direct-move.
11247 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
11250 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
11251 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
11252 Ensure negative shifts result in {0}.
11254 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
11256 PR rtl-optimization/79916
11257 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
11258 regs (if any) to define how to gnerate SD moves when LRA is in
11261 2018-04-13 Jakub Jelinek <jakub@redhat.com>
11263 PR rtl-optimization/85393
11264 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
11265 * except.c (expand_dw2_landing_pad_for_region): Make static.
11266 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
11267 a label and unconditional jump to old_bb, rather than
11268 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
11271 PR rtl-optimization/85376
11272 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
11273 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
11274 instead of a specific value.
11276 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11277 Bin Cheng <bin.cheng@arm.com>
11279 PR tree-optimization/82965
11280 PR tree-optimization/83991
11281 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
11282 by_profile_only parameter.
11283 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
11284 information if the loop was predicted to iterate too many times.
11285 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
11287 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11290 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
11293 2018-04-13 Martin Liska <mliska@suse.cz>
11294 Jakub Jelinek <jakub@redhat.com>
11296 PR middle-end/81657
11297 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
11298 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
11299 * builtins.c (expand_builtin_memory_copy_args): Use
11300 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
11301 handle dest_addr == pc_rtx.
11303 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
11306 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
11307 asked to not generate direct moves.
11308 (fix_trunc<mode>si2_stfiwx): Similar.
11309 (fix_trunc<mode>si2_internal): Similar.
11311 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11314 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
11315 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
11316 lookup if dest in some wider mode is known to be const0_rtx and
11317 if so, record permanent equivalence for it to be ZERO_EXTEND of
11318 the narrower mode destination.
11320 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11322 * lto-streamer-out.c (output_function): Revert 259346.
11323 * omp-expand.c (expand_oacc_for): Likewise.
11325 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
11327 PR rtl-optimization/85354
11328 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
11329 * sel-sched.c (sel_global_init): ... here.
11331 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
11334 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
11335 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11336 mode for PE-COFF targets.
11337 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
11338 (i386_pe_asm_lto_end): Likewise.
11339 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
11340 (TARGET_ASM_LTO_END): Likewise.
11341 * config/i386/winnt.c (saved_debug_info_level): New static variable.
11342 (i386_pe_asm_lto_start): New function.
11343 (i386_pe_asm_lto_end): Likewise.
11345 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11346 Richard Biener <rguenther@suse.de>
11348 PR middle-end/84955
11349 * lto-streamer-out.c (output_function): Fix CFG loop state before
11351 * omp-expand.c (expand_oacc_for): Handle calls to internal
11352 functions like regular functions.
11354 2018-04-12 Richard Biener <rguenther@suse.de>
11357 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
11358 for the early LTO debug to properly generate references to it
11359 during DIE emission. Do not re-use that for the skeleton for
11361 (dwarf2out_early_finish): Likewise.
11363 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11366 * config/i386/sse.md
11367 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
11368 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
11369 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
11370 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
11371 and output is a reg, avoid creating invalid lowpart subreg, but
11372 instead split into a 512-bit move. Don't split if not AVX512VL,
11373 input is xmm16+ reg and output is a mem.
11374 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
11375 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
11376 xmm16+ reg and output is a mem.
11378 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11380 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
11381 also for flag_dwarf2_cfi_asm.
11383 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11385 PR rtl-optimization/85342
11386 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
11387 a bool scalar var inside of the loop instead. Don't try to update
11388 recog_data.operand after failed apply_change_group.
11390 2018-04-12 Tom de Vries <tom@codesourcery.com>
11393 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
11394 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
11395 array with flexible array member as array without given dimension.
11396 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
11397 argument for undefined param to true.
11399 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
11402 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11403 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
11404 from PowerPC section.
11405 * config/rs6000/sysv4.opt (mcall-): Improve help text.
11406 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
11407 help text that is too long.
11408 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
11409 help text that is too long.
11410 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
11411 help text that is too long.
11413 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
11415 * config/alpha/alpha.md (stack_probe_internal): Rename
11416 from "probe_stack". Update all callers.
11418 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11420 PR rtl-optimization/84566
11421 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
11422 sched_macro_fuse_insns.
11424 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11427 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
11428 (compute_block_dependences): ... from here.
11430 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11432 PR tree-optimization/85331
11433 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
11434 from int to HOST_WIDE_INT.
11436 2018-04-11 Martin Jambor <mjambor@suse.cz>
11439 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
11440 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
11441 not the same as the source val.
11442 (cgraph_edge_brings_value_p): New parameter.
11443 (gather_edges_for_value): Pass destination value to
11444 cgraph_edge_brings_value_p.
11445 (perhaps_add_new_callers): Likewise.
11446 (get_info_about_necessary_edges): Likewise and exclude values brought
11447 only by self-recursive edges.
11448 (create_specialized_node): Redirect only clones of self-calling edges.
11449 (+self_recursive_pass_through_p): New function.
11450 (find_more_scalar_values_for_callers_subset): Use it.
11451 (find_aggregate_values_for_callers_subset): Likewise.
11452 (known_aggs_to_agg_replacement_list): Removed.
11453 (decide_whether_version_node): Re-calculate known constants for all
11454 remaining context clones.
11456 2018-04-11 Richard Biener <rguenther@suse.de>
11459 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
11460 from early DWARF output.
11461 (dwarf2out_early_finish): Output line info unconditionally into
11462 early DWARF and add reference to it.
11464 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11467 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
11468 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
11469 other than V2DFmode using iptr mode attribute.
11470 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
11472 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11474 PR rtl-optimization/84659
11475 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
11477 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11480 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
11482 (output_loc_list): Pass address of a dummy size variable even in the
11483 locview handling loop.
11484 (index_location_lists): Add comment on why skip_loc_list_entry can't
11487 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
11490 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
11493 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
11496 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11497 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
11498 and -mstring-compare-inline-limit.
11500 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11503 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
11504 for stack clash protection in a register whenever we need it to be in
11507 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11509 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
11510 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
11512 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11515 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
11517 (mlong-double-): Ditto.
11518 * config/rs6000/sysv4.opt (msdata=): Ditto.
11519 (mtls-size=): Ditto.
11521 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11523 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11524 erroneous entries for
11525 "vector int vec_ldl (int, long int *)", and
11526 "vector unsigned int vec_ldl (int, unsigned long int *)".
11527 Add comments and entries for
11528 "vector bool char vec_ldl (int, bool char *)",
11529 "vector bool short vec_ldl (int, bool short *)",
11530 "vector bool int vec_ldl (int, bool int *)",
11531 "vector bool long long vec_ldl (int, bool long long *)",
11532 "vector pixel vec_ldl (int, pixel *)",
11533 "vector long long vec_ldl (int, long long *)",
11534 "vector unsigned long long vec_ldl (int, unsigned long long *)".
11535 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
11536 type tree bool_long_long_type_node and correct definition of
11537 bool_V2DI_type_node to make reference to this new type tree.
11538 (rs6000_mangle_type): Replace erroneous reference to
11539 bool_long_type_node with bool_long_long_type_node.
11540 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
11541 comments to emphasize sign distinctions for char and int types and
11542 replace RS6000_BTI_bool_long constant with
11543 RS6000_BTI_bool_long_long constant. Also add comment to restrict
11544 use of RS6000_BTI_pixel.
11545 (bool_long_type_node): Remove this macro definition.
11546 (bool_long_long_type_node): New macro definition
11548 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11550 PR rtl-optimization/85300
11551 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
11552 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
11553 simplify_unary_operation fails.
11555 2018-04-10 Martin Liska <mliska@suse.cz>
11557 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
11558 cgraph_edge and ipa_ref.
11560 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11564 * config/i386/sse.md
11565 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
11566 computation of the VEC_MERGE selector from mask.
11567 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
11568 Fix decoding of the VEC_MERGE selector into mask.
11570 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11572 PR tree-optimization/85286
11573 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
11575 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11577 * final.c (final_1): Set insn_last_address as well as
11578 insn_current_address.
11580 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11583 * explow.c (emit_stack_probe): Call validize_mem on memory location
11584 before passing it to gen_probe_stack. Create address operand and
11585 legitimize it for the probe_stack_address case.
11587 2018-04-09 Jan Hubicka <jh@suse.cz>
11590 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
11591 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
11592 * tree.c (free_lang_data_in_type): Fix handling of binfos;
11594 (free_lang_data): Rebuild type inheritance graph.
11596 2018-04-09 Martin Sebor <msebor@redhat.com>
11598 * invoke.texi (-finline-small-functions): Mention other optimization
11600 (-findirect-inlining, -fpartial-inlining): Same.
11601 (-finline-functions-called-once): Same.
11602 (-freorder-blocks-and-partition): Same.
11604 2018-04-09 Jan Hubicka <jh@suse.cz>
11607 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
11608 jumps; choose last target that matches the criteria (i.e.
11609 no partition changes for non-crossing jumps).
11610 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
11611 support for redirecting crossing jumps to non-crossing.
11613 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
11615 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
11616 also for naked functions.
11618 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
11620 * config/arc/arc.md (add_shift): New pattern.
11621 (add_shift2): Likewise.
11622 (sub_shift): Likewise.
11623 (sub_shift_cmp0_noout): Likewise.
11624 (compare_si_ashiftsi): Likewise.
11625 (xbfu_cmp0_noout): New combine pattern.
11626 (xbfu_cmp0"): Likewise.
11627 (movsi_set_cc_insn): Place the predicable variant first.
11628 (commutative_binary_cmp0_noout): Remove clobber.
11629 (commutative_binary_cmp0): New pattern.
11630 (noncommutative_binary_cmp0): Likewise.
11631 (noncommutative_binary_cmp0_noout): Likewise.
11632 (noncommutative_binary_comparison_result_used): Removed.
11633 (rsub_cmp0): New pattern.
11634 (rsub_cmp0_noout): Likewise.
11635 (extzvsi): Changed, keep only meaningful variants.
11636 (SQH, SEZ): New iterators.
11637 (SQH_postfix): New mode attribute.
11638 (SEZ_prefix): New code attribute.
11639 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
11640 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
11641 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
11642 of numerical value.
11643 (noncommutative_operator): Check the availability of barrel
11646 2018-04-09 Richard Biener <rguenther@suse.de>
11648 PR tree-optimization/85284
11649 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
11650 Only use the niter constraining form of simple_iv when the exit
11651 is always executed.
11653 2018-04-09 Tom de Vries <tom@codesourcery.com>
11656 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11657 (define_expand "*memory_barrier"): New define_expand.
11658 (define_insn "memory_barrier"): New insn.
11660 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11662 PR rtl-optimization/80463
11663 PR rtl-optimization/83972
11664 PR rtl-optimization/83480
11666 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
11667 correct producer for the insn.
11668 (tidy_control_flow): Fixup seqnos in case of debug insns.
11670 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11672 PR rtl-optimization/83913
11674 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
11675 different sched-times when merging exprs.
11677 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11679 PR rtl-optimization/83962
11681 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
11682 tidy_fallthru_edge and tidy_control_flow.
11684 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11686 PR rtl-optimization/83530
11688 * sel-sched.c (force_next_insn): New global variable.
11689 (remove_insn_for_debug): When force_next_insn is true, also leave only
11690 next insn in the ready list.
11691 (sel_sched_region): When the region wasn't scheduled, make another pass
11692 over it with force_next_insn set to 1.
11694 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11696 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
11698 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11699 for interrupt control.
11700 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11701 functions for interrupt control.
11702 * config/nds32/nds32-intrinsic.md: Likewise.
11703 * config/nds32/nds32_intrinsic.h: Likewise.
11704 * config/nds32/nds32.h (nds32_builtins): Likewise.
11706 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11708 * config/nds32/nds32.c (nds32_init_machine_status,
11709 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
11710 strict_aligned_p field.
11711 (nds32_expand_to_rtl_hook): New function.
11712 (TARGET_EXPAND_TO_RTL_HOOK): Define.
11713 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
11715 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11716 Chung-Ju Wu <jasonwucj@gmail.com>
11718 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
11719 * config/nds32/nds32-n7.md: New file.
11720 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
11721 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
11723 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
11724 * config/nds32/nds32.md (pipeline_model): Add n7.
11725 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
11726 * config/nds32/pipelines.md: Include n7 settings.
11728 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11729 Chung-Ju Wu <jasonwucj@gmail.com>
11731 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
11732 * config/nds32/nds32-e8.md: New file.
11733 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
11734 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
11736 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
11737 * config/nds32/nds32.md (pipeline_model): Add e8.
11738 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
11739 * config/nds32/pipelines.md: Include e8 settings.
11741 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11742 Chung-Ju Wu <jasonwucj@gmail.com>
11744 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
11745 * config/nds32/nds32-n8.md: New file.
11746 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
11747 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
11749 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
11750 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
11751 * config/nds32/nds32.md (pipeline_model): Add n8.
11752 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
11753 * config/nds32/pipelines.md: Include n8 settings.
11755 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11756 Chung-Ju Wu <jasonwucj@gmail.com>
11758 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
11759 * config/nds32/nds32-n9-2r1w.md: New file.
11760 * config/nds32/nds32-n9-3r2w.md: New file.
11761 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
11762 nds32_register_ports): New or modify for cpu n9.
11763 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
11765 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
11766 * config/nds32/nds32-utils.c: New file.
11767 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11768 TARGET_MUL_SLOW): Define.
11769 * config/nds32/nds32.md (pipeline_model): New attribute.
11770 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11771 New options that support cpu n9.
11772 * config/nds32/pipelines.md: Include n9 settings.
11773 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11775 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11777 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11778 information if necessary.
11779 (output_cond_branch_compare_zero): Likewise.
11780 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11781 (nds32_target_alignment): Refine for alignment.
11782 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11783 (FUNCTION_BOUNDARY): Modify.
11784 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11786 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11788 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11790 * config/nds32/constants.md (unspec_volatile_element): Add values for
11791 TLB operation and data prefetch.
11792 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11793 functions for TLB operation and data prefetch.
11794 * config/nds32/nds32-intrinsic.md: Likewise.
11795 * config/nds32/nds32_intrinsic.h: Likewise.
11796 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11797 (nds32_print_operand): Likewise.
11798 * config/nds32/nds32.h (nds32_builtins): Likewise.
11800 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11801 Andrew Pinski <pinsika@gcc.gnu.org>
11803 PR middle-end/82976
11804 * match.pd: Use constant_boolean_node of correct type instead of
11805 boolean_true_node or boolean_false_node for simplifying
11806 pointer comparisons to zero.
11808 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11810 PR tree-optimization/80021
11811 * tree.c (verify_type_variant): Make error call in verify_variant_match
11812 translatable and remove final full stop.
11814 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11816 * config/nds32/constants.md (unspec_volatile_element): Add
11817 UNSPEC_VOLATILE_EH_RETURN.
11818 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11819 nds32_output_stack_pop): Support dwarf exception handling process.
11820 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11821 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11822 exception handling process.
11823 (nds32_compute_stack_frame): Likewise.
11824 (nds32_return_addr_rtx): Likewise.
11825 (nds32_initial_elimination_offset): Likewise.
11826 (nds32_expand_prologue): Likewise.
11827 (nds32_expand_epilogue): Likewise.
11828 (nds32_dynamic_chain_address): New function.
11829 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11830 exception handling.
11831 (DYNAMIC_CHAIN_ADDRESS): Define.
11832 (EH_RETURN_DATA_REGNO): Define.
11833 (EH_RETURN_STACKADJ_RTX): Define.
11834 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11835 patterns for dwarf exception handling.
11837 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11839 * config/nds32/nds32.h: Clean up obsolete macros.
11841 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11843 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11844 Add enum values for particular instructions.
11845 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11846 particular intrinsic functions.
11847 * config/nds32/nds32-intrinsic.md: Likewise.
11848 * config/nds32/nds32_intrinsic.h: Likewise.
11849 * config/nds32/nds32.h (nds32_builtins): Likewise.
11850 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11851 (btst, ave): New patterns for particular instructions.
11853 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11855 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11856 Add enum values for atomic load/store and memory sync.
11857 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11859 * config/nds32/nds32-intrinsic.md: Likewise.
11860 * config/nds32/nds32_intrinsic.h: Likewise.
11861 * config/nds32/nds32.h (nds32_builtins): Likewise.
11863 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11865 PR tree-optimization/85257
11866 * fold-const.c (native_encode_vector): If not all elts could fit
11867 and off is -1, return 0 rather than offset.
11868 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11869 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11870 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11871 adjust buffer in native_interpret_expr call.
11873 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11875 * config/nds32/constants.md (unspec_volatile_element): Add cache
11876 control enum values.
11877 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11878 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11879 * config/nds32/nds32.c (nds32_cctl_names): New.
11880 (nds32_print_operand): Handle cache control register names.
11881 * config/nds32/nds32.h (nds32_builtins): New enum values.
11882 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11884 * config/nds32/nds32.md (type): Add mmu.
11885 * config/nds32/pipelines.md (simple_insn): Add mmu.
11887 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11889 * config/nds32/nds32.md (type): Remove call.
11890 * config/nds32/pipelines.md (simple_insn): Likewise.
11892 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11894 * config/nds32/constants.md (unspec_volatile_element): Add
11895 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11896 UNSPEC_VOLATILE_FMFCFG.
11897 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11898 description for fmfcfg and fmfcsr.
11899 (bdesc_1arg): Add fmtcsr.
11900 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11901 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11902 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11903 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11904 unspec_fmfcfg): New patterns.
11905 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11906 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11907 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11908 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11909 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11910 __nds32__fmfcfg): Define.
11912 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11914 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11915 intrinsic register names.
11916 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11917 intrinsic register enum values and macros.
11919 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11921 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11922 for load/store addressing form.
11923 (nds32_print_operand_address): Likewise.
11925 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11928 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11929 based on LABEL_REF. Remove useless assertion.
11930 (pic_address_needs_scratch): Fix formatting.
11931 (sparc_legitimize_pic_address): Minor tweaks.
11932 (sparc_delegitimize_address): Adjust assertion accordingly.
11933 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11934 into symbolic_operand.
11935 (movsi_high_pic_label_ref): Likewise.
11936 (movsi_lo_sum_pic_label_ref): Likewise.
11937 (movdi_pic_label_ref): Likewise.
11938 (movdi_high_pic_label_ref): Likewise.
11939 (movdi_lo_sum_pic_label_ref): Likewise.
11941 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11943 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11944 custom LIB_SPEC setup.
11946 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11947 Kito Cheng <kito.cheng@gmail.com>
11949 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11950 * config/riscv/freebsd.h: New.
11952 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11954 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11955 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11958 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11959 Kito Cheng <kito.cheng@gmail.com>
11961 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11962 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11963 * config/nds32/nds32-protos.h (nds32_output_call,
11964 nds32_output_return): Declare.
11965 * config/nds32/nds32.md: Refine all the call and return patterns.
11967 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11970 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11971 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11973 PR rtl-optimization/84872
11974 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11975 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11976 EDGE_CROSSING edge.
11978 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11980 * expr.c (copy_blkmode_to_reg): Revert 254862.
11981 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11983 2018-04-06 Richard Biener <rguenther@suse.de>
11985 PR middle-end/85244
11986 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11987 after seeing a component reference with an adjacent field. Treat
11988 refs to arrays at struct end of external decls similar to
11989 refs to unconstrained commons.
11991 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11994 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11995 look through SAVE_EXPRs with non-side-effects argument. Adjust
11997 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
12000 2018-04-06 Richard Biener <rguenther@suse.de>
12002 PR middle-end/85180
12003 * alias.c (find_base_term): New wrapper around find_base_term
12004 unwinding CSELIB_VAL_PTR changes.
12005 (find_base_term): Do not restore CSELIB_VAL_PTR during the
12008 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12010 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
12012 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
12013 constant definitions.
12014 ("nop"): lr 0,0 -> nopr r0
12015 ("nop_lr0", "nop_lr1"): New insn definitions.
12017 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12019 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
12020 NDS32_V3PUSH_AVAILABLE_P macro.
12022 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
12023 Chung-Ju Wu <jasonwucj@gmail.com>
12025 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
12026 (nds32*-*-*): Add float and fpu_config into supported_defaults.
12027 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
12028 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
12029 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
12030 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
12031 * config/nds32/constraints.md: New constraints and checking for hard
12032 float configuration.
12033 * config/nds32/iterators.md: New mode iterator and attribute for hard
12034 float configuration.
12035 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
12037 * config/nds32/nds32-fpu.md: New file.
12038 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
12039 deal with hard float code generation.
12040 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
12042 (abi_type, float_reg_number): New enum type.
12043 * config/nds32/nds32-predicates.c: New predicates for hard float.
12044 * config/nds32/nds32-protos.h: Declare functions for hard float.
12045 * config/nds32/nds32.c: Implementation for hard float configuration.
12046 * config/nds32/nds32.h: Definitions for hard float configuration.
12047 * config/nds32/nds32.md: Include hard float machine description and
12048 modify patterns for hard float configuration.
12049 * config/nds32/nds32.opt: New options for hard float configuration.
12050 * config/nds32/predicates.md: New predicates for hard float
12053 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12055 * common/config/nds32/nds32-common.c
12056 (nds32_option_optimization_table): Enable -mreleax-hint by default.
12058 2018-04-05 Jakub Jelinek <jakub@redhat.com>
12060 PR middle-end/85195
12061 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
12062 CONSTRUCTOR_ELT (ctor, ...)->value.
12064 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
12067 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
12069 2018-04-05 Tom de Vries <tom@codesourcery.com>
12072 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
12075 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12076 Kito Cheng <kito.cheng@gmail.com>
12078 * config/nds32/constraints.md (U33): Fine-tune checking condition.
12079 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
12080 * config/nds32/nds32.h (nds32_16bit_address_type): Add
12081 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
12083 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12084 Kito Cheng <kito.cheng@gmail.com>
12086 * config/nds32/constraints.md (Ufe): New memory constraint.
12087 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
12088 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
12089 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
12091 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
12092 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
12094 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12096 * config/nds32/nds32.md: Use optimize_size in the condition for
12097 alu-shift instructions.
12099 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12101 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
12103 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12105 * config/nds32/nds32.md (negsi2): Refine pattern.
12107 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12108 Chung-Ju Wu <jasonwucj@gmail.com>
12110 * config/nds32/iterators.md (shift_rotate): New code iterator.
12111 (shift): New code attribute.
12112 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
12113 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
12114 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
12115 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
12116 bit-wise operations.
12117 (andsi3, *andsi3): Ditto.
12118 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
12119 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
12120 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
12121 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
12122 nds32_ior_operand, nds32_xor_operand): New predicates.
12124 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12126 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
12127 (addsi3, subsi3): ... this.
12129 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12131 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
12133 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12135 * config/nds32/nds32.md: Adjust indention.
12137 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12139 * config/nds32/nds32.md (feature): New attribute.
12141 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12143 * config/nds32/nds32.md (subtype): New attribute.
12145 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
12148 * config/arm/arm-builtins.c (arm_expand_builtin): Change
12149 expansion to perform a bitwise AND of the argument followed by a
12150 boolean negation of the result.
12152 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
12154 PR rtl-optimization/84878
12155 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
12156 the basic block. Assert the use reference is not artificial and that
12157 it has an associated insn.
12159 2018-04-04 Michael Matz <matz@suse.de>
12161 * builtins.c (compute_objsize): Pass correct operand
12162 to array_at_struct_end_p.
12164 2018-04-04 Richard Biener <rguenther@suse.de>
12167 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
12168 from contexts for DINFO_LEVEL_TERSE and below.
12170 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12172 * config/nds32/nds32-doubleword.md (move_<mode>): Require
12173 resiter_operand condition.
12174 * config/nds32/nds32.md (*move<mode>): Ditto.
12176 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12177 Monk Chiang <sh.chiang04@gmail.com>
12179 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
12181 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12183 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
12185 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12186 Kito Cheng <kito.cheng@gmail.com>
12188 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
12189 nds32_cond_code_str, output_cond_branch,
12190 output_cond_branch_compare_zero, nds32_expand_cbranch,
12191 nds32_expand_cstore, nds32_expand_movcc,
12192 nds32_output_cbranchsi4_equality_zero,
12193 nds32_output_cbranchsi4_equality_reg,
12194 nds32_output_cbranchsi4_equality_reg_or_const_int,
12195 nds32_output_cbranchsi4_greater_less_zero: New functions.
12196 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
12197 nds32_expand_cstore, nds32_expand_movcc,
12198 nds32_output_cbranchsi4_equality_zero,
12199 nds32_output_cbranchsi4_equality_reg,
12200 nds32_output_cbranchsi4_equality_reg_or_const_int,
12201 nds32_output_cbranchsi4_greater_less_zero): Declare.
12202 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
12203 nds32_rimm11s_operand): New predicates.
12204 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
12205 * config/nds32/nds32.md: Rewrite all the branch and conditional move
12208 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12210 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
12211 * config/nds32/nds32.md: Ditto.
12212 * config/nds32/pipelines.md: Ditto.
12214 2018-04-04 Richard Biener <rguenther@suse.de>
12216 PR tree-optimization/85168
12217 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
12218 propagating abnormals.
12220 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12222 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
12224 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12225 Kito Cheng <kito.cheng@gmail.com>
12227 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
12228 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
12229 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
12230 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
12231 * config/nds32/nds32.md (sibcall_internal): New.
12232 (sibcall_register): Remove.
12233 (sibcall_immediate): Remove.
12234 (sibcall_value_internal): New.
12235 (sibcall_value_register): Remove.
12236 (sibcall_value_immediate): Remove.
12237 * config/nds32/predicates.md (nds32_general_register_operand): New.
12238 (nds32_call_address_operand): New.
12240 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12242 PR rtl-optimization/85167
12243 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
12244 bb_defs if *split_p, instead preinitialize it to NULL.
12246 PR tree-optimization/85156
12247 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
12248 evaluating the argument multiple times.
12250 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
12252 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
12254 (_mm_cvtpd_ps): Likewise.
12255 (_mm_cvttpd_epi32): Likewise.
12256 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
12257 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
12258 vector, pixel, and bool following altivec.h include.
12260 2018-04-03 Martin Sebor <msebor@redhat.com>
12262 * doc/extend.texi (Common Function Attributes): Clarify.
12263 (const attribute): Likewise.
12264 (pure attribute): Likewise.
12266 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12269 * config/i386/i386.c (ix86_expand_vector_set): Use
12270 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
12272 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
12274 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
12275 instructions when changing rounding bits to preserve precision bits
12276 in the x87 control word.
12278 2018-04-03 Martin Liska <mliska@suse.cz>
12280 PR tree-optimization/82491
12281 * rtl.h (strip_offset_and_add): Replace += suboffset with
12282 poly_uint64 () + suboffset.
12284 2018-03-29 Martin Liska <mliska@suse.cz>
12285 Martin Jambor <mjambor@suse.cz>
12288 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
12289 param_type is not an integral or pointer type.
12291 2018-04-03 Richard Biener <rguenther@suse.de>
12293 * sese.h (recompute_all_dominators): Remove.
12295 2018-04-02 Martin Sebor <msebor@redhat.com>
12297 * doc/invoke.texi (-Wrestrict): Fix typos.
12299 2018-04-02 Jim Wilson <jimw@sifive.com>
12301 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
12302 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
12303 (<optab>di3, <optab>si3_extend): Likewise.
12304 (<optab>si3_mask, <optab>si3_mask_1): New.
12305 (<optab>di3_mask, <optab>di3_mask_1): New.
12306 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
12307 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
12308 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
12310 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
12312 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
12315 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
12317 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
12318 (nds32_canonicalize_comparison): New function.
12320 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12321 Kito Cheng <kito.cheng@gmail.com>
12322 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12324 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
12325 * config/nds32/constants.md (unspec_volatile_element): Add
12326 UNSPEC_VOLATILE_RELAX_GROUP.
12327 * config/nds32/nds32-relax-opt.c: New file.
12328 * config/nds32/nds32-predicates.c
12329 (nds32_symbol_load_store_p): New function.
12330 * config/nds32/nds32-protos.h
12331 (nds32_symbol_load_store_p): Declare function.
12332 (make_pass_nds32_relax_opt): Declare new rtl pass function.
12333 * config/nds32/nds32.c
12334 (nds32_register_pass): New function to register pass.
12335 (nds32_register_passes): New function to register passes.
12336 * config/nds32/nds32.md (relax_group): New pattern.
12337 * config/nds32/nds32.opt (mrelax-hint): New option.
12338 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
12340 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
12342 * config/nds32/t-nds32: Modify files dependency.
12344 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12346 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
12347 (PROFILE_HOOK): Define its implementation.
12349 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12351 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
12352 type and 32-bit size.
12354 2018-04-01 Jakub Jelinek <jakub@redhat.com>
12356 PR middle-end/85090
12357 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
12358 (V_128_256): New mode iterator.
12359 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
12360 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
12361 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
12363 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
12364 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
12366 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
12369 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
12370 NaN inputs correctly.
12372 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
12375 * config/rs6000/vsx.md (??r): New mode attribute.
12376 (*vsx_mov<mode>_64bit): Use it.
12377 (*vsx_mov<mode>_32bit): Likewise.
12379 2018-03-30 Martin Sebor <msebor@redhat.com>
12381 PR tree-optimization/84818
12382 * builtins.c (check_access): Use warning_n.
12384 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12387 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
12389 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
12392 2018-03-30 Julia Koval <julia.koval@intel.com>
12395 * x86-tune.def (movx, partial_reg_dependency): Enable for
12398 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
12400 PR inline-asm/84985
12401 * lra-constraints.c (process_alt_operands): Move setting
12402 this_alternative_matches below.
12404 2018-03-29 Martin Liska <mliska@suse.cz>
12407 * doc/invoke.texi: Document how LTO works with debug info.
12408 Describe auto-load support of binutils. Mention 'x86-64'
12409 as valid option value of -march option.
12411 2018-03-29 Jakub Jelinek <jakub@redhat.com>
12413 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
12416 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
12417 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
12418 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
12421 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12424 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
12425 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
12426 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
12427 for RS6000_BTM_POWERPC64.
12428 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
12429 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
12430 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
12433 (DIVDEU): Likewise.
12435 2018-03-28 Carl Love <cel@us.ibm.com>
12438 2017-09-27 Carl Love <cel@us.ibm.com>
12440 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
12441 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
12442 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
12445 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12447 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
12448 instead of __vector bool.
12449 (_mm_max_pu8): Likewise.
12450 (_mm_min_pi16): Likewise.
12452 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12455 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
12456 (DIVWEUO): Likewise.
12457 (DIVDEO): Likewise.
12458 (DIVDEUO): Likewise.
12459 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
12460 DIVWEUO and DIVDEUO.
12461 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
12462 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
12463 (div_extend): Likewise.
12464 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
12466 (__builtin_divweuo): Likewise.
12467 (__builtin_divdeo): Likewise.
12468 (__builtin_divdeuo): Likewise.
12470 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12473 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
12474 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
12476 PR tree-optimization/82004
12477 * gimple-match-head.c (optimize_pow_to_exp): New function.
12478 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
12479 Don't fold to exp if optimize_pow_to_exp is false.
12481 2018-03-28 Martin Liska <mliska@suse.cz>
12484 * calls.c (initialize_argument_information): Fix trailing space.
12485 * common.opt: Fix typo and provide better explanation for
12486 -fsanitize-coverage option.
12487 * config/i386/i386.opt: Fix typo.
12489 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12490 Martin Liska <mliska@suse.cz>
12493 * gimplify.c (asan_poison_variable): Don't do the check for
12494 gimplify_omp_ctxp here.
12495 (gimplify_decl_expr): Do it here.
12496 (gimplify_target_expr): Likewise.
12498 2018-03-28 Martin Liska <mliska@suse.cz>
12501 * config/i386/i386.c (ix86_function_arg_advance): Do not call
12502 chkp_type_bounds_count if MPX is not enabled.
12504 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
12506 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
12508 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
12511 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
12512 function to create the function decl for complex long double
12513 multiply and divide for -mabi=ieeelongdouble.
12514 (init_float128_ieee): Call it.
12516 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
12519 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
12520 -fcf-protection=branch -mibt.
12521 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
12523 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12526 * config/arm/arm.c (arm_valid_symbolic_address): Handle
12527 arm_word_relocations.
12529 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
12532 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
12533 extern array declarations.
12535 2018-03-27 Richard Biener <rguenther@suse.de>
12537 PR middle-end/84067
12538 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
12539 explicit single_use checks.
12541 2018-03-27 Richard Biener <rguenther@suse.de>
12543 PR tree-optimization/85082
12544 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
12547 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12549 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
12550 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
12551 Turn on fasynchronous-unwind-tables and funwind-tables.
12553 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
12556 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
12557 (*bmi_blsr_<mode>_ccz): Ditto.
12559 2018-03-26 Tom de Vries <tom@codesourcery.com>
12561 PR tree-optimization/85063
12562 * omp-general.c (offloading_function_p): New function. Factor out
12564 * omp-offload.c (pass_omp_target_link::gate): ... here.
12565 * omp-general.h (offloading_function_p): Declare.
12566 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
12567 with attribute omp declare target for offloading functions.
12569 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
12571 PR tree-optimization/84005
12572 * tree-data-ref.h (get_base_for_alignment): Declare.
12573 * tree-data-ref.c (get_base_for_alignment_1): New function.
12574 (get_base_for_alignment): Likewise.
12575 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
12576 get_base_for_alignment to find a suitable base object, instead
12577 of always using drb->base_address.
12579 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12581 PR inline-asm/85022
12582 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
12583 known size by default.
12585 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
12587 PR inline-asm/85030
12588 * lra-constraints.c (process_alt_operands): Don't match BLKmode
12589 and non BLKmode operands.
12591 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12594 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
12595 Clean up attributes.
12597 2018-03-23 Richard Biener <rguenther@suse.de>
12600 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
12601 we are going to emit early debug for LTO.
12603 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12605 PR inline-asm/85034
12606 * function.c (match_asm_constraints_1): Don't optimize if input
12607 doesn't satisfy general_operand predicate for output's mode.
12609 PR inline-asm/85022
12610 * alias.c (write_dependence_p): Don't require for x_canonicalized
12611 non-VOIDmode if x has VOIDmode.
12614 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
12615 just don't try to optimize it rather than assert it never happens.
12617 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12619 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
12620 macro expansions for definition of ST_INTERNAL_<mode> and
12621 LD_INTERNAL_<mode> builtins.
12622 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
12624 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
12626 (altivec_expand_st_builtin): Likewise.
12627 (altivec_expand_builtin): Remove calls to deleted functions.
12628 (rs6000_address_for_altivec): Delete this function.
12629 * config/rs6000/vector.md: Remove expands for
12630 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
12632 2018-03-22 Sudakshina Das <sudi.das@arm.com>
12635 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
12636 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
12637 re-computing once computed.
12638 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
12639 (arm_init_machine_status): Initialize
12640 machine->static_chain_stack_bytes.
12642 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12645 * doc/extend.texi: Add four new prototypes for vec_ld.
12646 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
12647 definitions for more logical presentation.
12648 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
12649 entries for V1TI variants of __builtin_altivec_ld builtin.
12650 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
12651 handling of V1TI variant of LVX icode pattern.
12652 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
12653 (rs6000_gimple_fold_builtin): Likewise.
12654 (altivec_init_builtins): Add code to define
12655 __builtin_altivec_lvx_v1ti function.
12657 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12659 PR inline-asm/84941
12660 * function.c (match_asm_constraints_1): Don't do the optimization
12661 if input isn't a REG, SUBREG, MEM or constant.
12663 2018-03-22 Tom de Vries <tom@codesourcery.com>
12665 PR tree-optimization/84956
12666 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
12667 bb_has_abnormal_pred.
12669 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12672 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
12673 DECL_INITIAL (decl) to decl at the end.
12674 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
12675 adjust the comment.
12677 2018-03-21 Joseph Myers <joseph@codesourcery.com>
12679 * doc/extend.texi (__builtin_tgmath): Document when complex
12680 integer types are treated as _Complex _Float64.
12682 2018-03-21 Tom de Vries <tom@codesourcery.com>
12684 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
12686 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12688 PR tree-optimization/84960
12689 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
12690 if it is ENTRY block, move them into single succ of ENTRY in that case.
12692 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
12694 PR tree-optimization/84811
12695 * poly-int.h (poly_span_traits): Remove the T3 parameter and
12696 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12697 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
12698 (known_subrange_p): Update accordingly. Cast each value involved
12699 in the size comparison, rather than casting the result of the
12702 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12704 PR tree-optimization/84982
12705 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
12706 by flipping the least significant bit rather than all bits from
12707 bitpos to bitpos + bitsize - 1.
12709 2018-03-21 Nathan Sidwell <nathan@acm.org>
12711 * doc/extend.texi (Deprecated Features): Remove mention of
12712 long-deleted deprecations.
12714 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12717 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
12718 * configure: Regenerate.
12720 2018-03-21 Tom de Vries <tom@codesourcery.com>
12722 PR tree-optimization/83126
12723 * tree-parloops.c (num_phis): New function.
12724 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
12726 2018-03-21 Nathan Sidwell <nathan@acm.org>
12728 * doc/extend.texi (Deprecated Features): Update deprecated flags,
12729 mention anon-struct/union members and trailing attributes.
12731 2018-03-21 Bin Cheng <bin.cheng@arm.com>
12733 PR tree-optimization/84969
12734 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
12735 builtin memset partitions if they set different rhs values.
12737 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12739 PR rtl-optimization/84989
12740 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
12741 VEC_DUPLICATE with scalar result mode.
12743 2018-03-21 Martin Liska <mliska@suse.cz>
12746 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
12747 not intended return statement.
12749 2018-03-21 Martin Liska <mliska@suse.cz>
12752 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
12753 (chkp_find_bound_slots_1): Limit number of iterations.
12755 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
12758 * Minor grammar fixes for x86 options.
12760 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12763 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
12764 holding REG_CFA_RESTORE notes, instead turn them into a USE.
12766 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12769 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12770 (altivec_lvx_<mode>_1op): Likewise.
12771 (altivec_stvx_<mode>_2op): Likewise.
12772 (altivec_stvx_<mode>_1op): Likewise.
12773 (altivec_lvx_<VM2:mode>): New define_expand.
12774 (altivec_stvx_<VM2:mode>): Likewise.
12775 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12776 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12777 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12778 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12779 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12780 (rs6000_gen_lvx): Likewise.
12781 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12782 (altivec_expand_stv_builtin): Likewise.
12783 (altivec_expand_builtin): Likewise.
12784 * config/rs6000/vector.md: Likewise.
12786 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12789 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12792 2018-03-20 Richard Biener <rguenther@suse.de>
12795 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12796 sign-conversions as zero, fall back to standard scalar_stmt
12799 2018-03-20 Martin Liska <mliska@suse.cz>
12802 * predict.c (rebuild_frequencies): Handle case when we have
12803 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12805 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12808 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12809 flag_section_anchors.
12810 * varasm.c (use_blocks_for_decl_p): Remove hack for
12811 dw2_force_const_mem.
12814 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12816 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12817 be created, use lowpart_subreg of operands[0] rather than operands[0]
12819 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12820 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12821 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12822 and n constraint instead of aarch64_shift_imm_di and Usd.
12823 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12824 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12826 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12829 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12830 to favor GPR over NEON registers.
12831 (<shift>di3_neon): Likewise.
12833 2018-03-20 Tom de Vries <tom@codesourcery.com>
12836 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12837 (nvptx_process_pars): Emit bar.sync asap and alap.
12839 2018-03-20 Tom de Vries <tom@codesourcery.com>
12842 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12843 seen_label if seen_label is already set.
12845 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12848 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12849 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12850 Use 1U instead of 1. Formatting fixes.
12853 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12854 instead of TREE_TYPE (s1) for the return value.
12856 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12858 PR tree-optimization/84946
12859 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12860 bitsize + bitsize in poly_uint64 rather than poly_int64.
12863 * dwarf2asm.c: Include fold-const.c.
12864 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12865 of decl rather than decl itself.
12867 PR rtl-optimization/84643
12868 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12870 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12873 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12874 calling assemble_variable.
12876 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12879 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12880 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12882 2018-03-19 Jim Wilson <jimw@sifive.com>
12885 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12886 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12887 (riscv_first_stack_step): Likewise.
12888 (riscv_option_override): Use STACK_BOUNDARY instead of
12889 MIN_STACK_BOUNDARY.
12890 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12891 MIN_STACK_BOUNDARY.
12892 (BIGGEST_ALIGNMENT): Set to 128.
12893 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12894 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12897 2018-03-19 Richard Biener <rguenther@suse.de>
12899 PR tree-optimization/84933
12900 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12901 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12903 2018-03-19 Richard Biener <rguenther@suse.de>
12905 PR tree-optimization/84859
12906 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12907 (cond_if_else_store_replacement): Perform sinking operation on
12908 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12909 Generalize what a BB with a single eligible store is.
12911 2018-03-19 Richard Biener <rguenther@suse.de>
12913 PR tree-optimization/84929
12914 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12915 chrec_is_positive against non-chrec arg.
12917 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12920 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12922 2018-03-18 Martin Liska <mliska@suse.cz>
12924 PR rtl-optimization/84635
12925 * regrename.c (build_def_use): Use matches_mode only when
12928 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12930 PR tree-optimization/84913
12931 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12932 vectorize chains of COND_EXPRs.
12934 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12936 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12938 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12940 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12942 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12944 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12946 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12947 Kito Cheng <kito.cheng@gmail.com>
12949 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12950 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12951 (nds32_adjust_reg_alloc_order): New function.
12952 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12954 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12956 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12957 nds32_print_operand, nds32_print_operand_address): Use
12958 HOST_WIDE_INT_PRINT_DEC instead.
12960 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12962 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12964 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12967 * config/i386/i386.c (initial_ix86_tune_features,
12968 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12969 unsigned long long.
12970 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12971 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12972 rather than 1u << ix86_tune. Formatting fix.
12973 (ix86_option_override_internal): Change ix86_arch_mask from
12974 unsigned int to unsigned HOST_WIDE_INT, initialize to
12975 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12976 (ix86_function_specific_restore): Likewise.
12978 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12981 * postreload.c (reload_combine_recognize_pattern): Perform
12982 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12983 truncate_int_for_mode the result for the destination's mode.
12986 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12987 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12989 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12992 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12995 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12996 regno_allocno_class_array and sorted_pseudos.
12997 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12998 insns where regno is used.
13000 2018-03-16 Martin Liska <mliska@suse.cz>
13003 * multiple_target.c (create_dispatcher_calls): Redirect
13004 reference in the symbol table.
13006 2018-03-16 Martin Liska <mliska@suse.cz>
13009 * multiple_target.c (create_dispatcher_calls): Redirect also
13012 2018-03-16 Jakub Jelinek <jakub@redhat.com>
13016 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
13017 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
13018 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
13020 2018-03-16 Julia Koval <julia.koval@intel.com>
13022 * doc/invoke.texi (Skylake Server): Add CLWB.
13023 Cannonlake): Remove CLWB.
13025 2018-03-16 Jakub Jelinek <jakub@redhat.com>
13027 PR tree-optimization/84841
13028 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
13030 (FLOAT_ONE_CONST_TYPE): Define.
13031 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
13032 (sort_by_operand_rank): Put entries with higher constant_type last
13033 rather than first to match comments.
13035 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
13037 * config/nios2/nios2.md (movsi_internal): Fix thinko in
13040 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13043 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
13044 check and use address of target always.
13046 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
13049 * config/i386/i386.c (indirect_thunk_needed): Update comments.
13050 (indirect_thunk_bnd_needed): Likewise.
13051 (indirect_thunks_used): Likewise.
13052 (indirect_thunks_bnd_used): Likewise.
13053 (indirect_return_needed): New.
13054 (indirect_return_bnd_needed): Likewise.
13055 (output_indirect_thunk_function): Add a bool argument for
13057 (output_indirect_thunk_function): Don't generate alias for
13058 function return thunk.
13059 (ix86_code_end): Call output_indirect_thunk_function to generate
13060 function return thunks.
13061 (ix86_output_function_return): Set indirect_return_bnd_needed
13062 and indirect_return_needed instead of indirect_thunk_bnd_needed
13063 and indirect_thunk_needed.
13065 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
13067 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
13068 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
13069 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
13071 2018-03-15 David Malcolm <dmalcolm@redhat.com>
13072 Paul Hua <paul.hua.gm@gmail.com>
13075 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
13077 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
13079 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
13080 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
13081 resp. SFmode cases.
13083 2018-03-15 Tamar Christina <tamar.christina@arm.com>
13086 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13087 instead of GET_MODE_SIZE when comparing Units.
13089 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13092 * varasm.c (hash_section): Return an unchangeble hash value
13093 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
13094 Return !aarch64_can_use_per_function_literal_pools_p ().
13096 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13099 * optabs.c (emit_conditional_move): Pass address of cmode's copy
13100 rather than address of cmode as last argument to prepare_cmp_insn.
13102 2018-03-15 Julia Koval <julia.koval@intel.com>
13104 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
13105 F_AVX512VNNI, F_AVX512BITALG): New.
13107 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
13110 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
13111 insn for floating-point loads and stores.
13113 2018-03-14 Carl Love <cel@us.ibm.com>
13115 * config/rs6000/rs6000-c.c: Add macro definitions for
13116 ALTIVEC_BUILTIN_VEC_PERMXOR.
13117 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
13118 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
13119 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
13120 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
13122 * config/doc/extend.texi: Add prototypes for vec_permxor.
13124 2018-03-14 David Malcolm <dmalcolm@redhat.com>
13127 * diagnostic-show-locus.c (class layout_point): Convert m_line
13128 from int to linenum_type.
13129 (line_span::comparator): Use linenum "compare" function when
13130 comparing line numbers.
13131 (test_line_span): New function.
13132 (layout_range::contains_point): Convert param "row" from int to
13134 (layout_range::intersects_line_p): Likewise.
13135 (layout::will_show_line_p): Likewise.
13136 (layout::print_source_line): Likewise.
13137 (layout::should_print_annotation_line_p): Likewise.
13138 (layout::print_annotation_line): Likewise.
13139 (layout::print_leading_fixits): Likewise.
13140 (layout::annotation_line_showed_range_p): Likewise.
13141 (struct line_corrections): Likewise for field m_row.
13142 (line_corrections::line_corrections): Likewise for param "row".
13143 (layout::print_trailing_fixits): Likewise.
13144 (layout::get_state_at_point): Likewise.
13145 (layout::get_x_bound_for_row): Likewise.
13146 (layout::print_line): Likewise.
13147 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
13148 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
13149 * input.c (selftest::test_linenum_comparisons): New function.
13150 (selftest::input_c_tests): Call it.
13151 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
13152 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
13153 * selftest.h (ASSERT_GT): New macro.
13154 (ASSERT_GT_AT): New macro.
13155 (ASSERT_LT): New macro.
13156 (ASSERT_LT_AT): New macro.
13158 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
13160 PR rtl-optimization/84780
13161 * combine.c (distribute_links): Don't make a link based on pc_rtx.
13163 2018-03-14 Martin Liska <mliska@suse.cz>
13165 * tree.c (record_node_allocation_statistics): Use
13166 get_stats_node_kind.
13167 (get_stats_node_kind): New function extracted from
13168 record_node_allocation_statistics.
13169 (free_node): Use get_stats_node_kind.
13171 2018-03-14 Richard Biener <rguenther@suse.de>
13173 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
13174 that the value-set of ANTIC_IN doesn't grow.
13177 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13179 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13180 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13182 2018-03-14 Julia Koval <julia.koval@intel.com>
13184 * config.gcc (icelake-client, icelake-server): New.
13186 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
13187 (initial_ix86_arch_features): Ditto.
13188 (PTA_SKYLAKE): Add SGX.
13189 (PTA_ICELAKE): Remove.
13190 (PTA_ICELAKE_CLIENT): New.
13191 (PTA_ICELAKE_SERVER): New.
13192 (ix86_option_override_internal): Split up icelake on icelake client and
13194 (get_builtin_code_for_version): Ditto.
13195 (fold_builtin_cpu): Ditto.
13196 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
13197 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
13198 * config/i386/i386.h (processor_type): Ditto.
13199 * doc/invoke.texi: Ditto.
13201 2018-03-14 Jakub Jelinek <jakub@redhat.com>
13204 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
13205 INTEGER_CST offset, add it together with bitpos / 8 and
13206 sign extend based on POINTER_SIZE.
13210 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
13213 * config/i386/constraints.md (Yc): New register constraint.
13214 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
13215 Use Yc constraint for alternative 2 of operand 0. Remove
13216 preferred_for_speed attribute.
13218 2018-03-14 Richard Biener <rguenther@suse.de>
13220 PR tree-optimization/84830
13221 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
13222 with the old one to avoid oscillations.
13224 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
13227 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
13229 (assign_by_spills): Return a flag of reload assignment failure.
13230 Do not process the reload assignment failures. Do not spill other
13231 reload pseudos if they has the same reg class. Update n if
13233 (lra_assign): Add a return arg. Set up from the result of
13234 assign_by_spills call.
13235 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13236 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13237 usage_insns if it is not NULL.
13238 (spill_hard_reg_in_range): New function.
13239 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13240 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13241 function prototypes.
13242 (lra_assign): Change prototype.
13243 * lra.c (lra): Add code to deal with fails by splitting hard reg
13246 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
13248 * config/riscv/riscv.opt (mrelax): New option.
13249 * config/riscv/riscv.c (riscv_file_start): Emit ".option
13250 "norelax" when riscv_mrelax is disabled.
13251 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
13253 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13256 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
13257 reassociation for int modes.
13259 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13261 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
13262 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
13264 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
13265 * config/aarch64/aarch64-sve.md
13266 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
13267 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
13268 (*extend<mode><Vwide>2): Rename to...
13269 (aarch64_sve_extend<mode><Vwide>2): ...this.
13270 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
13271 renaming the old pattern to...
13272 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
13274 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
13275 define_expand, renaming the old pattern to...
13276 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
13277 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
13278 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
13279 account when deciding which SVE instruction the optab should use.
13280 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
13282 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13284 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
13285 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
13286 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
13287 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
13288 (tlsdesc_small_<mode>): Turn a define_expand and use
13289 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
13290 (tlsdesc_small_advsimd_<mode>): ...this.
13291 (tlsdesc_small_sve_<mode>): New pattern.
13293 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13295 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
13296 (UNSPEC_UMUL_HIGHPART): New constants.
13297 (MUL_HIGHPART): New int iteraor.
13298 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
13299 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
13301 (*<su>mul<mode>3_highpart): New define_insn.
13303 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13306 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13309 2018-03-13 Martin Liska <mliska@suse.cz>
13312 * (sem_item_optimizer::sem_item_optimizer): Initialize new
13314 (sem_item_optimizer::~sem_item_optimizer): Release it.
13315 (sem_item_optimizer::merge_classes): Register variable aliases.
13316 (sem_item_optimizer::fixup_pt_set): New function.
13317 (sem_item_optimizer::fixup_points_to_sets): Likewise.
13318 * ipa-icf.h: Declare new variables and functions.
13320 2018-03-13 Jakub Jelinek <jakub@redhat.com>
13322 PR middle-end/84834
13323 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
13324 integer_pow2p@2 and test integer_pow2p in condition.
13325 (A < 0 ? C : 0): Similarly for @1.
13327 PR middle-end/84831
13328 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
13329 characters starting at p contain '\0' character, don't look beyond
13333 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
13334 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
13337 * reg-stack.c (change_stack): Change update_end var from int to
13338 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
13339 also call set_block_for_insn on the newly added insns and rescan.
13342 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
13343 on the last operand.
13346 * tree.c (stabilize_reference_1): Return save_expr (e) for
13347 STATEMENT_LIST even if it doesn't have side-effects.
13349 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
13351 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
13353 2018-03-12 Renlin Li <renlin.li@arm.com>
13355 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
13356 aarch64_output_scalar_simd_mov_immediate.
13358 2018-03-12 Martin Sebor <msebor@redhat.com>
13360 PR tree-optimization/83456
13361 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
13362 for perfectly overlapping calls to memcpy.
13363 (gimple_fold_builtin_memory_chk): Same.
13364 (gimple_fold_builtin_strcpy): Handle no-warning.
13365 (gimple_fold_builtin_stxcpy_chk): Same.
13366 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
13368 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13370 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
13371 parameter. Use it for SFmode.
13372 (rs6000_function_arg_advance_1): Adjust.
13373 (rs6000_function_arg): Adjust.
13374 (rs6000_gimplify_va_arg): Pass false for that new parameter.
13376 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13378 PR rtl-optimization/84169
13379 PR rtl-optimization/84780
13380 * combine.c (can_combine_p): Check for a 2-insn combination whether
13381 the destination register is used between the two insns, too.
13383 2018-03-12 Richard Biener <rguenther@suse.de>
13385 PR tree-optimization/84803
13386 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
13387 for refs DR analysis didn't process.
13389 2018-03-12 Richard Biener <rguenther@suse.de>
13391 PR tree-optimization/84777
13392 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
13393 force-vectorize loops ignore whether we are optimizing for size.
13395 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
13397 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
13398 (TARGET_MD_ASM_ADJUST): Define.
13400 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
13401 Kito Cheng <kito.cheng@gmail.com>
13402 Chung-Ju Wu <jasonwucj@gmail.com>
13404 * config/nds32/nds32.c (nds32_compute_stack_frame,
13405 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
13406 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
13407 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
13408 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
13409 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
13410 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
13411 * config/nds32/nds32.md (prologue, epilogue): Use macro
13412 NDS32_V3PUSH_AVAILABLE_P to do checking.
13414 2018-03-11 Jakub Jelinek <jakub@redhat.com>
13417 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
13418 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
13419 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
13420 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
13421 addition of most attributes on !orig_type_die or the attribute not
13422 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
13424 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13425 Chung-Ju Wu <jasonwucj@gmail.com>
13427 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
13428 __NDS32_VH__ macro.
13429 * config/nds32/nds32.opt (mvh): New option.
13431 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13432 Chung-Ju Wu <jasonwucj@gmail.com>
13434 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
13436 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
13437 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
13440 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13441 Chung-Ju Wu <jasonwucj@gmail.com>
13443 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
13445 * config/nds32/nds32-multiple.md (strlensi): New pattern.
13446 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
13448 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
13449 Kito Cheng <kito.cheng@gmail.com>
13450 Chung-Ju Wu <jasonwucj@gmail.com>
13452 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
13453 UNSPEC_FFMISM and UNSPEC_FLMISM.
13454 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
13455 for ffb, ffmism and flmism.
13456 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
13457 (unspec_ffmism): Ditto.
13458 (unspec_flmism): Ditto.
13459 (nds32_expand_builtin_impl): Check if string extension is available.
13460 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
13461 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
13463 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
13466 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13469 * lra-assigns.c (assign_by_spills): Return a flag of reload
13470 assignment failure. Do not process the reload assignment
13471 failures. Do not spill other reload pseudos if they has the same
13473 (lra_assign): Add a return arg. Set up from the result of
13474 assign_by_spills call.
13475 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13476 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13477 usage_insns if it is not NULL.
13478 (spill_hard_reg_in_range): New function.
13479 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13480 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13481 function prototypes.
13482 (lra_assign): Change prototype.
13483 * lra.c (lra): Add code to deal with fails by splitting hard reg
13486 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
13489 * config/i386/i386.opt: Replace Enforcment with Enforcement.
13491 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
13494 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
13495 (dw_val_node): Add val_symbolic_view.
13496 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
13497 (symview_upper_bound): New.
13498 (new_line_info_table): Initialize symviews_since_reset.
13499 (dwarf2out_source_line): Count symviews_since_reset and set
13500 symview_upper_bound.
13501 (dw_val_equal_p): Handle symview.
13502 (add_AT_symview): New.
13503 (print_dw_val): Handle symview.
13504 (attr_checksum, attr_checksum_ordered): Likewise.
13505 (same_dw_val_p, size_of_die): Likewise.
13506 (value_format, output_die): Likewise.
13507 (add_high_low_attributes): Use add_AT_symview for entry_view.
13508 (dwarf2out_finish): Reset symview_upper_bound, clear
13511 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
13514 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
13515 Add strict argument and use it.
13516 (rs6000_split_multireg_move): Update for new strict argument.
13517 (mem_operand_gpr): Disallow all non-offsettable addresses.
13518 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
13520 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13523 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
13524 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
13525 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
13528 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
13529 decl, use remap_type if we want to use the type.
13531 2018-03-09 Martin Sebor <msebor@redhat.com>
13533 PR tree-optimization/84526
13534 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13536 (builtin_access::generic_overlap): Be prepared to handle non-array
13539 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
13541 PR rtl-optimization/84682
13542 * lra-constraints.c (process_address_1): Check is_address flag
13543 for address constraints.
13544 (process_alt_operands): Likewise.
13545 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
13546 preprocess_constraints.
13547 * recog.h (preprocess_constraints): Add oploc parameter.
13549 * recog.c (preprocess_constraints): Test address_operand for
13550 CT_ADDRESS constraints.
13552 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13555 * lra-assigns.c (assign_by_spills): Return a flag of reload
13556 assignment failure. Do not process the reload assignment
13557 failures. Do not spill other reload pseudos if they has the same
13559 (lra_assign): Add a return arg. Set up from the result of
13560 assign_by_spills call.
13561 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13562 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13563 usage_insns if it is not NULL.
13564 (spill_hard_reg_in_range): New function.
13565 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13566 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13567 function prototypes.
13568 (lra_assign): Change prototype.
13569 * lra.c (lra): Add code to deal with fails by splitting hard reg
13572 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13575 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
13576 Accept complain bool parameter. Only emit errors if it is true.
13577 (arm_parse_cpu_option_name): Likewise.
13578 (arm_target_thumb_only): Adjust callers of the above.
13579 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
13580 prototype to take a default true bool parameter.
13581 (arm_parse_arch_option_name): Likewise.
13583 2018-03-09 David Malcolm <dmalcolm@redhat.com>
13584 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
13588 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
13589 * configure: Regenerate.
13590 * configure.ac ("linker --version-script option"): New.
13591 ("linker soname option"): New.
13593 2018-03-09 Richard Biener <rguenther@suse.de>
13595 PR tree-optimization/84775
13596 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
13597 immediate uses of predicate stmts and mark them modified.
13600 PR tree-optimization/84178
13601 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13603 (version_loop_for_if_conversion): Delay update_ssa call.
13604 (tree_if_conversion): Delay update_ssa until after predicate
13607 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
13610 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
13611 when the function accesses prior frames.
13613 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13616 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
13617 gen_llsym, otherwise call maybe_gen_llsym.
13619 PR inline-asm/84742
13620 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
13621 has ',' character inside of it.
13623 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13626 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
13627 as clobbering CC_REGNUM.
13629 2018-03-08 Richard Biener <rguenther@suse.de>
13631 PR middle-end/84552
13632 * tree-scalar-evolution.c: Include tree-into-ssa.h.
13633 (follow_copies_to_constant): Do not follow SSA names registered
13636 2018-03-08 Richard Biener <rguenther@suse.de>
13638 PR tree-optimization/84178
13639 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13641 (version_loop_for_if_conversion): Delay update_ssa call.
13642 (tree_if_conversion): Delay update_ssa until after predicate
13645 2018-03-08 David Malcolm <dmalcolm@redhat.com>
13647 PR tree-optimization/84178
13648 * tree-if-conv.c (release_bb_predicate): Remove the
13649 the assertion that the stmts have NULL use_ops.
13650 Discard the statements, asserting that they haven't
13651 yet been added to a BB.
13653 2018-03-08 Richard Biener <rguenther@suse.de>
13655 PR tree-optimization/84746
13656 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
13657 (phi_translate): Pass in destination ANTIC_OUT set.
13658 (phi_translate_1): Likewise. For a simplified result lookup
13659 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
13660 (phi_translate_set): Adjust.
13661 (do_pre_regular_insertion): Likewise.
13662 (do_pre_partial_partial_insertion): Likewise.
13664 2018-03-08 Martin Liska <mliska@suse.cz>
13666 PR gcov-profile/84735
13667 * doc/gcov.texi: Document usage of profile files.
13668 * gcov-io.h: Document changes in the format.
13670 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
13674 * dwarf2out.c (struct dw_line_info_table): Update comments for
13676 (FORCE_RESET_NEXT_VIEW): New.
13677 (FORCE_RESETTING_VIEW_P): New.
13678 (RESETTING_VIEW_P): Check for -1 too.
13679 (ZERO_VIEW_P): Likewise.
13680 (new_line_info_table): Force-reset next view.
13681 (dwarf2out_begin_function): Likewise.
13682 (dwarf2out_source_line): Simplify zero_view_p initialization.
13683 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
13684 view directly. Omit view when omitting .loc at line 0.
13686 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13688 PR tree-optimization/84740
13689 * tree-switch-conversion.c (process_switch): Call build_constructors
13690 only if info.phi_count is non-zero.
13692 PR tree-optimization/84739
13693 * tree-tailcall.c (find_tail_calls): Check call arguments against
13694 DECL_ARGUMENTS (current_function_decl) rather than
13695 DECL_ARGUMENTS (func) when checking for tail recursion.
13697 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13699 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
13700 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
13701 Volker Reichelt's entry and add entries for people that perform
13702 GCC fuzzy testing and report numerous bugs.
13704 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
13707 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
13708 readonly data in sdata, if that is disabled.
13709 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
13710 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
13711 -mreadonly-in-sdata option.
13713 2018-03-07 Martin Sebor <msebor@redhat.com>
13715 PR tree-optimization/84468
13716 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
13717 basic block when looking for nul assignment.
13719 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
13722 * except.h (output_function_exception_table): Adjust prototype.
13723 * except.c (output_function_exception_table): Remove FNNAME parameter
13724 and add SECTION parameter. Ouput one part of the table at a time.
13725 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
13726 the first part of the exception table and emit unwind directives.
13727 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
13728 (i386_pe_seh_cold_init): Likewise.
13729 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
13730 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
13731 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
13732 (ix86_output_call_insn): Emit a nop in one more case for SEH.
13733 * config/i386/winnt.c: Include except.h.
13734 (struct seh_frame_state): Add reg_offset, after_prologue and
13735 in_cold_section fields.
13736 (i386_pe_seh_end_prologue): Set seh->after_prologue.
13737 (i386_pe_seh_cold_init): New function.
13738 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
13739 to seh->in_cold_section.
13740 (seh_emit_push): Record the offset of the push.
13741 (seh_emit_save): Record the offet of the save.
13742 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
13743 Test seh->after_prologue to disregard the epilogue.
13744 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
13745 (i386_pe_end_cold_function): New function.
13747 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13750 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
13751 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
13754 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
13756 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
13757 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
13759 PR middle-end/84723
13760 * multiple_target.c: Include tree-inline.h and intl.h.
13761 (expand_target_clones): Diagnose and fail if node->definition and
13762 !tree_versionable_function_p (node->decl).
13764 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
13766 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13768 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13769 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13770 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13772 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13775 * combine.c (try_combine): Use reg_or_subregno instead of handling
13776 just paradoxical SUBREGs and REGs.
13778 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13780 * config/arc/arc.c (arc_finalize_pic): Remove function.
13781 (arc_must_save_register): We use single base PIC register, remove
13782 checks to save/restore the PIC register.
13783 (arc_expand_prologue): Likewise.
13784 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13786 (arc_verify_short): Likewise.
13787 (arc_attr_type): Likewise.
13788 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13789 (walk_stores): Likewise.
13790 (arc_address_cost): Make it static.
13791 (arc_verify_short): Likewise.
13792 (branch_dest): Likewise.
13793 (arc_attr_type): Likewise.
13794 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13795 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13796 (arc_final_prescan_insn): Remove inserting the nops due to
13797 hardware hazards. It is done in reorg step.
13798 (insn_length_variant_t): Remove.
13799 (insn_length_parameters_t): Likewise.
13800 (arc_insn_length_parameters): Likewise.
13801 (arc_get_insn_variants): Likewise.
13802 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13804 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13806 PR inline-asm/84683
13807 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13810 PR tree-optimization/84687
13811 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13813 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13815 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13817 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13818 Rename to ppc_speculation_barrier.
13819 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13820 __builtin_ppc_speculation_barrier.
13822 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13825 * combine.c (combine_simplify_rtx): Don't try to simplify if
13826 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13829 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13831 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13832 to 32 bytes when compiling for POWER9.
13834 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13837 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13838 regparm >= 3 with no arg reg available also for calls with
13839 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13842 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13844 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13846 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13849 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13851 2018-03-05 Richard Biener <rguenther@suse.de>
13853 PR tree-optimization/84486
13854 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13855 When inserting a __builtin_assume_aligned call set the LHS
13856 SSA name alignment info accordingly.
13858 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13860 PR tree-optimization/84114
13861 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13862 Avoid reassociation of FLOAT_MODE addition.
13864 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13866 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13867 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13868 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13869 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13870 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13871 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13872 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13874 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13875 __builtin_ia32_wbinvd): New builtins.
13876 (SPECIAL_ARGS2): New.
13877 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13878 (SPECIAL_ARGS2): New.
13879 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13880 (ix86_valid_target_attribute_inner_p): Ditto.
13881 (ix86_init_mmx_sse_builtins): Add special_args2.
13882 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13883 TARGET_WBNOINVD_P): New.
13884 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13885 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13886 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13887 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13888 * config/i386/pconfigintrin.h: New file.
13889 * config/i386/wbnoinvdintrin.h: Ditto.
13890 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13892 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13894 2018-03-05 Richard Biener <rguenther@suse.de>
13896 PR tree-optimization/84670
13897 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13899 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13900 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13901 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13902 doesn't grow if all successors (recursively) were visited at least
13905 2018-03-05 Richard Biener <rguenther@suse.de>
13907 PR tree-optimization/84650
13908 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13909 if executed in the loop pipeline.
13911 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13913 * doc/configfiles.texi (Configuration Files): Move info about
13914 conditionalizing $target-protos.h to...
13915 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13916 differs from $target-protos.h.
13918 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13919 Chung-Ju Wu <jasonwucj@gmail.com>
13921 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13922 * config/nds32/nds32-multiple.md (setmemsi): Define.
13923 * config/nds32/nds32-memory-manipulation.c
13924 (nds32_gen_dup_4_byte_to_word_value): New.
13925 (emit_setmem_word_loop): New.
13926 (emit_setmem_byte_loop): New.
13927 (nds32_expand_setmem_loop): New.
13928 (nds32_expand_setmem_loop_v3m): New.
13929 (nds32_expand_setmem_unroll): New.
13930 (nds32_expand_setmem): New.
13932 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13933 Chung-Ju Wu <jasonwucj@gmail.com>
13935 * config/nds32/nds32-memory-manipulation.c
13936 (nds32_emit_load_store): New.
13937 (nds32_emit_post_inc_load_store): New.
13938 (nds32_emit_mem_move): New.
13939 (nds32_emit_mem_move_block): New.
13940 (nds32_expand_movmemsi_loop_unknown_size): New.
13941 (nds32_expand_movmemsi_loop_known_size): New.
13942 (nds32_expand_movmemsi_loop): New.
13943 (nds32_expand_movmemsi_unroll): New.
13944 (nds32_expand_movmemqi): Rename ...
13945 (nds32_expand_movmemsi): ... to this.
13946 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13947 (movmemsi): ... to this.
13948 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13949 (nds32_expand_movmemsi): ... to this.
13951 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13952 Monk Chiang <sh.chiang04@gmail.com>
13953 Chung-Ju Wu <jasonwucj@gmail.com>
13955 * config/nds32/nds32-protos.h
13956 (nds32_expand_load_multiple): New arguments.
13957 (nds32_expand_store_multiple): Ditto.
13958 (nds32_valid_multiple_load_store): Rename ...
13959 (nds32_valid_multiple_load_store_p): ... to this.
13960 * config/nds32/nds32-memory-manipulation.c
13961 (nds32_expand_load_multiple): Refine implementation.
13962 (nds32_expand_store_multiple): Ditto.
13963 * config/nds32/nds32-multiple.md
13964 (load_multiple): Update nds32_expand_load_multiple interface.
13965 (store_multiple): Update nds32_expand_store_multiple interface.
13966 * config/nds32/nds32-predicates.c
13967 (nds32_valid_multiple_load_store): Rename ...
13968 (nds32_valid_multiple_load_store_p): ... to this and refine
13970 * config/nds32/predicates.md
13971 (nds32_load_multiple_and_update_address_operation): New predicate.
13972 (nds32_store_multiple_and_update_address_operation): New predicate.
13974 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13975 Chung-Ju Wu <jasonwucj@gmail.com>
13977 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13978 (combo): New attribute.
13979 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13981 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13983 * config/nds32/nds32.opt: Change -mcmodel= default value.
13985 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13986 Monk Chiang <sh.chiang04@gmail.com>
13987 Chung-Ju Wu <jasonwucj@gmail.com>
13989 * config/nds32/constants.md (unspec_element): New enum.
13990 * config/nds32/constraints.md (Umw): New constraint.
13991 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13992 * config/nds32/nds32-intrinsic.md: Likewise.
13993 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13994 (nds32_valid_smw_lwm_base_p): New.
13995 (nds32_output_smw_single_word): New.
13996 (nds32_output_lmw_single_word): New.
13997 (nds32_expand_unaligned_load): New.
13998 (nds32_expand_unaligned_store): New.
13999 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
14000 (nds32_output_smw_single_word): Declare.
14001 (nds32_output_lmw_single_word): Declare.
14002 (nds32_expand_unaligned_load): Declare.
14003 (nds32_expand_unaligned_store): Declare.
14004 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
14005 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
14006 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
14007 NDS32_BUILTIN_UASTORE_DW.
14008 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
14011 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
14012 Kito Cheng <kito.cheng@gmail.com>
14013 Chung-Ju Wu <jasonwucj@gmail.com>
14015 * config/nds32/nds32-intrinsic.c
14016 (nds32_expand_builtin_null_ftype_reg): Delete.
14017 (nds32_expand_builtin_reg_ftype_imm): Ditto.
14018 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
14019 (nds32_read_argument): New.
14020 (nds32_legitimize_target): Ditto.
14021 (nds32_legitimize_argument): Ditto.
14022 (nds32_check_constant_argument): Ditto.
14023 (nds32_expand_unop_builtin): Ditto.
14024 (nds32_expand_unopimm_builtin): Ditto.
14025 (nds32_expand_binop_builtin): Ditto.
14026 (nds32_builtin_decl_impl): Ditto.
14027 (builtin_description): Ditto.
14028 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
14029 (nds32_init_builtins_impl): Ditto.
14030 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
14031 (nds32_builtin_decl): New.
14032 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
14033 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
14035 2018-03-02 Jeff Law <law@redhat.com>
14037 * reorg.c (stop_search_p): Handle DEBUG_INSN.
14038 (redundant_insn, fill_simple_delay_slots): Likewise.
14039 (fill_slots_from_thread): Likewise.
14040 * resource.c (mark_referenced_resources): Likewise.
14041 (mark_set_resources, find_dead_or_set_registers): Likewise.
14043 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14045 * substring-locations.h (format_warning_va): Formatting fix for
14046 ATTRIBUTE_GCC_DIAG.
14047 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
14049 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
14050 * substring-locations.c: Include intl.h.
14051 (format_warning_va): Turned into small wrapper around
14052 format_warning_n_va, renamed to ...
14053 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
14054 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
14056 (format_warning_at_substring_n): New function.
14057 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
14058 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
14059 format_warning_at_substring with just a shorter name instead of
14060 const function pointer.
14061 (fmtwarn_n): New function.
14062 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
14063 appropriate, get rid of all the fmtstr temporaries, move conditionals
14064 with G_() wrapped string literals directly into fmtwarn arguments,
14065 cast dir.len to (int), formatting fixes.
14067 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
14069 * doc/invoke.texi: Remove "Cilk Plus" references.
14071 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14072 Richard Biener <rguenther@suse.de>
14075 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
14076 for error or warning attributes if CALL_FROM_THUNK_P is set.
14079 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14082 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
14083 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
14086 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
14087 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
14089 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
14090 instead of -1U in last predictors element's probability member.
14092 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14095 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14096 arguments if they are comparable.
14098 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14100 PR tree-optimization/84634
14101 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
14102 masks and masked_loop_p with a single loop_masks, making sure it's
14103 null for bb vectorization.
14105 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14107 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14108 (vect_analyze_data_ref_access): Use loop->safe_len rather than
14109 loop->force_vectorize to check whether there is no alias.
14111 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14114 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
14116 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
14118 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
14119 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
14120 instead of a loop around prev_real_insn.
14121 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
14124 PR inline-asm/84625
14125 * config/i386/i386.c (ix86_print_operand): Use conditional
14126 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
14129 2018-03-02 Richard Biener <rguenther@suse.de>
14131 PR tree-optimization/84427
14132 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
14133 (bitmap_set_subtract_values): Rewrite to handle multiple
14136 (prune_clobbered_mems): Likewise.
14137 (phi_translate): Take edge instead of pred/phiblock.
14138 (phi_translate_1): Likewise.
14139 (phi_translate_set): Likewise. Insert all translated
14140 exprs for a value into the set, keeping possibly multiple
14141 expressions per value.
14142 (compute_antic_aux): Adjust for phi_translate changes.
14143 When intersecting union the expressions and prune those
14144 not in the final value set, keeping possibly multiple
14145 expressions per value. Do not use value-insertion
14146 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
14147 all expressions. Add verification that the value-sets
14148 only shrink during iteration.
14149 (compute_partial_antic_aux): Adjust for the phi_translate changes.
14150 (do_pre_regular_insertion): Likewise.
14151 (do_pre_partial_partial_insertion): Likewise.
14153 2018-03-02 Richard Biener <rguenther@suse.de>
14156 * config/darwin.c (saved_debug_info_level): New static global.
14157 (darwin_asm_lto_start): Disable debug info generation for LTO out.
14158 (darwin_asm_lto_end): Restore debug info generation settings.
14160 2018-03-01 Martin Liska <mliska@suse.cz>
14163 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
14164 volatile arguments.
14166 2018-03-01 Richard Biener <rguenther@suse.de>
14169 * dwarf2out.c (gen_variable_die): Properly handle late VLA
14170 type annotation with LTO when debug was disabled at compile-time.
14172 2018-03-01 Matthew Fortune <mfortune@gmail.com>
14174 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
14176 (mips_final_postscan_insn): Likewise.
14178 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
14180 PR rtl-optimization/84528
14181 * alias.c (init_alias_target): Add commentary.
14182 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
14183 a unique base value if the frame pointer is not eliminated
14184 to the stack pointer.
14186 2018-03-01 Tom de Vries <tom@codesourcery.com>
14188 PR rtl-optimization/83327
14189 * lra-int.h (hard_regs_spilled_into): Declare.
14190 * lra.c (hard_regs_spilled_into): Define.
14191 (init_reg_info): Init hard_regs_spilled_into.
14192 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
14193 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
14194 (process_bb_lives): Handle hard_regs_spilled_into.
14195 (lra_create_live_ranges_1): Before doing liveness propagation, clear
14196 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
14198 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
14200 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
14201 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
14202 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
14203 * config/rs6000/aix72.h: New file.
14205 2018-02-28 Jakub Jelinek <jakub@redhat.com>
14207 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
14208 instead of warning_at with conditional singular and plural messages
14212 * stor-layout.c (update_alignment_for_field): For
14213 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
14214 && !DECL_PACKED (field), do the alignment update, just use
14215 only desired_align instead of MAX (type_align, desired_align)
14217 (place_field): Don't do known_align < desired_align handling
14218 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
14219 is non-NULL, instead do it after rli->prev_field handling and
14220 only if not within a bitfield word. For DECL_PACKED (field)
14221 use type_align of BITS_PER_UNIT.
14223 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
14225 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
14226 superfluous parentheses and trailing spaces.
14228 2018-02-28 Richard Biener <rguenther@suse.de>
14230 PR tree-optimization/84584
14231 * graphite-scop-detection.c (scop_detection::add_scop): Discard
14232 SCoPs with fake exit edge.
14234 2018-02-28 Martin Liska <mliska@suse.cz>
14237 * timevar.c (timer::print): Fix format to properly print 100%
14240 2018-02-28 Richard Biener <rguenther@suse.de>
14242 PR middle-end/84607
14243 * genmatch.c (capture_info::walk_match): Do not mark
14244 captured expressions without operands as expr_p given
14245 they act more like predicates and should be subject to
14246 "lost tail" side-effect preserving.
14248 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
14250 PR rtl-optimization/81611
14251 * auto-inc-dec.c (attempt_change): Move dead note from
14252 mem_insn if it's the next use of regno
14253 (find_address): Take address use of reg holding
14254 non-incremented value. Add parm to limit search to the named
14256 (merge_in_block): Attempt to use a mem insn that is the next
14257 use of the original regno.
14259 2018-02-27 Martin Sebor <msebor@redhat.com>
14262 * doc/invoke.texi (-Wmissing-attributes): New option.
14263 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
14265 2018-02-27 Martin Sebor <msebor@redhat.com>
14267 PR translation/84207
14268 * diagnostic-core.h (warning_n, error_n, inform_n): Change
14269 n argument to unsigned HOST_WIDE_INT.
14270 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
14271 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
14272 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
14273 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
14275 2018-02-27 Richard Biener <rguenther@suse.de>
14277 PR tree-optimization/84512
14278 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
14279 Do not use the estimate returned from record_stmt_cost for
14280 the scalar iteration cost but sum properly using add_stmt_cost.
14282 2018-02-27 Richard Biener <rguenther@suse.de>
14284 PR tree-optimization/84466
14285 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14286 Adjust last change to less strictly validate use operands.
14288 2018-02-27 Martin Liska <mliska@suse.cz>
14290 PR gcov-profile/84548
14291 * gcov.c (process_file): Allow partial overlap and consider it
14292 also as group functions.
14293 (output_lines): Properly calculate range of lines for a group.
14295 2018-02-27 Martin Liska <mliska@suse.cz>
14297 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
14298 'ggc' suffixes. Change first column width.
14299 (timer::print): Fix formatting of the column.
14301 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
14303 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
14304 preserve inline entry blocks for the sake of debug inline
14305 entry point markers alone.
14306 (remove_unused_locals): Suggest in comments a better place to
14307 force the preservation of inline entry blocks that are
14308 otherwise unused, but do not preserve them.
14310 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14312 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
14314 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14317 * config/i386/constraints.md (Bs): Replace
14318 ix86_indirect_branch_register with
14319 TARGET_INDIRECT_BRANCH_REGISTER.
14321 * config/i386/i386.md (indirect_jump): Likewise.
14322 (tablejump): Likewise.
14323 (*sibcall_memory): Likewise.
14324 (*sibcall_value_memory): Likewise.
14325 Peepholes of indirect call and jump via memory: Likewise.
14326 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
14327 (*sibcall_value_GOT_32): Likewise.
14328 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14329 (GOT_memory_operand): Likewise.
14330 (call_insn_operand): Likewise.
14331 (sibcall_insn_operand): Likewise.
14332 (GOT32_symbol_operand): Likewise.
14333 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
14335 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14337 PR rtl-optimization/83496
14338 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
14339 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
14340 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
14341 redundant insn, if any.
14342 (relax_delay_slots): Likewise.
14343 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
14345 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
14347 PR tree-optimization/83965
14348 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
14349 that grouped statements are part of a reduction chain. Return
14350 true if the statement is not marked as a reduction itself but
14351 is part of a group.
14352 (vect_recog_dot_prod_pattern): Don't check whether the statement
14353 is part of a group here.
14354 (vect_recog_sad_pattern): Likewise.
14355 (vect_recog_widen_sum_pattern): Likewise.
14357 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14360 * final.c (rest_of_clean_state): Also look for calls inside sequences.
14362 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14365 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
14367 (ix86_output_indirect_function_return): New prototype.
14368 (ix86_split_simple_return_pop_internal): Likewise.
14369 * config/i386/i386.c (indirect_return_via_cx): New.
14370 (indirect_return_via_cx_bnd): Likewise.
14371 (indirect_thunk_name): Handle return va CX_REG.
14372 (output_indirect_thunk_function): Create alias for
14373 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
14374 (ix86_output_indirect_jmp): Remove the bool argument.
14375 (ix86_output_indirect_function_return): New function.
14376 (ix86_split_simple_return_pop_internal): Likewise.
14377 * config/i386/i386.md (*indirect_jump): Don't pass false
14378 to ix86_output_indirect_jmp.
14379 (*tablejump_1): Likewise.
14380 (simple_return_pop_internal): Change it to define_insn_and_split.
14381 Call ix86_split_simple_return_pop_internal to split it for
14382 -mfunction-return=.
14383 (simple_return_indirect_internal): Call
14384 ix86_output_indirect_function_return instead of
14385 ix86_output_indirect_jmp.
14387 2018-02-26 Jakub Jelinek <jakub@redhat.com>
14390 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
14391 memset and value initialization afterwards.
14393 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
14395 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
14397 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14400 * common/config/aarch64/aarch64-common.c
14401 (aarch_option_optimization_table[]): Switch
14402 off fomit-frame-pointer
14404 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14405 Chung-Ju Wu <jasonwucj@gmail.com>
14407 * config/nds32/nds32-multiple.md (load_multiple): Disallow
14409 (store_multiple): Ditto.
14411 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14413 * config.gcc: Add --with-cpu support for nds32 target.
14414 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
14415 * config/nds32/nds32.opt: Add -mcpu= option.
14417 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
14419 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
14420 isel=yes): Warn for these deprecated options.
14422 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
14424 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
14425 ISA_2_5_MASKS_EMBEDDED.
14427 2018-02-23 Jakub Jelinek <jakub@redhat.com>
14429 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
14430 p->max as pointers rather than using iterative_hash_expr.
14432 2018-02-23 Carl Love <cel@us.ibm.com>
14434 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
14435 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
14437 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
14438 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
14439 P8V_BUILTIN_VEC_VUNSIGNED2.
14441 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
14444 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
14445 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
14447 * lra-constraints.c (curr_insn_transform): Set up
14448 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
14450 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
14451 * lra-eliminations.c (spill_pseudos): Ditto.
14452 (process_insn_for_elimination): Ditto.
14453 * lra-lives.c (reg_early_clobber_p): Use the new macros.
14454 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
14455 LRA_NON_CLOBBERED_ALT.
14457 2018-02-22 Martin Sebor <msebor@redhat.com>
14459 PR tree-optimization/84480
14460 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
14461 to maybe_diag_stxncpy_trunc. Call it.
14462 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
14463 from gimple_fold_builtin_strcpy. Print inlining stack.
14464 (handle_builtin_stxncpy): Print inlining stack.
14465 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
14467 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
14470 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
14471 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
14472 and -fcheck-pointer-bounds are used together.
14473 (indirect_thunk_prefix): New enum.
14474 (indirect_thunk_need_prefix): New function.
14475 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
14476 "_nt" instead of "_bnd" for NOTRACK prefix.
14477 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
14478 (output_indirect_thunk_function): Likewise.
14480 (ix86_code_end): Update output_indirect_thunk_function calls.
14481 (ix86_output_indirect_branch_via_reg): Replace
14482 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
14483 (ix86_output_indirect_branch_via_push): Likewise.
14484 (ix86_output_function_return): Likewise.
14485 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
14486 incompatible with -fcf-protection=branch and
14487 -fcheck-pointer-bounds.
14489 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14492 * config/aarch64/aarch64.c (aarch64_print_address_internal):
14493 Change gcc_assert call to output_operand_lossage.
14495 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14497 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
14499 2018-02-22 DJ Delorie <dj@redhat.com>
14500 Sebastian Perta <sebastian.perta@renesas.com>
14501 Oleg Endo <olegendo@gcc.gnu.org>
14503 * config/rx/rx.c (rx_rtx_costs): New function.
14504 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
14506 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
14508 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
14510 2018-02-22 Martin Liska <mliska@suse.cz>
14513 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
14514 Add "native" as a possible value.
14516 2018-02-22 Martin Liska <mliska@suse.cz>
14519 * config/i386/i386.c (ix86_option_override_internal):
14520 Add "native" as a possible value for -march and -mtune.
14522 2018-02-22 Jakub Jelinek <jakub@redhat.com>
14525 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
14526 to all type variants.
14528 PR tree-optimization/84503
14529 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
14530 width as info->bitpos + info->bitsize - start.
14531 (merged_store_group::merge_overlapping): Simplify width computation.
14532 (check_no_overlap): New function.
14533 (imm_store_chain_info::try_coalesce_bswap): Compute expected
14534 start + width and last_order of the group, fail if check_no_overlap
14536 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
14537 to group if check_no_overlap fails.
14539 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14541 * config/rs6000/altivec.md: Delete contraint arguments to
14542 define_expand, define_split, and define_peephole2, and in
14543 define_insn_and_split if always unused.
14544 * config/rs6000/darwin.md: Ditto.
14545 * config/rs6000/dfp.md: Ditto.
14546 * config/rs6000/rs6000.md: Ditto.
14547 * config/rs6000/sync.md: Ditto.
14548 * config/rs6000/vector.md: Ditto.
14549 * config/rs6000/vsx.md: Ditto.
14551 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14553 * config/rs6000/altivec.md: Write output control strings as braced
14554 blocks instead of double-quoted strings.
14555 * config/rs6000/darwin.md: Ditto.
14556 * config/rs6000/rs6000.md: Ditto.
14557 * config/rs6000/vector.md: Ditto.
14558 * config/rs6000/vsx.md: Ditto.
14560 2018-02-21 Jason Merrill <jason@redhat.com>
14562 PR c++/84314 - ICE with templates and fastcall attribute.
14563 * attribs.c (build_type_attribute_qual_variant): Remove assert.
14565 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14567 * ipa-cp.c (determine_versionability): Fix comment typos.
14569 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14572 * ipa-cp.c (determine_versionability): Do not version functions caling
14575 2018-02-21 Martin Liska <mliska@suse.cz>
14578 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
14579 Add "native" as a possible value.
14580 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
14581 the macro when native cpu detection is available.
14583 2018-02-21 Martin Liska <mliska@suse.cz>
14586 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
14587 Add "native" as a possible value.
14588 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
14589 when native cpu detection is available.
14591 2018-02-21 Jakub Jelinek <jakub@redhat.com>
14592 Martin Sebor <msebor@redhat.com>
14594 PR tree-optimization/84478
14595 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
14597 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
14598 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
14599 support which is conservatively correct, for 2 only stay conservative
14600 for maxlen. Formatting and comment capitalization fixes. Add STRICT
14601 argument to the 2 argument get_range_strlen, adjust 6 arg
14602 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
14604 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
14605 (gimple_fold_builtin_strlen): Pass true as last argument to
14608 2018-02-20 Martin Sebor <msebor@redhat.com>
14610 PR middle-end/84095
14611 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
14612 (builtin_memref::set_base_and_offset): Same. Handle inner references.
14613 (builtin_memref::builtin_memref): Factor out parts into
14614 set_base_and_offset and call it.
14616 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14618 PR middle-end/84406
14619 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
14620 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
14621 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
14622 search at the associated MODE_INT.
14624 2018-02-20 Jeff Law <law@redhat.com>
14626 PR middle-end/82123
14627 PR tree-optimization/81592
14628 PR middle-end/79257
14629 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
14630 for range data rather than using global data.
14631 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
14632 range data rather than using global data.
14633 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
14634 pass it to children as needed.
14635 (struct directive::fmtresult): Similarly.
14636 (struct directive::set_width): Similarly.
14637 (struct directive::set_precision): Similarly.
14638 (format_integer, format_directive, parse_directive): Similarly.
14639 (format_none): Accept unnamed vr_values parameter.
14640 (format_percent, format_floating, format_character): Similarly.
14641 (format_string, format_plain): Similarly.
14642 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
14643 the EVRP range analyzer for range data rather than using global data.
14644 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
14645 gimple-ssa-evrp-analyze.h
14646 (class sprintf_dom_walker): Add after_dom_children member function.
14647 Add evrp_range_analyzer member.
14648 (sprintf_dom_walker::before_dom_children): Call into the EVRP
14649 range analyzer as needed.
14650 (sprintf_dom_walker::after_dom_children): New member function.
14651 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
14653 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
14654 (evrp_range_analyzer::pop_to_marker): Likewise.
14656 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14658 PR tree-optimization/84419
14659 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
14660 with the required type if its current type is compatible but
14663 2018-02-20 Jakub Jelinek <jakub@redhat.com>
14665 PR middle-end/82004
14666 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
14667 after vectorization.
14669 2018-02-20 Martin Liska <mliska@suse.cz>
14672 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
14673 possible values if we don't have a hint.
14675 2018-02-20 Martin Liska <mliska@suse.cz>
14679 * final.c (shorten_branches): Build align_tab array with one
14681 * opts.c (finish_options): Add alignment option limit check.
14682 (MAX_CODE_ALIGN): Likewise.
14683 (MAX_CODE_ALIGN_VALUE): Likewise.
14684 * doc/invoke.texi: Document maximum allowed option value for
14685 all -falign-* options.
14687 2018-02-19 Jakub Jelinek <jakub@redhat.com>
14690 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
14691 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
14692 * var-tracking.c (emit_note_insn_var_location): Remove all references
14693 to NOTE_INSN_CALL_ARG_LOCATION.
14694 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
14695 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
14696 Use copy_rtx_if_shared.
14697 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
14698 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
14699 (dwarf2out_var_location): Remove handling of
14700 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
14702 * final.c (final_scan_insn): Remove all references to
14703 NOTE_INSN_CALL_ARG_LOCATION.
14704 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
14705 before dumping final insns.
14706 * except.c (emit_note_eh_region_end): Remove all references to
14707 NOTE_INSN_CALL_ARG_LOCATION.
14708 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
14709 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
14710 * config/arc/arc.c (hwloop_optimize): Likewise.
14711 * config/arm/arm.c (create_fix_barrier): Likewise.
14712 * config/s390/s390.c (s390_chunkify_start): Likewise.
14713 * config/sh/sh.c (find_barrier): Likewise.
14714 * config/i386/i386.c (rest_of_insert_endbranch,
14715 ix86_seh_fixup_eh_fallthru): Likewise.
14716 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
14717 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
14718 * config/frv/frv.c (frv_function_prologue): Likewise.
14719 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
14721 (note_outside_basic_block_p): Remove all references to
14722 NOTE_INSN_CALL_ARG_LOCATION.
14723 * gengtype.c (adjust_field_rtx_def): Likewise.
14724 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
14726 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
14727 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
14730 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
14733 PR tree-optimization/84452
14734 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
14735 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
14738 2018-02-19 Martin Liska <mliska@suse.cz>
14741 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
14743 2018-02-19 Martin Liska <mliska@suse.cz>
14744 Richard Sandiford <richard.sandiford@linaro.org>
14746 PR tree-optimization/82491
14747 * gimple-fold.c (get_base_constructor): Make earlier bail out
14750 2018-02-19 Carl Love <cel@us.ibm.com>
14752 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
14753 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
14755 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
14756 P8V_BUILTIN_VEC_NEG.
14758 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
14760 * config/rl78/rl78.md (movdf): New define expand.
14762 2018-02-19 Martin Liska <mliska@suse.cz>
14765 * doc/invoke.texi: Fix typo.
14766 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14768 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14770 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14771 handle rs6000_single_float and rs6000_double_float specially for
14774 2018-02-16 Jeff Law <law@redhat.com>
14776 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14777 (pop_regs): Use it.
14779 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14782 * ipa-inline.c (inline_small_functions): Fix a typo.
14784 2018-02-16 Nathan Sidwell <nathan@acm.org>
14786 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14788 2018-02-16 Carl Love <cel@us.ibm.com>
14790 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14791 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14792 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14793 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14794 expansion to P8V_BUILTIN_VEC_FLOAT2.
14796 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14798 PR rtl-optimization/70023
14799 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14800 src_regno into account.
14802 2018-02-16 Carl Love <cel@us.ibm.com>
14804 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14805 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14806 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14807 * config/rs6000/rs6000.c: Remove case statements for
14808 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14809 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14810 and P9V_BUILTIN_VEC_VINSERT4B.
14811 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14812 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14813 * config/rs6000/vsx.md:
14814 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14817 2018-02-16 Carl Love <cel@us.ibm.com>
14819 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14821 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14823 * config/rs6000/rs6000-c.c: Add the definitions for
14824 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14825 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14826 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14827 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14828 definition for insert4b and define insn *insert3b_internal.
14829 * doc/extend.texi: Add documentation for vec_extract4b.
14831 2018-02-16 Nathan Sidwell <nathan@acm.org>
14833 * doc/extend.texi (Backwards Compatibility): Mention friend
14834 injection. Note for-scope is deprecated.
14835 * doc/invoke.texi (-ffriend-injection): Deprecate.
14837 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14839 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14840 that moved to I2, also allow destinations that are a paradoxical
14841 subreg (instead of a normal reg).
14843 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14846 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14849 2018-02-16 Richard Biener <rguenther@suse.de>
14851 PR tree-optimization/84037
14852 PR tree-optimization/84016
14854 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14855 Adjust vec_construct for the fact we need additional higher latency
14856 128bit inserts for AVX256 and AVX512 vector builds.
14857 (ix86_add_stmt_cost): Scale vector construction cost for
14860 2018-02-16 Richard Biener <rguenther@suse.de>
14862 PR tree-optimization/84417
14863 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14864 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14865 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14867 2018-02-16 Martin Liska <mliska@suse.cz>
14870 * internal-fn.def (ASAN_CHECK): Set proper flags.
14871 (ASAN_MARK): Likewise.
14873 2018-02-16 Julia Koval <julia.koval@intel.com>
14875 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14876 from PTA_CANNONLAKE.
14878 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14881 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14882 Use ++iter rather than iter++ for std::list iterators.
14883 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14884 defer deleting them until all nodes in the forest are processed. Do
14885 free even leaf nodes. Change to_process into auto_vec.
14888 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14889 * vec.h (vec_default_construct): Use memset instead of placement new
14890 if BROKEN_VALUE_INITIALIZATION is defined.
14891 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14892 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14895 PR rtl-optimization/83723
14896 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14897 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14898 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14900 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14902 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14904 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14906 PR rtl-optimization/81443
14907 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14908 from inner REGs to paradoxical SUBREGs.
14910 2018-02-16 Richard Biener <rguenther@suse.de>
14912 PR tree-optimization/84399
14913 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14914 For operands we can analyze at their definition make sure we can
14915 analyze them at each use as well.
14917 2018-02-16 Richard Biener <rguenther@suse.de>
14919 PR tree-optimization/84190
14920 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14921 volatile accesses if the decl isn't volatile.
14923 2018-02-15 Jason Merrill <jason@redhat.com>
14925 PR c++/84314 - ICE with templates and fastcall attribute.
14926 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14927 TYPE_CANONICAL on an existing type.
14929 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14931 PR tree-optimization/84383
14932 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14933 dstoff nor call operand_equal_p if dstbase is NULL.
14935 PR tree-optimization/84334
14936 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14937 also a CONSTANT_CLASS_P, punt.
14939 2018-02-14 Jim Wilson <jimw@sifive.com>
14941 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14942 first SMALL_OPERAND check. New local min_second_step. Move assert
14943 to where locals are set. Add TARGET_RVC support.
14944 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14946 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14948 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14950 2018-02-14 Martin Sebor <msebor@redhat.com>
14952 PR tree-optimization/83698
14953 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14954 arrays constrain the offset range to their bounds.
14955 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14956 (builtin_access::overlap): Avoid setting the size of overlap if it's
14958 (maybe_diag_overlap): Also consider arrays when deciding what values
14959 of offsets to include in diagnostics.
14961 2018-02-14 Martin Sebor <msebor@redhat.com>
14964 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14965 that correspond to the kind of a declaration.
14967 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14970 * config/pa/pa.md: Load address of PIC label using the linkage table
14971 if the label is nonlocal.
14973 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14976 warning message if user requests -maltivec=be.
14977 * doc/invoke.texi: Document deprecation of -maltivec=be.
14979 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14982 * config/rs6000/rs6000-c.c: Update definitions for
14983 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14984 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14986 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14989 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14990 add _get_ssp intrinsics. Remove argument from
14991 __builtin_ia32_rdssp[d|q].
14992 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14993 * config/i386/i386-builtin.def: Remove argument from
14994 __builtin_ia32_rdssp[d|q].
14995 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14996 ix86_expand_special_args_builtin for _rdssp[d|q].
14997 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14998 Clear register before usage.
14999 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
15000 Add documentation for new _get_ssp and _inc_ssp intrinsics.
15002 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
15004 PR tree-optimization/84357
15005 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
15006 operand 1 of an ARRAY_REF too.
15008 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
15011 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
15012 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
15014 (set_of_reg): New struct.
15015 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
15016 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
15017 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
15019 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
15020 Split into bitclr, bitset, bitinvert patterns if appropriate.
15021 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
15022 use rx_fuse_in_memory_bitop.
15023 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
15024 to named insn, correct maximum insn length.
15026 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
15029 * machmode.def: Define a complex mode for PARTIAL_INT.
15030 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
15032 * doc/rtl.texi: Document CSPImode.
15033 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
15035 (msp430_hard_regno_nregs_with_padding): Likewise.
15037 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
15040 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
15042 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
15044 PR rtl-optimization/84169
15045 * combine.c (try_combine): New variable split_i2i3. Set it to true if
15046 we generated a parallel as new i3 and we split that to new i2 and i3
15047 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
15048 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
15049 those to i2, not i1. Partially rewrite this scan code.
15051 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15054 * stor-layout.c (place_field): For variable length fields, adjust
15055 offset_align afterwards not just based on the field's alignment,
15056 but also on the size.
15058 PR middle-end/84309
15059 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
15060 of exps and logs in the use_exp2 case.
15062 2018-02-13 Jeff Law <law@redhat.com>
15064 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
15065 entry for "vector".
15067 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
15070 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
15074 * common.opt (gas-loc-support, gas-locview-support): New.
15075 (ginline-points, ginternal-reset-location-views): New.
15076 * doc/invoke.texi: Document them. Use @itemx where intended.
15077 (gvariable-location-views): Adjust.
15078 * target.def (reset_location_view): New.
15079 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
15080 (TARGET_RESET_LOCATION_VIEW): New.
15081 * doc/tm.texi: Rebuilt.
15082 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
15083 (dwarf2out_default_as_locview_support): New.
15084 (output_asm_line_debug_info): Use option variables.
15085 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
15086 (output_loc_list): Likewise.
15087 (add_high_low_attributes): Check option variables.
15088 Don't output entry view attribute in strict mode.
15089 (gen_inlined_subroutine_die): Check option variables.
15090 (dwarf2out_inline_entry): Likewise.
15091 (init_sections_and_labels): Likewise.
15092 (dwarf2out_early_finish): Likewise.
15093 (maybe_reset_location_view): New, from...
15094 (dwarf2out_var_location): ... here. Call it.
15095 * debug.h (dwarf2out_default_as_loc_support): Declare.
15096 (dwarf2out_default_as_locview_support): Declare.
15097 * hooks.c (hook_int_rtx_insn_0): New.
15098 * hooks.h (hook_int_rtx_insn_0): Declare.
15099 * toplev.c (process_options): Take -gas-loc-support and
15100 -gas-locview-support from dwarf2out. Enable
15101 -gvariable-location-views by default only with locview
15102 assembler support. Enable -ginternal-reset-location-views by
15103 default only if the target defines the corresponding hook.
15104 Enable -ginline-points by default if location views are
15105 enabled; force it disabled if statement frontiers are
15107 * tree-inline.c (expand_call_inline): Check option variables.
15108 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
15110 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
15112 PR tree-optimization/84321
15113 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
15114 handling. Also check whether the anti-range contains any values
15115 that satisfy the mask; switch to a VR_RANGE if not.
15117 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
15120 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
15122 2018-02-13 Martin Jambor <mjambor@suse.cz>
15125 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
15126 of call statements, also set location of a load to a temporary.
15128 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
15130 * config/rl78/rl78.c (add_vector_labels): New function.
15131 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
15132 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
15133 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
15134 which checks that no arguments are passed.
15135 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
15136 * doc/extend.texi: Documentation for the new attribute.
15138 2018-02-13 Andreas Schwab <schwab@suse.de>
15140 * config/riscv/linux.h (CPP_SPEC): Define.
15142 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15145 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
15146 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
15147 OPTION_MASK_ISA_AES as first argument to def_builtin_const
15148 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
15149 instead of OPTION_MASK_ISA_PCLMUL as first argument to
15150 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
15151 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
15152 temporarily for AES and PCLMUL builtins.
15154 PR tree-optimization/84339
15155 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
15156 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
15159 PR middle-end/84309
15160 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
15161 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
15162 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
15164 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
15166 * omp-simd-clone.h: New file.
15167 * omp-simd-clone.c: Include omp-simd-clone.h.
15168 (expand_simd_clones): No longer static.
15169 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
15170 cgraph.h and omp-simd-clone.h.
15171 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
15172 (vect_recog_widen_shift_pattern): Formatting fix.
15173 (vect_pattern_recog_1): Don't check optab for calls.
15176 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
15177 operands[2] into a REG before using gen_lowpart on it.
15179 2018-02-12 Jeff Law <law@redhat.com>
15182 * config/sh/sh.c (find_barrier): Consider a sibling call
15185 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
15186 successfully back substituting a reg.
15188 2018-02-12 Richard Biener <rguenther@suse.de>
15190 PR tree-optimization/84037
15191 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
15192 parameter, move visited init to caller.
15193 (vect_slp_analyze_operations): Separate cost from validity
15194 check, initialize visited once for all instances.
15195 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
15197 * tree-vect-stmts.c (vect_model_simple_cost): Make early
15199 (vect_model_promotion_demotion_cost): Likewise.
15200 (vectorizable_bswap): Guard cost modeling with !slp_node
15201 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
15203 (vectorizable_call): Likewise.
15204 (vectorizable_conversion): Likewise.
15205 (vectorizable_assignment): Likewise.
15206 (vectorizable_shift): Likewise.
15207 (vectorizable_operation): Likewise.
15208 (vectorizable_store): Likewise.
15209 (vectorizable_load): Likewise.
15210 (vectorizable_condition): Likewise.
15211 (vectorizable_comparison): Likewise.
15213 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
15216 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
15217 (ASAN_MARK): Fix fnspec to account for return value, change pointer
15218 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
15220 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15222 PR middle-end/83665
15223 * params.def (inline-min-speedup): Increase from 8 to 15.
15224 (max-inline-insns-auto): Decrease from 40 to 30.
15225 * ipa-split.c (consider_split): Add some buffer for function to
15226 be considered inlining candidate.
15227 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
15230 2018-02-12 Richard Biener <rguenther@suse.de>
15232 PR tree-optimization/84037
15233 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
15234 matched stmts if we cannot swap the non-matched ones.
15236 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15238 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
15239 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
15240 _mm_maskz_scalef_round_ss): New intrinsics.
15241 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
15242 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
15243 __builtin_ia32_scalefss_round): Remove.
15244 (__builtin_ia32_scalefsd_mask_round,
15245 __builtin_ia32_scalefss_mask_round): New intrinsics.
15246 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
15247 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
15248 ((match_operand:VF_128 2 "<round_nimm_predicate>"
15249 "<round_constraint>")): Changed to ...
15250 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
15251 "<round_scalar_constraint>")): ... this.
15252 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
15253 %0, %1, %2<round_op3>}"): Changed to ...
15254 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
15255 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
15256 %2<round_scalar_mask_op3>}"): ... this.
15257 * config/i386/subst.md (round_scalar_nimm_predicate): New.
15259 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15261 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
15262 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
15263 (_mm_maskz_sqrt_round_ss): New intrinsics.
15264 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
15265 (__builtin_ia32_sqrtsd_mask_round)
15266 (__builtin_ia32_sqrtss_mask_round): New builtins.
15267 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
15268 (__builtin_ia32_sqrtss_round): Remove.
15269 (__builtin_ia32_sqrtsd_mask_round)
15270 (__builtin_ia32_sqrtss_mask_round): New builtins.
15271 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
15272 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
15273 ((match_operand:VF_128 1 "vector_operand"
15274 "xBm,<round_constraint>")): Changed to ...
15275 ((match_operand:VF_128 1 "vector_operand"
15276 "xBm,<round_scalar_constraint>")): ... this.
15277 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
15278 %0, %2, %<iptr>1<round_op3>}): Changed to ...
15279 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
15280 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
15281 %<iptr>1<round_scalar_mask_op3>}): ... this.
15282 ((set_attr "prefix" "<round_prefix>")): Changed to ...
15283 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
15285 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
15288 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
15289 Cast vec_cmpeq result to correct type.
15290 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
15291 Cast vec_cmpgt result to correct type.
15293 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
15295 * final.c (final_scan_insn_1): Renamed from...
15296 (final_scan_insn): ... this. New wrapper, to recover
15297 seen from the outermost call in recursive ones.
15298 * config/sparc/sparc.c (output_return): Drop seen from call.
15299 (output_sibcall): Likewise.
15300 * config/visium/visium.c (output_branch): Likewise.
15302 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
15304 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
15307 2018-02-10 Alan Modra <amodra@gmail.com>
15310 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
15311 Specify LR as an input.
15313 2018-02-10 Jakub Jelinek <jakub@redhat.com>
15316 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
15317 remove_member_access_dummy_vars): New functions.
15318 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
15319 lower_omp_1, execute_lower_omp): Use them.
15321 PR rtl-optimization/84308
15322 * shrink-wrap.c (spread_components): Release todo vector.
15324 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
15326 PR rtl-optimization/57193
15327 * ira-color.c (struct allocno_color_data): Add member
15328 conflict_allocno_hard_prefs.
15329 (update_conflict_allocno_hard_prefs): New.
15330 (bucket_allocno_compare_func): Add a preference based on
15331 conflict_allocno_hard_prefs.
15332 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
15333 (color_allocnos): Remove a dead code. Initiate
15334 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
15336 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15339 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
15340 constraint from =wa to wa. Avoid a subreg on the output operand,
15341 instead use a pseudo and subreg it in a move.
15342 (p9_xxbrd_<mode>): Changed to ...
15343 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
15344 (p9_xxbrd_v2df): New expander.
15345 (p9_xxbrw_<mode>): Changed to ...
15346 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
15347 (p9_xxbrw_v4sf): New expander.
15349 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15351 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
15353 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
15356 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
15357 multiply in 32-bit mode.
15358 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
15359 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
15362 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15364 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
15365 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
15366 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
15367 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
15369 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
15372 * dwarf2out.c (is_trivial_indirect_ref): New function.
15373 (dwarf2out_late_global_decl): Do not generate a location
15374 attribute for variables that have a non-trivial DECL_VALUE_EXPR
15375 and that are not defined in the current unit.
15377 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15379 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
15380 instead of a libcall for UNORDERED.
15382 2018-02-09 Tamar Christina <tamar.christina@arm.com>
15385 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
15386 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
15388 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15391 * config/s390/s390.c (s390_set_current_function): Invoke
15392 s390_indirect_branch_settings also if fndecl didn't change.
15394 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15396 * config/rs6000/rs6000.md (blockage): Set length to zero.
15398 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15400 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
15402 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15405 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
15406 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
15410 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
15411 PARALLEL incoming that failed vt_get_decl_and_offset check.
15413 PR middle-end/84237
15414 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
15415 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
15417 (get_variable_section): For decls in named .bss* sections pass true as
15418 second argument to bss_initializer_p.
15420 2018-02-09 Marek Polacek <polacek@redhat.com>
15421 Jakub Jelinek <jakub@redhat.com>
15424 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
15425 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
15426 Sync some changes from cxx_fold_indirect_ref.
15428 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15430 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
15432 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
15433 (BLOCK_INLINE_ENTRY_LABEL): New.
15434 (dwarf2out_var_location): Disregard inline entry markers.
15435 (inline_entry_data): New struct.
15436 (inline_entry_data_hasher): New hashtable type.
15437 (inline_entry_data_hasher::hash): New.
15438 (inline_entry_data_hasher::equal): New.
15439 (inline_entry_data_table): New variable.
15440 (add_high_low_attributes): Add DW_AT_entry_pc and
15441 DW_AT_GNU_entry_view attributes if a pending entry is found
15442 in inline_entry_data_table. Add old entry_pc attribute only
15443 if debug nonbinding markers are disabled.
15444 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
15445 markers are enabled.
15446 (block_within_block_p, dwarf2out_inline_entry): New.
15447 (dwarf2out_finish): Check that no entries remained in
15448 inline_entry_data_table.
15449 * final.c (reemit_insn_block_notes): Handle inline entry notes.
15450 (final_scan_insn, notice_source_line): Likewise.
15451 (rest_of_clean_state): Skip inline entry markers.
15452 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
15454 * gimple.c (gimple_build_debug_inline_entry): New.
15455 * gimple.h (enum gimple_debug_subcode): Add
15456 GIMPLE_DEBUG_INLINE_ENTRY.
15457 (gimple_build_debug_inline_entry): Declare.
15458 (gimple_debug_inline_entry_p): New.
15459 (gimple_debug_nonbind_marker_p): Adjust.
15460 * insn-notes.def (INLINE_ENTRY): New.
15461 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
15462 inline entry marker notes.
15463 (print_insn): Likewise.
15464 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
15465 (INSN_DEBUG_MARKER_KIND): Likewise.
15466 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
15467 * tree-inline.c (expand_call_inline): Build and insert
15468 debug_inline_entry stmt.
15469 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
15470 inline entry blocks early, if nonbind markers are enabled.
15471 (dump_scope_block): Dump fragment info.
15472 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
15473 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
15474 (gimple_build_debug_inline_entry): New.
15475 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
15476 Enable/disable inline entry points too.
15477 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
15478 (DEBUG_INSN): Describe inline entry markers.
15480 * common.opt (gvariable-location-views): New.
15481 (gvariable-location-views=incompat5): New.
15482 * config.in: Rebuilt.
15483 * configure: Rebuilt.
15484 * configure.ac: Test assembler for view support.
15485 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
15486 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
15487 * dwarf2out.c (var_loc_view): New typedef.
15488 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
15489 (dwarf2out_locviews_in_attribute): New.
15490 (dwarf2out_locviews_in_loclist): New.
15491 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
15492 (enum dw_line_info_opcode): Add LI_adv_address.
15493 (struct dw_line_info_table): Add view.
15494 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
15495 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
15496 (zero_view_p): New variable.
15497 (ZERO_VIEW_P): New macro.
15498 (output_asm_line_debug_info): New.
15499 (struct var_loc_node): Add view.
15500 (add_AT_view_list, AT_loc_list): New.
15501 (add_var_loc_to_decl): Add view param. Test it against last.
15502 (new_loc_list): Add view params. Record them.
15503 (AT_loc_list_ptr): Handle loc and view lists.
15504 (view_list_to_loc_list_val_node): New.
15505 (print_dw_val): Handle dw_val_class_view_list.
15506 (size_of_die): Likewise.
15507 (value_format): Likewise.
15508 (loc_list_has_views): New.
15509 (gen_llsym): Set vl_symbol too.
15510 (maybe_gen_llsym, skip_loc_list_entry): New.
15511 (dwarf2out_maybe_output_loclist_view_pair): New.
15512 (output_loc_list): Output view list or entries too.
15513 (output_view_list_offset): New.
15514 (output_die): Handle dw_val_class_view_list.
15515 (output_dwarf_version): New.
15516 (output_compilation_unit_header): Use it.
15517 (output_skeleton_debug_sections): Likewise.
15518 (output_rnglists, output_line_info): Likewise.
15519 (output_pubnames, output_aranges): Update version comments.
15520 (output_one_line_info_table): Output view numbers in asm comments.
15521 (dw_loc_list): Determine current endview, pass it to new_loc_list.
15522 Call maybe_gen_llsym.
15523 (loc_list_from_tree_1): Adjust.
15524 (add_AT_location_description): Create view list attribute if
15525 needed, check it's absent otherwise.
15526 (convert_cfa_to_fb_loc_list): Adjust.
15527 (maybe_emit_file): Call output_asm_line_debug_info for test.
15528 (dwarf2out_var_location): Reset views as needed. Precompute
15529 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
15530 attribute. Set view.
15531 (new_line_info_table): Reset next view.
15532 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
15533 (dwarf2out_source_line): Likewise. Output view resets and labels to
15534 the assembler, or select appropriate line info opcodes.
15535 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
15536 (optimize_string_length): Catch it. Adjust.
15537 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
15538 dw_val_class_view_list, and remove it if no longer needed.
15539 (hash_loc_list): Hash view numbers.
15540 (loc_list_hasher::equal): Compare them.
15541 (optimize_location_lists): Check whether a view list symbol is
15542 needed, and whether the locview attribute is present, and
15543 whether they match. Remove the locview attribute if no longer
15545 (index_location_lists): Call skip_loc_list_entry for test.
15546 (dwarf2out_finish): Call output_asm_line_debug_info for test.
15547 Use output_dwarf_version.
15548 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
15549 (struct dw_val_node): Add val_view_list.
15550 * final.c (SEEN_NEXT_VIEW): New.
15551 (set_next_view_needed): New.
15552 (clear_next_view_needed): New.
15553 (maybe_output_next_view): New.
15554 (final_start_function): Rename to...
15555 (final_start_function_1): ... this. Take pointer to FIRST,
15556 add SEEN parameter. Emit param bindings in the initial view.
15557 (final_start_function): Reintroduce SEEN-less interface.
15558 (final): Rename to...
15559 (final_1): ... this. Take SEEN parameter. Output final pending
15560 next view at the end.
15561 (final): Reintroduce seen-less interface.
15562 (final_scan_insn): Output pending next view before switching
15563 sections or ending a block. Mark the next view as needed when
15564 outputting variable locations. Notify debug backend of section
15565 changes, and of location view changes.
15566 (rest_of_handle_final): Adjust.
15567 * toplev.c (process_options): Autodetect value for debug variable
15568 location views option. Warn on incompat5 without -gdwarf-5.
15569 * doc/invoke.texi (gvariable-location-views): New.
15570 (gvariable-location-views=incompat5): New.
15571 (gno-variable-location-views): New.
15573 2018-02-08 David Malcolm <dmalcolm@redhat.com>
15575 PR tree-optimization/84136
15576 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
15577 that the result of find_edge is non-NULL.
15579 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15582 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
15583 storing integer register in SImode. Fix cost of 256 and 512
15584 byte aligned SSE register store.
15586 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15588 * config/i386/i386.c (ix86_multiplication_cost): Fix
15589 multiplication cost for TARGET_AVX512DQ.
15591 2018-02-08 Marek Polacek <polacek@redhat.com>
15593 PR tree-optimization/84238
15594 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
15597 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15599 PR tree-optimization/84265
15600 * tree-vect-stmts.c (vectorizable_store): Don't treat
15601 VMAT_CONTIGUOUS accesses as grouped.
15602 (vectorizable_load): Likewise.
15604 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15606 PR tree-optimization/81635
15607 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
15608 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
15609 (test_round_for_mask): New functions.
15610 (wide_int_cc_tests): Call test_round_for_mask.
15611 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
15612 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
15613 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
15614 range returned by get_range_info.
15616 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15619 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
15620 * symtab.c: Include builtins.h
15621 (symtab_node::output_to_lto_symbol_table_p): Move here
15622 from lto-streamer-out.c:output_symbol_p.
15623 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
15624 (output_symbol_p): Move all logic to symtab.c
15625 (produce_symtab): Update.
15627 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15629 * config/s390/s390-opts.h (enum indirect_branch): Define.
15630 * config/s390/s390-protos.h (s390_return_addr_from_memory)
15631 (s390_indirect_branch_via_thunk)
15632 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
15633 (enum s390_indirect_branch_type): Define.
15634 * config/s390/s390.c (struct s390_frame_layout, struct
15635 machine_function): Remove.
15636 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
15637 (indirect_branch_table_label_no, indirect_branch_table_name):
15639 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
15640 (enum s390_indirect_branch_option): Define.
15641 (s390_return_addr_from_memory): New function.
15642 (s390_handle_string_attribute): New function.
15643 (s390_attribute_table): Add new attribute handler.
15644 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
15645 (s390_indirect_branch_via_thunk): New function.
15646 (s390_indirect_branch_via_inline_thunk): New function.
15647 (s390_function_ok_for_sibcall): When jumping via thunk disallow
15648 sibling call optimization for non z10 compiles.
15649 (s390_emit_call): Force indirect branch target to be a single
15650 register. Add r1 clobber for non-z10 compiles.
15651 (s390_emit_epilogue): Emit return jump via return_use expander.
15652 (s390_reorg): Handle JUMP_INSNs as execute targets.
15653 (s390_option_override_internal): Perform validity checks for the
15654 new command line options.
15655 (s390_indirect_branch_attrvalue): New function.
15656 (s390_indirect_branch_settings): New function.
15657 (s390_set_current_function): Invoke s390_indirect_branch_settings.
15658 (s390_output_indirect_thunk_function): New function.
15659 (s390_code_end): Implement target hook.
15660 (s390_case_values_threshold): Implement target hook.
15661 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
15663 * config/s390/s390.h (struct s390_frame_layout)
15664 (struct machine_function): Move here from s390.c.
15665 (TARGET_INDIRECT_BRANCH_NOBP_RET)
15666 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
15667 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
15668 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
15669 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
15670 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
15671 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
15672 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
15673 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
15674 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
15675 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
15676 (mnemonic attribute): Add values which aren't recognized
15678 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
15679 pattern for branch conversion. Fix mnemonic attribute.
15680 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
15681 indirect branch via thunk if requested.
15682 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
15683 ("*indirect_jump"): Disable for branch conversion using out of
15685 ("indirect_jump_via_thunk<mode>_z10")
15686 ("indirect_jump_via_thunk<mode>")
15687 ("indirect_jump_via_inlinethunk<mode>_z10")
15688 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
15689 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
15690 ("casesi_jump_via_inlinethunk<mode>_z10")
15691 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
15692 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
15693 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
15694 ("*indirect2_jump"): Disable for branch conversion.
15695 ("casesi_jump"): Turn into expander and expand patterns for branch
15697 ("return_use"): New expander.
15698 ("*return"): Emit return via thunk and rename it to ...
15699 ("*return<mode>"): ... this one.
15700 * config/s390/s390.opt: Add new options and and enum for the
15703 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15705 * lra-constraints.c (match_reload): Unconditionally use
15706 gen_lowpart_SUBREG, rather than selecting between that
15707 and equivalent gen_rtx_SUBREG code.
15709 2018-02-08 Richard Biener <rguenther@suse.de>
15711 PR tree-optimization/84233
15712 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
15713 changed flag instead of boguously re-using phi_inserted.
15715 2018-02-08 Martin Jambor <mjambor@suse.cz>
15717 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
15718 static local variables.
15720 2018-02-08 Richard Biener <rguenther@suse.de>
15722 PR tree-optimization/84278
15723 * tree-vect-stmts.c (vectorizable_store): When looking for
15724 smaller vector types to perform grouped strided loads/stores
15725 make sure the mode is supported by the target.
15726 (vectorizable_load): Likewise.
15728 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15730 * config/aarch64/aarch64.c (aarch64_components_for_bb):
15731 Increase LDP/STP opportunities by adding adjacent callee-saves.
15733 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15735 PR rtl-optimization/84068
15736 PR rtl-optimization/83459
15737 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
15739 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
15741 PR tree-optimization/84224
15742 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
15743 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
15744 non-zero arguments.
15746 2018-02-07 Iain Sandoe <iain@codesourcery.com>
15749 * config/rs6000/altivec.md (*restore_world): Remove LR use.
15750 * config/rs6000/predicates.md (restore_world_operation): Adjust op
15751 count, remove one USE.
15753 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
15755 * doc/install.texi (Configuration): Document the
15756 --with-long-double-format={ibm,ieee} PowerPC configuration
15760 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
15761 Convert from define_expand to be define_insn_and_split. Rework
15762 float/double/_Float128 conversions to QI/HI/SImode to work with
15763 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
15764 conversions to QI/HImode types did a store and then a load to
15765 truncate the value. For conversions to VSX registers, don't split
15766 the insn, instead emit the code directly. Use the code iterator
15767 any_fix to combine signed and unsigned conversions.
15768 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15769 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15770 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15771 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15772 (fix_<mode>di2_hw): Likewise.
15773 (fixuns_<mode>di2_hw): Likewise.
15774 (fix_<mode>si2_hw): Likewise.
15775 (fixuns_<mode>si2_hw): Likewise.
15776 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15777 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15778 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15779 fix<uns>_trunc<SFDF:mode>si2_p8.
15780 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15781 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15782 (fix<uns>_<mode>_mem): Likewise.
15783 (fctiw<u>z_<mode>_mem): Likewise.
15784 (fix<uns>_<mode>_mem): Likewise.
15785 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15786 the register allocator from doing a direct move to the GPRs to do
15787 a store, and instead use the ISA 3.0 store byte/half-word from
15788 vector register instruction. For IEEE 128-bit floating point,
15789 also optimize stores of 32-bit ints.
15790 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15792 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15794 * genextract.c (push_pathstr_operand): New function to support
15796 (walk_rtx): Call push_pathstr_operand.
15797 (print_path): Support [a-zA-Z].
15799 2018-02-07 Richard Biener <rguenther@suse.de>
15801 PR tree-optimization/84037
15802 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15803 (cse_and_gimplify_to_preheader): Declare.
15804 (vect_get_place_in_interleaving_chain): Likewise.
15805 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15807 (_loop_vec_info::~_loop_vec_info): Delete it.
15808 (cse_and_gimplify_to_preheader): New function.
15809 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15810 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15811 (vectorizable_load): Likewise. For grouped stores always base
15812 the IV on the first element.
15813 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15814 condition before gimplifying.
15816 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15818 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15819 *DIV_EXPR and *MOD_EXPR.
15821 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15824 * config/i386/i386.c (ix86_option_override_internal): Mask out
15825 the CF_SET bit when checking -fcf-protection.
15827 2018-02-07 Tom de Vries <tom@codesourcery.com>
15830 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15833 2018-02-07 Richard Biener <rguenther@suse.de>
15835 PR tree-optimization/84204
15836 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15839 PR tree-optimization/84205
15840 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15841 special-case isl_ast_op_zdiv_r.
15843 PR tree-optimization/84223
15844 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15845 Only add conditions from within the region.
15846 (gather_bbs::after_dom_children): Adjust.
15848 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15851 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15852 * config/avr/avr.md: Only post-reload split REG-REG moves if
15853 either register is GENERAL_REG_P.
15855 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15857 PR tree-optimization/84235
15858 * tree-ssa-scopedtables.c
15859 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15860 if the subtraction is performed in floating point type where NaNs are
15861 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15862 build 1. Formatting fix.
15864 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15867 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15868 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15869 and skip it regardless of bb boundaries. Use CALL_P macro,
15870 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15871 unnecessarily, formatting fix.
15873 2018-02-06 Michael Collison <michael.collison@arm.com>
15875 * config/arm/thumb2.md:
15876 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15877 (*thumb_mov_notscc): Ditto.
15879 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15882 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15883 unsigned_fix, not "s".
15885 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15887 * configure.ac (gcc_fn_eh_frame_ro): New function.
15888 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15889 correct .eh_frame permissions.
15890 * configure: Regenerate.
15892 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15894 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15895 irrelevant options.
15897 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15899 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15900 Display warning message for -mno-speculate-indirect-jumps.
15902 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15904 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15906 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15908 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15910 PR tree-optimization/84225
15911 * tree-eh.c (find_trapping_overflow): Only call
15912 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15914 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15917 * config/i386/i386.c: Reimplement the check of possible options
15918 -mibt/-mshstk conbination. Change error messages.
15919 * doc/invoke.texi: Fix a typo: remove extra '='.
15921 2018-02-06 Marek Polacek <polacek@redhat.com>
15923 PR tree-optimization/84228
15924 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15926 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15929 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15930 emitted arch directives.
15931 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15932 __ARM_FEATURE_COPROC before changing architectures.
15934 2018-02-06 Richard Biener <rguenther@suse.de>
15936 * config/i386/i386.c (print_reg): Fix typo.
15937 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15939 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15941 * configure: Regenerate.
15943 2018-02-05 Martin Sebor <msebor@redhat.com>
15945 PR tree-optimization/83369
15946 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15949 2018-02-05 Martin Liska <mliska@suse.cz>
15951 * doc/invoke.texi: Cherry-pick upstream r323995.
15953 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15955 * ira.c (ira_init_register_move_cost): Adjust comment.
15957 2018-02-05 Martin Liska <mliska@suse.cz>
15959 PR gcov-profile/84137
15960 * doc/gcov.texi: Fix typo in documentation.
15962 2018-02-05 Martin Liska <mliska@suse.cz>
15964 PR gcov-profile/83879
15965 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15966 using dlopen functionality.
15968 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15970 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15971 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15972 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15973 _mm_maskz_range_round_ss): New intrinsics.
15974 (__builtin_ia32_rangesd128_round)
15975 (__builtin_ia32_rangess128_round): Remove.
15976 (__builtin_ia32_rangesd128_mask_round,
15977 __builtin_ia32_rangess128_mask_round): New builtins.
15978 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15979 __builtin_ia32_rangess128_round): Remove.
15980 (__builtin_ia32_rangesd128_mask_round,
15981 __builtin_ia32_rangess128_mask_round): New builtins.
15982 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15983 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15984 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15985 "<round_saeonly_constraint>")): Changed to ...
15986 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15987 "<round_saeonly_scalar_constraint>")): ... this.
15988 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15989 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15990 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15991 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15992 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15994 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15996 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15998 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15999 Remove all values except native, 8540 and 8548.
16001 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
16003 * config/i386/i386.c (ix86_output_function_return): Pass
16004 INVALID_REGNUM, instead of -1, as invalid register number to
16005 indirect_thunk_name and output_indirect_thunk.
16007 2018-02-02 Julia Koval <julia.koval@intel.com>
16009 * config.gcc: Add -march=icelake.
16010 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
16011 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
16012 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
16013 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
16014 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
16015 (processor_target_table): Add icelake.
16016 (ix86_option_override_internal): Handle new PTAs.
16017 (get_builtin_code_for_version): Handle icelake.
16018 (M_INTEL_COREI7_ICELAKE): New.
16019 (fold_builtin_cpu): Handle icelake.
16020 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
16021 * doc/invoke.texi: Add -march=icelake.
16023 2018-02-02 Julia Koval <julia.koval@intel.com>
16025 * config/i386/i386.c (ix86_option_override_internal): Change flags type
16026 to wide_int_bitmask.
16027 * wide-int-bitmask.h: New.
16029 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16032 * config/i386/i386.md: Replace Pmode with word_mode in
16033 builtin_setjmp_setup and builtin_longjmp to support x32.
16035 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
16039 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
16041 (rs6000_supported_cpu_names): New static variable.
16042 (linux_cpu_translation_table): Likewise.
16043 (elf_platform) <cpu>: Define new static variable and use it.
16044 Translate kernel AT_PLATFORM name to canonical name if needed.
16045 Error if platform name is unknown.
16047 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
16050 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
16052 2018-02-01 Jeff Law <law@redhat.com>
16055 * config/i386/i386.c (release_scratch_register_on_entry): Add new
16056 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
16057 the scratch if RELEASE_VIA_POP is false.
16058 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
16059 If we have to save a temporary register, decrement SIZE appropriately.
16060 Pass new arguments to release_scratch_register_on_entry.
16061 (ix86_adjust_stack_and_probe): Likewise.
16062 (ix86_emit_probe_stack_range): Pass new arguments to
16063 release_scratch_register_on_entry.
16065 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
16067 PR rtl-optimization/84157
16068 * combine.c (change_zero_ext): Use REG_P predicate in
16069 front of HARD_REGISTER_P predicate.
16071 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
16073 * config/avr/avr.c (avr_option_override): Move disabling of
16074 -fdelete-null-pointer-checks to...
16075 * common/config/avr/avr-common.c (avr_option_optimization_table):
16078 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16080 PR tree-optimization/81635
16081 * tree-data-ref.c (split_constant_offset_1): For types that
16082 wrap on overflow, try to use range info to prove that wrapping
16085 2018-02-01 Renlin Li <renlin.li@arm.com>
16088 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
16089 TAILCALL_ADDR_REGS.
16090 (aarch64_register_move_cost): Likewise.
16091 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
16092 TAILCALL_ADDR_REGS.
16093 (REG_CLASS_NAMES): Likewise.
16094 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
16095 TAILCALL_ADDR_REGS. Remove IP registers.
16096 * config/aarch64/aarch64.md (Ucs): Update register constraint.
16098 2018-02-01 Richard Biener <rguenther@suse.de>
16100 * domwalk.h (dom_walker::dom_walker): Add additional constructor
16101 for specifying RPO order and allow NULL for that.
16102 * domwalk.c (dom_walker::dom_walker): Likewise.
16103 (dom_walker::walk): Handle NULL RPO order.
16104 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
16106 (rewrite_update_dom_walker): Likewise.
16107 (mark_def_dom_walker): Likewise.
16109 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16111 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
16112 (aarch64_maybe_expand_sve_subreg_move): Declare.
16113 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
16114 * config/aarch64/predicates.md (aarch64_any_register_operand): New
16116 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
16117 that are semantically a reverse operation.
16118 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
16119 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
16120 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
16122 (aarch64_can_change_mode_class): For big-endian, forbid changes
16123 between two SVE modes if they have different element sizes.
16125 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16127 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
16128 the TImode handling for big-endian targets.
16130 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16132 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
16133 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
16135 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
16136 Remove BSWAP handing for big-endian targets and use the form of
16137 LD1RQ appropariate for the mode.
16139 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16141 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
16142 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
16143 duplicated element.
16145 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16148 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
16149 check for operands that need to go through aarch64_sve_reload_be.
16151 2018-02-01 Jakub Jelinek <jakub@redhat.com>
16153 PR tree-optimization/81661
16154 PR tree-optimization/84117
16155 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
16156 * tree-eh.c: Include gimplify.h.
16157 (find_trapping_overflow, replace_trapping_overflow,
16158 rewrite_to_non_trapping_overflow): New functions.
16159 * tree-vect-loop.c: Include tree-eh.h.
16160 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
16161 * tree-data-ref.c: Include tree-eh.h.
16162 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
16164 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
16166 PR rtl-optimization/84123
16167 * combine.c (change_zero_ext): Check if hard register satisfies
16168 can_change_dest_mode before calling gen_lowpart_SUBREG.
16170 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
16173 * ira.c (ira_init_register_move_cost): Remove assert.
16175 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16177 PR rtl-optimization/84071
16178 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
16179 * doc/tm.texi: Regenerate.
16181 2018-01-31 Richard Biener <rguenther@suse.de>
16183 PR tree-optimization/84132
16184 * tree-data-ref.c (analyze_miv_subscript): Properly
16185 check whether evolution_function_is_affine_multivariate_p
16186 before calling gcd_of_steps_may_divide_p.
16188 2018-01-31 Julia Koval <julia.koval@intel.com>
16191 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
16192 * config/i386/i386.md (rdpid_rex64) New.
16193 (rdpid): Make 32bit only.
16195 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
16198 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
16199 an IDENTIFIER_NODE for FUNCTION_TYPE's.
16201 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16204 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16206 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16208 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16210 PR rtl-optimization/84071
16211 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
16212 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
16214 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16216 * config/arc/arc.c (arc_handle_aux_attribute): New function.
16217 (arc_attribute_table): Add 'aux' attribute.
16218 (arc_in_small_data_p): Consider aux like variables.
16219 (arc_is_aux_reg_p): New function.
16220 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
16221 (arc_get_aux_arg): New function.
16222 (prepare_move_operands): Handle aux-register access.
16223 (arc_handle_aux_attribute): New function.
16224 * doc/extend.texi (ARC Variable attributes): Add subsection.
16226 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16228 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
16229 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
16230 (arc_attribute_table): Add 'uncached' attribute.
16231 (arc_print_operand): Print '.di' flag for uncached memory
16233 (arc_in_small_data_p): Do not consider for small data the uncached
16235 (arc_is_uncached_mem_p): New function.
16236 * config/arc/predicates.md (compact_store_memory_operand): Check
16237 for uncached memory accesses.
16238 (nonvol_nonimm_operand): Likewise.
16239 * doc/extend.texi (ARC Type Attribute): New subsection.
16241 2018-01-31 Jakub Jelinek <jakub@redhat.com>
16244 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
16245 falign-loops=): Add Optimization flag.
16247 2018-01-30 Jeff Law <law@redhat.com>
16250 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
16251 INT_REGISTERS_SAVED. Check it prior to calling
16252 get_scratch_register_on_entry.
16253 (ix86_adjust_stack_and_probe): Similarly.
16254 (ix86_emit_probe_stack_range): Similarly.
16255 (ix86_expand_prologue): Corresponding changes.
16257 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16260 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
16261 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
16263 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
16266 * lra-constraints.c (curr_insn_transform): Process AND in the
16269 2018-01-30 Jakub Jelinek <jakub@redhat.com>
16271 PR rtl-optimization/83986
16272 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
16273 dependence against last_pending_memory_flush in addition to
16274 pending_jump_insns.
16276 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
16278 PR tree-optimization/81611
16279 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
16282 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16285 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
16288 2018-01-30 Richard Biener <rguenther@suse.de>
16289 Jakub Jelinek <jakub@redhat.com>
16291 PR tree-optimization/84111
16292 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
16293 inner loops added during recursion, as they don't have up-to-date
16296 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16299 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
16300 (can_inline_edge_by_limits_p): ... here.
16301 (can_early_inline_edge_p, check_callers,
16302 update_caller_keys, update_callee_keys, recursive_inlining,
16303 add_new_edges_to_heap, speculation_useful_p,
16304 inline_small_functions,
16305 inline_small_functions, flatten_function,
16306 inline_to_all_callers_1): Update.
16308 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16310 * profile-count.c (profile_count::combine_with_ipa_count): Handle
16313 2018-01-30 Richard Biener <rguenther@suse.de>
16315 PR tree-optimization/83008
16316 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
16317 invariant and constant vector uses in stmts when they need
16318 more than one stmt.
16320 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16323 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
16324 * configure: Regenerate.
16326 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16328 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
16330 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
16331 Use gen_rtx_REG rather than gen_lowpart.
16333 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16335 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
16336 rather than 0 when creating partial subregs.
16338 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16340 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
16343 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
16346 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
16347 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
16348 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
16349 flags. This restores the settings used before the 2017-07-24.
16350 Turning off pre increment/decrement/modify allows IVOPTS to
16351 optimize DF/SF loops where the index is an int.
16353 2018-01-29 Richard Biener <rguenther@suse.de>
16354 Kelvin Nilsen <kelvin@gcc.gnu.org>
16357 * tree-vect-stmts.c (vectorizable_call): Don't call
16358 targetm.vectorize_builtin_md_vectorized_function if callee is
16361 2018-01-22 Carl Love <cel@us.ibm.com>
16363 * doc/extend.tex: Fix typo in second arg in
16364 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
16366 2018-01-29 Richard Biener <rguenther@suse.de>
16368 PR tree-optimization/84086
16369 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
16370 (flush_ssaname_freelist): When SSA names were released reset
16371 the SCEV hash table.
16373 2018-01-29 Richard Biener <rguenther@suse.de>
16375 PR tree-optimization/84057
16376 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
16377 removed paths when removing edges.
16379 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
16381 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
16382 -mfunction-return=@var{choice}.
16384 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
16386 PR diagnostic/84034
16387 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
16388 Handle CR like TAB.
16389 (layout::print_source_line): Likewise.
16390 (test_get_line_width_without_trailing_whitespace): Add test cases.
16392 2018-01-27 Jakub Jelinek <jakub@redhat.com>
16394 PR middle-end/84040
16395 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
16398 2018-01-26 Jim Wilson <jimw@sifive.com>
16400 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
16402 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
16405 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16407 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
16408 and CMP + SUB-immediate -> SUBS.
16410 2018-01-26 Martin Sebor <msebor@redhat.com>
16412 PR tree-optimization/83896
16413 * tree-ssa-strlen.c (get_string_len): Rename...
16414 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
16415 Avoid assuming length is constant.
16416 (handle_char_store): Use HOST_WIDE_INT for string length.
16418 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
16421 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
16422 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
16424 2018-01-26 Richard Biener <rguenther@suse.de>
16426 PR rtl-optimization/84003
16427 * dse.c (record_store): Only record redundant stores when
16428 the earlier store aliases at least all accesses the later one does.
16430 2018-01-26 Jakub Jelinek <jakub@redhat.com>
16432 PR rtl-optimization/83985
16433 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
16434 REG_CFA_RESTORE insns.
16435 (delete_unmarked_insns): Don't ignore separate shrink wrapping
16436 REG_CFA_RESTORE insns here.
16439 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
16440 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
16442 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16444 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
16445 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
16446 (arc_init): Likewise.
16447 (arc_override_options): Likewise.
16448 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
16450 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
16452 * config/arc/arc.h (TARGET_DBNZ): Define.
16453 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
16454 properly set the tune attribute.
16455 (dbnz): Use TARGET_DBNZ guard.
16456 * config/arc/arc.opt (mtune): Add core3 option.
16458 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16460 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
16461 recognize new pic like addresses.
16462 (arc_delegitimize_address): Clean up.
16464 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16466 * config/arc/arc-arches.def: Option mrf16 valid for all
16468 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
16469 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
16470 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
16471 * config/arc/arc-tables.opt: Regenerate.
16472 * config/arc/arc.c (arc_conditional_register_usage): Handle
16473 reduced register file case.
16474 (arc_file_start): Set must have build attributes.
16475 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
16476 mrf16 option value.
16477 * config/arc/arc.opt (mrf16): Add new option.
16478 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
16479 * config/arc/genmultilib.awk: Handle new mrf16 option.
16480 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
16481 * config/arc/t-multilib: Regenerate.
16482 * doc/invoke.texi (ARC Options): Document mrf16 option.
16484 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16486 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
16487 * config/arc/arc.c (arc_handle_secure_attribute): New function.
16488 (arc_attribute_table): Add 'secure_call' attribute.
16489 (arc_print_operand): Print secure call operand.
16490 (arc_function_ok_for_sibcall): Don't optimize tail calls when
16492 (arc_is_secure_call_p): New function. * config/arc/arc.md
16493 (call_i): Add support for sjli instruction.
16494 (call_value_i): Likewise.
16495 * config/arc/constraints.md (Csc): New constraint.
16497 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16498 John Eric Martin <John.Martin@emmicro-us.com>
16500 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
16501 * config/arc/arc.c (_arc_jli_section): New struct.
16502 (arc_jli_section): New type.
16503 (rc_jli_sections): New static variable.
16504 (arc_handle_jli_attribute): New function.
16505 (arc_attribute_table): Add jli_always and jli_fixed attribute.
16506 (arc_file_end): New function.
16507 (TARGET_ASM_FILE_END): Define.
16508 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
16509 (arc_add_jli_section): New function.
16510 (jli_call_scan): Likewise.
16511 (arc_reorg): Call jli_call_scan.
16512 (arc_output_addsi): Remove 'S' from printing asm operand.
16513 (arc_is_jli_call_p): New function.
16514 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
16516 (movhi_insn): Likewise.
16517 (movsi_insn): Likewise.
16518 (movsi_set_cc_insn): Likewise.
16519 (loadqi_update): Likewise.
16520 (load_zeroextendqisi_update): Likewise.
16521 (load_signextendqisi_update): Likewise.
16522 (loadhi_update): Likewise.
16523 (load_zeroextendhisi_update): Likewise.
16524 (load_signextendhisi_update): Likewise.
16525 (loadsi_update): Likewise.
16526 (loadsf_update): Likewise.
16527 (movsicc_insn): Likewise.
16528 (bset_insn): Likewise.
16529 (bxor_insn): Likewise.
16530 (bclr_insn): Likewise.
16531 (bmsk_insn): Likewise.
16532 (bicsi3_insn): Likewise.
16533 (cmpsi_cc_c_insn): Likewise.
16534 (movsi_ne): Likewise.
16535 (movsi_cond_exec): Likewise.
16536 (clrsbsi2): Likewise.
16537 (norm_f): Likewise.
16547 (call_i): Remove 'S' asm letter, add jli instruction.
16548 (call_value_i): Likewise.
16549 * config/arc/arc.op (mjli-always): New option.
16550 * config/arc/constraints.md (Cji): New constraint.
16551 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
16553 (subsf3_fpx): Likewise.
16554 (mulsf3_fpx): Likewise.
16555 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
16557 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
16558 function attrbutes.
16559 * doc/invoke.texi (ARC): Document mjli-always option.
16561 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
16563 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
16564 avoid addition with 0 and use incw and decw where possible.
16566 2018-01-26 Richard Biener <rguenther@suse.de>
16568 PR tree-optimization/81082
16569 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
16570 association if it requires casting to unsigned.
16571 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
16572 from fold_plusminus_mult_expr to catch important cases late when
16573 range info is available.
16575 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16577 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
16578 * configure.ac (hidden_linkonce): New test.
16579 * configure: Regenerate.
16580 * config.in: Regenerate.
16582 2018-01-26 Julia Koval <julia.koval@intel.com>
16584 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
16585 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
16586 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
16587 _mm_mask_bitshuffle_epi64_mask): Fix type.
16588 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
16589 USI_FTYPE_V4DI_V4DI_USI): Remove.
16590 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
16591 __builtin_ia32_vpshufbitqmb256_mask,
16592 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
16593 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
16594 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
16596 2018-01-26 Alan Modra <amodra@gmail.com>
16599 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
16600 UNSPEC_VBPERMQ. Sort other unspecs.
16602 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
16604 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
16606 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
16608 PR middle-end/83055
16609 * predict.c (drop_profile): Do not push/pop cfun; update also
16611 (handle_missing_profiles): Fix logic looking for zero profiles.
16613 2018-01-25 Jakub Jelinek <jakub@redhat.com>
16615 PR middle-end/83977
16616 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
16617 on functions with #pragma omp declare simd or functions with simd
16619 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
16620 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
16621 Remove trailing \n from warning_at calls.
16623 2018-01-25 Tom de Vries <tom@codesourcery.com>
16626 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16627 for neutered workers.
16629 2018-01-24 Joseph Myers <joseph@codesourcery.com>
16632 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
16633 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
16635 2018-01-24 Jeff Law <law@redhat.com>
16638 * i386.c (get_probe_interval): Move to earlier point.
16639 (ix86_compute_frame_layout): If -fstack-clash-protection and
16640 the frame is larger than the probe interval, then use pushes
16641 to save registers rather than reg->mem moves.
16642 (ix86_expand_prologue): Remove conditional for int_registers_saved
16645 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
16648 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
16649 min/max for never referenced object.
16651 2018-01-24 Jakub Jelinek <jakub@redhat.com>
16653 PR middle-end/83977
16654 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
16656 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
16657 attributes from DECL_ATTRIBUTES (decl) without affecting
16658 DECL_ATTRIBUTES (current_function_decl).
16659 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
16660 functions with non-NULL DECL_ABSTRACT_ORIGIN.
16662 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
16664 PR tree-optimization/83979
16665 * fold-const.c (fold_comparison): Use constant_boolean_node
16666 instead of boolean_{true,false}_node.
16668 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
16670 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
16673 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16675 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16676 Simplify the clause that sets the length attribute.
16677 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16678 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
16679 clause that sets the length attribute.
16680 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16682 2018-01-24 Tom de Vries <tom@codesourcery.com>
16685 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
16686 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
16687 Add strict parameter.
16688 (prevent_branch_around_nothing): Insert dummy insn between branch to
16689 label and label with no ptx insn inbetween.
16690 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
16692 2018-01-24 Tom de Vries <tom@codesourcery.com>
16695 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16696 for neutered threads in warp.
16697 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
16699 2018-01-24 Richard Biener <rguenther@suse.de>
16701 PR tree-optimization/83176
16702 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
16705 2018-01-24 Richard Biener <rguenther@suse.de>
16707 PR tree-optimization/82819
16708 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
16709 code generating pluses that are no-ops in the target precision.
16711 2018-01-24 Richard Biener <rguenther@suse.de>
16713 PR middle-end/84000
16714 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
16716 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16718 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
16719 to merge probabilities.
16720 * predict.c (probably_never_executed): Also mark as cold functions
16721 with global 0 profile and guessed local profile.
16722 * profile-count.c (profile_probability::combine_with_count): New
16724 * profile-count.h (profile_probability::operator*,
16725 profile_probability::operator*=, profile_probability::operator/,
16726 profile_probability::operator/=): Reduce precision to adjusted
16727 and set value to guessed on contradictory divisions.
16728 (profile_probability::combine_with_freq): Remove.
16729 (profile_probability::combine_wiht_count): Declare.
16730 (profile_count::force_nonzero):: Set to adjusted.
16731 (profile_count::probability_in):: Set quality to adjusted.
16732 * tree-ssa-tail-merge.c (replace_block_by): Use
16733 combine_with_count.
16735 2018-01-23 Andrew Waterman <andrew@sifive.com>
16736 Jim Wilson <jimw@sifive.com>
16738 * config/riscv/riscv.c (riscv_stack_boundary): New.
16739 (riscv_option_override): Set riscv_stack_boundary. Handle
16740 riscv_preferred_stack_boundary_arg.
16741 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
16742 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
16743 (STACK_BOUNDARY): Set to riscv_stack_boundary.
16744 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
16745 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
16746 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
16748 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
16751 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
16752 of struct ix86_frame.
16753 (ix86_expand_epilogue): Likewise. Add a local variable for
16754 the reg_save_offset field in struct ix86_frame.
16756 2018-01-23 Bin Cheng <bin.cheng@arm.com>
16758 PR tree-optimization/82604
16759 * tree-loop-distribution.c (enum partition_kind): New enum item
16760 PKIND_PARTIAL_MEMSET.
16761 (partition_builtin_p): Support above new enum item.
16762 (generate_code_for_partition): Ditto.
16763 (compute_access_range): Differentiate cases that equality can be
16764 proven at all loops, the innermost loops or no loops.
16765 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
16766 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16767 (finalize_partitions, distribute_loop): Don't fuse partition of
16768 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16769 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16770 parloop is enabled.
16772 2018-01-23 Martin Liska <mliska@suse.cz>
16774 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16775 order to ignore the predictor.
16776 (PRED_POLYMORPHIC_CALL): Likewise.
16777 (PRED_RECURSIVE_CALL): Likewise.
16779 2018-01-23 Martin Liska <mliska@suse.cz>
16781 * tree-profile.c (tree_profiling): Print function header to
16782 aware reader which function we are working on.
16783 * value-prof.c (gimple_find_values_to_profile): Do not print
16784 not interesting value histograms.
16786 2018-01-23 Martin Liska <mliska@suse.cz>
16788 * profile-count.h (enum profile_quality): Add
16789 profile_uninitialized as the first value. Do not number values
16790 as they are zero based.
16791 (profile_count::verify): Update sanity check.
16792 (profile_probability::verify): Likewise.
16794 2018-01-23 Nathan Sidwell <nathan@acm.org>
16796 * doc/invoke.texi (ffor-scope): Deprecate.
16798 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16800 PR tree-optimization/83510
16801 * domwalk.c (set_all_edges_as_executable): New function.
16802 (dom_walker::dom_walker): Convert bool param
16803 "skip_unreachable_blocks" to enum reachability. Move setup of
16804 edge flags to set_all_edges_as_executable and only do it when
16805 reachability is REACHABLE_BLOCKS.
16806 * domwalk.h (enum dom_walker::reachability): New enum.
16807 (dom_walker::dom_walker): Convert bool param
16808 "skip_unreachable_blocks" to enum reachability.
16809 (set_all_edges_as_executable): New decl.
16810 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16811 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16813 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16814 but converting true to REACHABLE_BLOCKS.
16815 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16817 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16818 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16819 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16821 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16822 if check_all_array_refs will be called.
16824 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16826 * tree.c (selftest::test_location_wrappers): Add more test
16829 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16831 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16832 (selftest::test_bit_in_range): Likewise.
16834 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16837 * doc/sourcebuild.texi (vect_float): Say that the selector
16838 only describes the situation when -funsafe-math-optimizations is on.
16839 (vect_float_strict): Document.
16841 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16843 PR tree-optimization/83965
16844 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16845 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16846 instead of checking only for a reduction.
16847 (vect_recog_widen_sum_pattern): Likewise.
16849 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16851 * predict.c (probably_never_executed): Only use precise profile info.
16852 (compute_function_frequency): Skip after inlining hack since we now
16853 have quality checking.
16855 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16857 * profile-count.h (profile_probability::very_unlikely,
16858 profile_probability::unlikely, profile_probability::even): Set
16859 precision to guessed.
16861 2018-01-23 Richard Biener <rguenther@suse.de>
16863 PR tree-optimization/83963
16864 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16865 Properly terminate dominator walk when crossing the exit edge not
16866 when visiting its source block.
16868 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16871 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16872 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16874 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16876 PR tree-optimization/83957
16877 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16878 semicolon after for body surrounded by braces.
16880 PR tree-optimization/83081
16881 * profile-count.h (profile_probability::split): New method.
16882 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16883 Use profile_probability::split.
16884 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16885 when splitting a single conditional jump into 2.
16887 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16889 PR tree-optimization/69452
16890 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16893 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16895 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16896 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16897 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16899 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16901 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16903 * config/rl78/rl78.md (movdi): New define_expand.
16904 * config/rl78/rl78.c (rl78_split_movdi): New function.
16906 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16909 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16911 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16912 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16913 128-bit to produce an UNSPEC move to get the double word with the
16914 signbit and then a shift directly to do signbit.
16915 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16916 implementation with a new version that just does either a direct
16917 move or a regular move. Move memory interface to separate insns.
16918 Move insns so they are next to the expander.
16919 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16920 with signbit move. Split big and little endian case.
16921 (signbit<mode>2_dm_mem_le): Likewise.
16922 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16923 (signbit<mode>2_dm2): Likewise.
16925 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16927 * config/rl78/rl78.md (anddi3): New define_expand.
16929 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16931 * config/rl78/rl78.md (umindi3): New define_expand.
16933 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16935 * config/rl78/rl78.md (smindi3): New define_expand.
16937 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16939 * config/rl78/rl78.md (smaxdi3): New define_expand.
16941 2018-01-22 Carl Love <cel@us.ibm.com>
16943 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16944 LVX_V1TI): Add macro expansion.
16945 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16946 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16947 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16948 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16949 Change check to determine if the instruction is a byte reversing
16950 entry. Fix typo in comment.
16951 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16952 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16953 Add def_builtin calls for new builtins.
16954 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16955 Add define_insn expansion.
16957 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16959 * config/rl78/rl78.md (umaxdi3): New define_expand.
16961 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16963 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16964 for non-QImode registers.
16966 2018-01-22 Richard Biener <rguenther@suse.de>
16968 PR tree-optimization/83963
16969 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16970 including the loop exit block.
16971 (scop_detection::merge_sese): Likewise.
16972 (scop_detection::add_scop): Do it here instead.
16974 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16976 * doc/sourcebuild.texi (arm_softfloat): Document.
16978 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16981 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16982 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16983 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16985 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16986 David Edelsohn <dje.gcc@gmail.com>
16989 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16990 Change "crset eq" to "crset 2".
16991 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16992 (*call_indirect_aix<mode>_nospec): Likewise.
16993 (*call_value_indirect_aix<mode>_nospec): Likewise.
16994 (*call_indirect_elfv2<mode>_nospec): Likewise.
16995 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16996 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16997 change assembly output from . to $.
16998 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16999 (indirect_jump<mode>_nospec): Change assembly output from . to $.
17000 (*tablejump<mode>_internal1_nospec): Likewise.
17002 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
17005 * config/sh/sh_optimize_sett_clrt.cc:
17006 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
17008 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
17010 PR tree-optimization/83940
17011 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
17012 offset_dt to vect_constant_def rather than vect_unknown_def_type.
17013 (vect_check_load_store_mask): Add a mask_dt_out parameter and
17014 use it to pass back the definition type.
17015 (vect_check_store_rhs): Likewise rhs_dt_out.
17016 (vect_build_gather_load_calls): Add a mask_dt argument and use
17017 it instead of a call to vect_is_simple_use.
17018 (vectorizable_store): Update calls to vect_check_load_store_mask
17019 and vect_check_store_rhs. Use the dt returned by the latter instead
17020 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
17021 instead of calls to vect_is_simple_use. Pass the scalar rather
17022 than the vector operand to vect_is_simple_use when handling
17023 second and subsequent copies of an rhs value.
17024 (vectorizable_load): Update calls to vect_check_load_store_mask
17025 and vect_build_gather_load_calls. Use the cached mask_dt and
17026 gs_info.offset_dt instead of calls to vect_is_simple_use.
17028 2018-01-20 Jakub Jelinek <jakub@redhat.com>
17030 PR middle-end/83945
17031 * tree-emutls.c: Include gimplify.h.
17032 (lower_emutls_2): New function.
17033 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
17034 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
17035 it before further processing.
17038 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
17039 UINTVAL (trueop1) instead of INTVAL (op1).
17041 2018-01-19 Jakub Jelinek <jakub@redhat.com>
17045 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
17046 INCOMING_FRAME_SP_OFFSET if not defined.
17047 (scan_trace): Add ENTRY argument. If true and
17048 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
17049 emit a note to adjust the CFA offset.
17050 (create_cfi_notes): Adjust scan_trace callers.
17051 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
17052 INCOMING_FRAME_SP_OFFSET in the CIE.
17053 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
17054 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
17056 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
17057 * doc/tm.texi: Regenerated.
17059 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17061 PR rtl-optimization/83147
17062 * lra-constraints.c (remove_inheritance_pseudos): Use
17063 lra_substitute_pseudo_within_insn.
17065 2018-01-19 Tom de Vries <tom@codesourcery.com>
17066 Cesar Philippidis <cesar@codesourcery.com>
17069 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
17071 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
17074 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
17075 spaces for function labels.
17077 2018-01-19 Martin Liska <mliska@suse.cz>
17079 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
17080 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
17081 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
17082 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
17083 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
17084 (PRED_CONST_RETURN): Change from 69 to 65.
17085 (PRED_NULL_RETURN): Change from 91 to 71.
17086 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
17087 (PRED_LOOP_GUARD): Change from 66 to 73.
17089 2018-01-19 Martin Liska <mliska@suse.cz>
17091 * predict.c (predict_insn_def): Add new assert.
17092 (struct branch_predictor): Change type to signed integer.
17093 (test_prediction_value_range): Amend test to cover
17094 PROB_UNINITIALIZED.
17095 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
17096 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
17097 (PRED_LOOP_ITERATIONS_MAX): Likewise.
17098 (PRED_LOOP_IV_COMPARE): Likewise.
17099 * predict.h (PROB_UNINITIALIZED): Define new constant.
17101 2018-01-19 Martin Liska <mliska@suse.cz>
17103 * predict.c (dump_prediction): Add new format for
17104 analyze_brprob.py script which is enabled with -details
17106 * profile-count.h (precise_p): New function.
17108 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17110 PR tree-optimization/83922
17111 * tree-vect-loop.c (vect_verify_full_masking): Return false if
17112 there are no statements that need masking.
17113 (vect_active_double_reduction_p): New function.
17114 (vect_analyze_loop_operations): Use it when handling phis that
17115 are not in the loop header.
17117 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17119 PR tree-optimization/83914
17120 * tree-vect-loop.c (vectorizable_induction): Don't convert
17121 init_expr or apply the peeling adjustment for inductions
17122 that are nested within the vectorized loop.
17124 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17126 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
17129 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17133 * function.h (gimplify_parameters): Add gimple_seq * argument.
17134 * function.c: Include gimple.h and options.h.
17135 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
17136 for the added local temporaries if needed.
17137 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
17138 if there are any parameter cleanups, wrap whole body into a
17139 try/finally with the cleanups.
17141 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
17144 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
17145 Use GET_MODE_CLASS for scalar floating point.
17147 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17151 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
17152 Fix call of call_cgraph_insertion_hooks.
17154 2018-01-18 Martin Sebor <msebor@redhat.com>
17156 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
17158 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17161 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
17164 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
17167 * common.opt: (-ffile-prefix-map): New option.
17168 * opts.c (common_handle_option): Defer it.
17169 * opts-global.c (handle_common_deferred_options): Handle it.
17170 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
17171 * file-prefix-map.h: New file.
17172 (remap_debug_filename, add_debug_prefix_map): ...here.
17173 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
17174 * final.c (debug_prefix_map, add_debug_prefix_map
17175 remap_debug_filename): Move to...
17176 * file-prefix-map.c: New file.
17177 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
17178 generalize, get rid of alloca(), use strrchr() instead of strchr().
17179 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
17180 Implement in terms of add_prefix_map().
17181 (remap_macro_filename, remap_debug_filename): Implement in term of
17183 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
17184 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
17185 * dbxout.c: Include file-prefix-map.h.
17186 * varasm.c: Likewise.
17187 * vmsdbgout.c: Likewise.
17188 * xcoffout.c: Likewise.
17189 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
17190 * doc/cppopts.texi (-fmacro-prefix-map): Document.
17191 * doc/invoke.texi (-ffile-prefix-map): Document.
17192 (-fdebug-prefix-map): Update description.
17194 2018-01-18 Martin Liska <mliska@suse.cz>
17196 * config/i386/i386.c (indirect_thunk_name): Document that also
17198 (output_indirect_thunk): Document why both instructions
17199 (pause and lfence) are generated.
17201 2018-01-18 Richard Biener <rguenther@suse.de>
17203 PR tree-optimization/83887
17204 * graphite-scop-detection.c
17205 (scop_detection::get_nearest_dom_with_single_entry): Remove.
17206 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
17207 (scop_detection::merge_sese): Re-implement with a flood-fill
17208 algorithm that properly finds a SESE region if it exists.
17210 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17213 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
17214 pointer_diff optimizations use view_convert instead of convert.
17216 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17218 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17219 Generate different code for -mno-speculate-indirect-jumps.
17220 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17221 (*call_indirect_aix<mode>): Disable for
17222 -mno-speculate-indirect-jumps.
17223 (*call_indirect_aix<mode>_nospec): New define_insn.
17224 (*call_value_indirect_aix<mode>): Disable for
17225 -mno-speculate-indirect-jumps.
17226 (*call_value_indirect_aix<mode>_nospec): New define_insn.
17227 (*sibcall_nonlocal_sysv<mode>): Generate different code for
17228 -mno-speculate-indirect-jumps.
17229 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17231 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17233 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
17234 long double type, set the flags for noting the default long double
17235 type, even if we don't pass or return a long double type.
17237 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
17240 * ipa-inline.c (flatten_function): Do not overwrite final inlining
17243 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
17245 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
17246 support for merge[hl].
17247 (fold_mergehl_helper): New helper function.
17248 (tree-vector-builder.h): New #include for tree_vector_builder usage.
17249 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
17250 (altivec_vmrglw_direct): Add xxmrglw insn.
17252 2018-01-17 Andrew Waterman <andrew@sifive.com>
17254 * config/riscv/riscv.c (riscv_conditional_register_usage): If
17255 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
17257 2018-01-17 David Malcolm <dmalcolm@redhat.com>
17260 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
17261 call the lto_location_cache before reading the
17262 DECL_SOURCE_LOCATION of the types.
17264 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
17265 Richard Sandiford <richard.sandiford@linaro.org>
17267 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
17268 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
17269 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
17270 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
17271 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
17273 * config/aarch64/constraints.md (aarch64_movti_operand):
17275 * config/aarch64/predicates.md (Uti): Add new constraint.
17277 2018-01-17 Carl Love <cel@us.ibm.com>
17279 * config/rs6000/vsx.md (define_expand xl_len_r,
17280 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
17281 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
17283 (define_expand, define_insn): Move the shift left from the
17284 define_insn to the define_expand for lxvl and stxvl instructions.
17285 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
17286 and XL_LEN_R definitions to PURE.
17288 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
17290 * config/i386/i386.c (indirect_thunk_name): Declare regno
17291 as unsigned int. Compare regno with INVALID_REGNUM.
17292 (output_indirect_thunk): Ditto.
17293 (output_indirect_thunk_function): Ditto.
17294 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
17295 in the call to output_indirect_thunk_function.
17297 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
17299 PR middle-end/83884
17300 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
17301 rather than the size of inner_type to determine the stack slot size
17302 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
17304 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
17307 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
17310 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
17312 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
17313 endian Linux systems to optionally enable multilibs for selecting
17314 the long double type if the user configured an explicit type.
17315 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
17316 have no long double multilibs if not defined.
17317 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
17318 warn if the user used -mabi={ieee,ibm}longdouble and we built
17319 multilibs for long double.
17320 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
17321 appropriate multilib option.
17322 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
17324 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
17325 for building long double multilibs.
17326 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
17328 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
17330 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
17333 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
17335 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
17338 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
17341 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
17344 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17346 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
17347 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
17349 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
17351 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
17352 different rtl trees depending on TARGET_64BIT.
17353 (rs6000_gen_lvx): Likewise.
17355 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17357 * config/visium/visium.md (nop): Tweak comment.
17358 (hazard_nop): Likewise.
17360 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17362 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
17363 -mspeculate-indirect-jumps.
17364 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
17365 for -mno-speculate-indirect-jumps.
17366 (*call_indirect_elfv2<mode>_nospec): New define_insn.
17367 (*call_value_indirect_elfv2<mode>): Disable for
17368 -mno-speculate-indirect-jumps.
17369 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
17370 (indirect_jump): Emit different RTL for
17371 -mno-speculate-indirect-jumps.
17372 (*indirect_jump<mode>): Disable for
17373 -mno-speculate-indirect-jumps.
17374 (*indirect_jump<mode>_nospec): New define_insn.
17375 (tablejump): Emit different RTL for
17376 -mno-speculate-indirect-jumps.
17377 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
17378 (tablejumpsi_nospec): New define_expand.
17379 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
17380 (tablejumpdi_nospec): New define_expand.
17381 (*tablejump<mode>_internal1): Disable for
17382 -mno-speculate-indirect-jumps.
17383 (*tablejump<mode>_internal1_nospec): New define_insn.
17384 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
17387 2018-01-16 Artyom Skrobov tyomitch@gmail.com
17389 * caller-save.c (insert_save): Drop unnecessary parameter. All
17392 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17393 Richard Biener <rguenth@suse.de>
17396 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
17397 return early, inline manually is_gimple_sizepos. Make sure if we
17398 call gimplify_expr we don't end up with a gimple constant.
17399 * tree.c (variably_modified_type_p): Don't return true for
17400 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
17401 * gimplify.h (is_gimple_sizepos): Remove.
17403 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17405 PR tree-optimization/83857
17406 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
17407 vectorizable_live_operation for pure SLP statements.
17408 (vectorizable_live_operation): Handle PHIs.
17410 2018-01-16 Richard Biener <rguenther@suse.de>
17412 PR tree-optimization/83867
17413 * tree-vect-stmts.c (vect_transform_stmt): Precompute
17414 nested_in_vect_loop_p since the scalar stmt may get invalidated.
17416 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17419 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
17420 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
17421 If off is not INTEGER_CST, issue a may not be aligned warning
17422 rather than isn't aligned. Use isn%'t rather than isn't.
17423 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
17425 <case MULT_EXPR>: Improve the case when bottom and one of the
17426 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
17427 operand, in that case check if the other operand is multiple of
17428 bottom divided by the INTEGER_CST operand.
17430 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17433 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
17434 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
17435 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
17436 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
17437 * config/pa/pa.c (pa_function_arg_advance): Likewise.
17438 (pa_function_arg, pa_arg_partial_bytes): Likewise.
17439 (pa_function_arg_size): New function.
17441 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17443 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
17444 in a separate statement.
17446 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17448 PR tree-optimization/83847
17449 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
17450 group gathers and scatters.
17452 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17454 PR rtl-optimization/86620
17455 * params.def (max-sched-ready-insns): Bump minimum value to 1.
17457 PR rtl-optimization/83213
17458 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
17459 to last if both are JUMP_INSNs.
17461 PR tree-optimization/83843
17462 * gimple-ssa-store-merging.c
17463 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
17464 store_immediate_info for bswap/nop orig_stores.
17466 2018-01-15 Andrew Waterman <andrew@sifive.com>
17468 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
17470 <UDIV>: Increase cost if !TARGET_DIV.
17472 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
17474 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
17475 (define_attr "cr_logical_3op"): New.
17476 (cceq_ior_compare): Adjust.
17477 (cceq_ior_compare_complement): Adjust.
17478 (*cceq_rev_compare): Adjust.
17479 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
17480 (is_cracked_insn): Adjust.
17481 (insn_must_be_first_in_group): Adjust.
17482 * config/rs6000/40x.md: Adjust.
17483 * config/rs6000/440.md: Adjust.
17484 * config/rs6000/476.md: Adjust.
17485 * config/rs6000/601.md: Adjust.
17486 * config/rs6000/603.md: Adjust.
17487 * config/rs6000/6xx.md: Adjust.
17488 * config/rs6000/7450.md: Adjust.
17489 * config/rs6000/7xx.md: Adjust.
17490 * config/rs6000/8540.md: Adjust.
17491 * config/rs6000/cell.md: Adjust.
17492 * config/rs6000/e300c2c3.md: Adjust.
17493 * config/rs6000/e500mc.md: Adjust.
17494 * config/rs6000/e500mc64.md: Adjust.
17495 * config/rs6000/e5500.md: Adjust.
17496 * config/rs6000/e6500.md: Adjust.
17497 * config/rs6000/mpc.md: Adjust.
17498 * config/rs6000/power4.md: Adjust.
17499 * config/rs6000/power5.md: Adjust.
17500 * config/rs6000/power6.md: Adjust.
17501 * config/rs6000/power7.md: Adjust.
17502 * config/rs6000/power8.md: Adjust.
17503 * config/rs6000/power9.md: Adjust.
17504 * config/rs6000/rs64.md: Adjust.
17505 * config/rs6000/titan.md: Adjust.
17507 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17509 * config/i386/predicates.md (indirect_branch_operand): Rewrite
17510 ix86_indirect_branch_register logic.
17512 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17514 * config/i386/constraints.md (Bs): Update
17515 ix86_indirect_branch_register check. Don't check
17516 ix86_indirect_branch_register with GOT_memory_operand.
17518 * config/i386/predicates.md (GOT_memory_operand): Don't check
17519 ix86_indirect_branch_register here.
17520 (GOT32_symbol_operand): Likewise.
17522 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17524 * config/i386/predicates.md (constant_call_address_operand):
17525 Rewrite ix86_indirect_branch_register logic.
17526 (sibcall_insn_operand): Likewise.
17528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17530 * config/i386/constraints.md (Bs): Replace
17531 ix86_indirect_branch_thunk_register with
17532 ix86_indirect_branch_register.
17534 * config/i386/i386.md (indirect_jump): Likewise.
17535 (tablejump): Likewise.
17536 (*sibcall_memory): Likewise.
17537 (*sibcall_value_memory): Likewise.
17538 Peepholes of indirect call and jump via memory: Likewise.
17539 * config/i386/i386.opt: Likewise.
17540 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17541 (GOT_memory_operand): Likewise.
17542 (call_insn_operand): Likewise.
17543 (sibcall_insn_operand): Likewise.
17544 (GOT32_symbol_operand): Likewise.
17546 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17548 PR middle-end/83837
17549 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
17550 type rather than type addr's type points to.
17551 (expand_omp_atomic_mutex): Likewise.
17552 (expand_omp_atomic): Likewise.
17554 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17557 * config/i386/i386.c (output_indirect_thunk_function): Use
17558 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
17559 for __x86_return_thunk.
17561 2018-01-15 Richard Biener <rguenther@suse.de>
17563 PR middle-end/83850
17564 * expmed.c (extract_bit_field_1): Fix typo.
17566 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17569 * config/arm/iterators.md (VF): New mode iterator.
17570 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
17571 Remove integer-related logic from pattern.
17572 (neon_vabd<mode>_3): Likewise.
17574 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17576 PR middle-end/82694
17577 * common.opt (fstrict-overflow): No longer an alias.
17578 (fwrapv-pointer): New option.
17579 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
17580 also for pointer types based on flag_wrapv_pointer.
17581 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
17582 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
17583 opts->x_flag_wrapv got set.
17584 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
17585 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
17586 POINTER_TYPE_OVERFLOW_UNDEFINED.
17587 * match.pd: Likewise in address comparison pattern.
17588 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
17590 2018-01-15 Richard Biener <rguenther@suse.de>
17593 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
17594 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
17595 Reset type names to their identifier if their TYPE_DECL doesn't
17596 have linkage (and thus is used for ODR and devirt).
17597 (save_debug_info_for_decl): Remove.
17598 (save_debug_info_for_type): Likewise.
17599 (add_tree_to_fld_list): Adjust.
17600 * tree-pretty-print.c (dump_generic_node): Make dumping of
17601 type names more robust.
17603 2018-01-15 Richard Biener <rguenther@suse.de>
17605 * BASE-VER: Bump to 8.0.1.
17607 2018-01-14 Martin Sebor <msebor@redhat.com>
17610 * builtins.c (check_access): Avoid warning when the no-warning bit
17613 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
17615 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
17616 * ira-color (allocno_hard_regs_compare): Likewise.
17618 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
17621 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
17622 Use .pushsection/.popsection.
17624 2018-01-14 Martin Sebor <msebor@redhat.com>
17627 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
17629 2018-01-14 Jakub Jelinek <jakub@redhat.com>
17631 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
17632 entry from extra_headers.
17633 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
17634 extra_headers, make the list bitwise identical to the i?86-*-* one.
17636 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17638 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
17639 -mcmodel=large with -mindirect-branch=thunk,
17640 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
17641 -mfunction-return=thunk-extern.
17642 * doc/invoke.texi: Document -mcmodel=large is incompatible with
17643 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
17644 -mfunction-return=thunk and -mfunction-return=thunk-extern.
17646 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17648 * config/i386/i386.c (print_reg): Print the name of the full
17649 integer register without '%'.
17650 (ix86_print_operand): Handle 'V'.
17651 * doc/extend.texi: Document 'V' modifier.
17653 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17655 * config/i386/constraints.md (Bs): Disallow memory operand for
17656 -mindirect-branch-register.
17658 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17659 (GOT_memory_operand): Likewise.
17660 (call_insn_operand): Likewise.
17661 (sibcall_insn_operand): Likewise.
17662 (GOT32_symbol_operand): Likewise.
17663 * config/i386/i386.md (indirect_jump): Call convert_memory_address
17664 for -mindirect-branch-register.
17665 (tablejump): Likewise.
17666 (*sibcall_memory): Likewise.
17667 (*sibcall_value_memory): Likewise.
17668 Disallow peepholes of indirect call and jump via memory for
17669 -mindirect-branch-register.
17670 (*call_pop): Replace m with Bw.
17671 (*call_value_pop): Likewise.
17672 (*sibcall_pop_memory): Replace m with Bs.
17673 * config/i386/i386.opt (mindirect-branch-register): New option.
17674 * doc/invoke.texi: Document -mindirect-branch-register option.
17676 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17678 * config/i386/i386-protos.h (ix86_output_function_return): New.
17679 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
17680 set function_return_type.
17681 (indirect_thunk_name): Add ret_p to indicate thunk for function
17683 (output_indirect_thunk_function): Pass false to
17684 indirect_thunk_name.
17685 (ix86_output_indirect_branch_via_reg): Likewise.
17686 (ix86_output_indirect_branch_via_push): Likewise.
17687 (output_indirect_thunk_function): Create alias for function
17688 return thunk if regno < 0.
17689 (ix86_output_function_return): New function.
17690 (ix86_handle_fndecl_attribute): Handle function_return.
17691 (ix86_attribute_table): Add function_return.
17692 * config/i386/i386.h (machine_function): Add
17693 function_return_type.
17694 * config/i386/i386.md (simple_return_internal): Use
17695 ix86_output_function_return.
17696 (simple_return_internal_long): Likewise.
17697 * config/i386/i386.opt (mfunction-return=): New option.
17698 (indirect_branch): Mention -mfunction-return=.
17699 * doc/extend.texi: Document function_return function attribute.
17700 * doc/invoke.texi: Document -mfunction-return= option.
17702 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17704 * config/i386/i386-opts.h (indirect_branch): New.
17705 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
17706 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
17707 with local indirect jump when converting indirect call and jump.
17708 (ix86_set_indirect_branch_type): New.
17709 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
17710 (indirectlabelno): New.
17711 (indirect_thunk_needed): Likewise.
17712 (indirect_thunk_bnd_needed): Likewise.
17713 (indirect_thunks_used): Likewise.
17714 (indirect_thunks_bnd_used): Likewise.
17715 (INDIRECT_LABEL): Likewise.
17716 (indirect_thunk_name): Likewise.
17717 (output_indirect_thunk): Likewise.
17718 (output_indirect_thunk_function): Likewise.
17719 (ix86_output_indirect_branch_via_reg): Likewise.
17720 (ix86_output_indirect_branch_via_push): Likewise.
17721 (ix86_output_indirect_branch): Likewise.
17722 (ix86_output_indirect_jmp): Likewise.
17723 (ix86_code_end): Call output_indirect_thunk_function if needed.
17724 (ix86_output_call_insn): Call ix86_output_indirect_branch if
17726 (ix86_handle_fndecl_attribute): Handle indirect_branch.
17727 (ix86_attribute_table): Add indirect_branch.
17728 * config/i386/i386.h (machine_function): Add indirect_branch_type
17729 and has_local_indirect_jump.
17730 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
17732 (tablejump): Likewise.
17733 (*indirect_jump): Use ix86_output_indirect_jmp.
17734 (*tablejump_1): Likewise.
17735 (simple_return_indirect_internal): Likewise.
17736 * config/i386/i386.opt (mindirect-branch=): New option.
17737 (indirect_branch): New.
17740 (thunk-inline): Likewise.
17741 (thunk-extern): Likewise.
17742 * doc/extend.texi: Document indirect_branch function attribute.
17743 * doc/invoke.texi: Document -mindirect-branch= option.
17745 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
17748 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
17750 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
17752 * ipa-inline.c (want_inline_small_function_p): Return false if
17753 inlining has already failed with CIF_FINAL_ERROR.
17754 (update_caller_keys): Call want_inline_small_function_p before
17756 (update_callee_keys): Likewise.
17758 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17760 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
17762 (rs6000_quadword_masked_address_p): Likewise.
17763 (quad_aligned_load_p): Likewise.
17764 (quad_aligned_store_p): Likewise.
17765 (const_load_sequence_p): Add comment to describe the outer-most loop.
17766 (mimic_memory_attributes_and_flags): New function.
17767 (rs6000_gen_stvx): Likewise.
17768 (replace_swapped_aligned_store): Likewise.
17769 (rs6000_gen_lvx): Likewise.
17770 (replace_swapped_aligned_load): Likewise.
17771 (replace_swapped_load_constant): Capitalize argument name in
17772 comment describing this function.
17773 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17774 and stores that access quad-word aligned addresses and replace
17775 with stvx or lvx instructions when appropriate.
17776 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17777 New function prototype.
17778 (rs6000_quadword_masked_address_p): Likewise.
17779 (rs6000_gen_lvx): Likewise.
17780 (rs6000_gen_stvx): Likewise.
17781 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17782 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17783 when memory address is aligned.
17784 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17785 this split to select lvx instruction when memory address is aligned.
17786 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17787 instruction when memory address is aligned.
17788 (*vsx_le_perm_load_v16qi): Likewise.
17789 (four unnamed splitters): Modify to select the stvx instruction
17790 when memory is aligned.
17792 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17794 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17795 which appears in the queue multiple times.
17797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17798 Alan Hayward <alan.hayward@arm.com>
17799 David Sherwood <david.sherwood@arm.com>
17801 * tree-vectorizer.h (vec_lower_bound): New structure.
17802 (_loop_vec_info): Add check_nonzero and lower_bounds.
17803 (LOOP_VINFO_CHECK_NONZERO): New macro.
17804 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17805 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17806 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17807 fields. Make seg_len the distance travelled, not including the
17809 (dr_direction_indicator): Declare.
17810 (dr_zero_step_indicator): Likewise.
17811 (dr_known_forward_stride_p): Likewise.
17812 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17814 (runtime_alias_check_p): Allow runtime alias checks with
17816 (operator ==): Compare access_size and align.
17817 (prune_runtime_alias_test_list): Rework for new distinction between
17818 the access_size and seg_len.
17819 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17821 (get_segment_min_max): New function.
17822 (create_intersect_range_checks): Use it.
17823 (dr_step_indicator): New function.
17824 (dr_direction_indicator): Likewise.
17825 (dr_zero_step_indicator): Likewise.
17826 (dr_known_forward_stride_p): Likewise.
17827 * tree-loop-distribution.c (data_ref_segment_size): Return
17828 DR_STEP * (niters - 1).
17829 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17831 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17832 (vect_preserves_scalar_order_p): New function, split out from...
17833 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17834 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17835 (vect_vfa_access_size): New function.
17836 (vect_vfa_align): Likewise.
17837 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17838 (dump_lower_bound): New function.
17839 (vect_check_lower_bound): Likewise.
17840 (vect_small_gap_p): Likewise.
17841 (vectorizable_with_step_bound_p): Likewise.
17842 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17843 depencies if the vectorization factor is 1. Convert the checks
17844 for nonzero steps into checks on the bounds of DR_STEP. Try using
17845 a bunds check for variable steps if the minimum required step is
17846 relatively small. Update calls to the dr_with_seg_len
17847 constructor and to vect_compile_time_alias.
17848 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17850 (vect_loop_versioning): Call it.
17851 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17853 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17856 Alan Hayward <alan.hayward@arm.com>
17857 David Sherwood <david.sherwood@arm.com>
17859 * doc/sourcebuild.texi (vect_scatter_store): Document.
17860 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17862 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17864 * genopinit.c (main): Add supports_vec_scatter_store and
17865 supports_vec_scatter_store_cached to target_optabs.
17866 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17867 IFN_MASK_SCATTER_STORE.
17868 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17870 * internal-fn.h (internal_store_fn_p): Declare.
17871 (internal_fn_stored_value_index): Likewise.
17872 * internal-fn.c (scatter_store_direct): New macro.
17873 (expand_scatter_store_optab_fn): New function.
17874 (direct_scatter_store_optab_supported_p): New macro.
17875 (internal_store_fn_p): New function.
17876 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17877 IFN_MASK_SCATTER_STORE.
17878 (internal_fn_mask_index): Likewise.
17879 (internal_fn_stored_value_index): New function.
17880 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17881 for scatter stores.
17882 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17883 * optabs-query.c (supports_vec_scatter_store_p): New function.
17884 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17885 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17886 true for scatter stores.
17887 (vect_gather_scatter_fn_p): Handle scatter stores too.
17888 (vect_check_gather_scatter): Consider using scatter stores if
17889 supports_vec_scatter_store_p.
17890 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17891 scatter stores too.
17892 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17893 internal_fn_stored_value_index.
17894 (check_load_store_masking): Handle scatter stores too.
17895 (vect_get_store_rhs): Make public.
17896 (vectorizable_call): Use internal_store_fn_p.
17897 (vectorizable_store): Handle scatter store internal functions.
17898 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17899 when deciding whether the end of the group has been reached.
17900 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17901 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17902 (mask_scatter_store<mode>): New insns.
17904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17905 Alan Hayward <alan.hayward@arm.com>
17906 David Sherwood <david.sherwood@arm.com>
17908 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17909 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17910 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17912 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17913 Use vect_truncate_gather_scatter_offset if we can't treat the
17914 operation as a normal gather load or scatter store.
17915 (get_group_load_store_type): Take the gather_scatter_info
17916 as argument. Try using a gather load or scatter store for
17917 single-element groups.
17918 (get_load_store_type): Update calls to get_group_load_store_type
17919 and vect_use_strided_gather_scatters_p.
17921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17922 Alan Hayward <alan.hayward@arm.com>
17923 David Sherwood <david.sherwood@arm.com>
17925 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17926 optional tree argument.
17927 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17929 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17930 but continue to use the current value as a fallback.
17931 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17932 to compare the updates.
17933 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17934 (get_load_store_type): Use it when handling a strided access.
17935 (vect_get_strided_load_store_ops): New function.
17936 (vect_get_data_ptr_increment): Likewise.
17937 (vectorizable_load): Handle strided gather loads. Always pass
17938 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17941 Alan Hayward <alan.hayward@arm.com>
17942 David Sherwood <david.sherwood@arm.com>
17944 * doc/md.texi (gather_load@var{m}): Document.
17945 (mask_gather_load@var{m}): Likewise.
17946 * genopinit.c (main): Add supports_vec_gather_load and
17947 supports_vec_gather_load_cached to target_optabs.
17948 * optabs-tree.c (init_tree_optimization_optabs): Use
17949 ggc_cleared_alloc to allocate target_optabs.
17950 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17951 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17953 * internal-fn.h (internal_load_fn_p): Declare.
17954 (internal_gather_scatter_fn_p): Likewise.
17955 (internal_fn_mask_index): Likewise.
17956 (internal_gather_scatter_fn_supported_p): Likewise.
17957 * internal-fn.c (gather_load_direct): New macro.
17958 (expand_gather_load_optab_fn): New function.
17959 (direct_gather_load_optab_supported_p): New macro.
17960 (direct_internal_fn_optab): New function.
17961 (internal_load_fn_p): Likewise.
17962 (internal_gather_scatter_fn_p): Likewise.
17963 (internal_fn_mask_index): Likewise.
17964 (internal_gather_scatter_fn_supported_p): Likewise.
17965 * optabs-query.c (supports_at_least_one_mode_p): New function.
17966 (supports_vec_gather_load_p): Likewise.
17967 * optabs-query.h (supports_vec_gather_load_p): Declare.
17968 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17969 and memory_type field.
17970 (NUM_PATTERNS): Bump to 15.
17971 * tree-vect-data-refs.c: Include internal-fn.h.
17972 (vect_gather_scatter_fn_p): New function.
17973 (vect_describe_gather_scatter_call): Likewise.
17974 (vect_check_gather_scatter): Try using internal functions for
17975 gather loads. Recognize existing calls to a gather load function.
17976 (vect_analyze_data_refs): Consider using gather loads if
17977 supports_vec_gather_load_p.
17978 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17979 (vect_get_gather_scatter_offset_type): Likewise.
17980 (vect_convert_mask_for_vectype): Likewise.
17981 (vect_add_conversion_to_patterm): Likewise.
17982 (vect_try_gather_scatter_pattern): Likewise.
17983 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17984 (vect_vect_recog_func_ptrs): Add it.
17985 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17986 internal_fn_mask_index and internal_gather_scatter_fn_p.
17987 (check_load_store_masking): Take the gather_scatter_info as an
17988 argument and handle gather loads.
17989 (vect_get_gather_scatter_ops): New function.
17990 (vectorizable_call): Check internal_load_fn_p.
17991 (vectorizable_load): Likewise. Handle gather load internal
17993 (vectorizable_store): Update call to check_load_store_masking.
17994 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17995 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17996 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17997 (aarch64_gather_scale_operand_d): New predicates.
17998 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17999 (mask_gather_load<mode>): New insns.
18001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18002 Alan Hayward <alan.hayward@arm.com>
18003 David Sherwood <david.sherwood@arm.com>
18005 * optabs.def (fold_left_plus_optab): New optab.
18006 * doc/md.texi (fold_left_plus_@var{m}): Document.
18007 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
18008 * internal-fn.c (fold_left_direct): Define.
18009 (expand_fold_left_optab_fn): Likewise.
18010 (direct_fold_left_optab_supported_p): Likewise.
18011 * fold-const-call.c (fold_const_fold_left): New function.
18012 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
18013 * tree-parloops.c (valid_reduction_p): New function.
18014 (gather_scalar_reductions): Use it.
18015 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
18016 (vect_finish_replace_stmt): Declare.
18017 * tree-vect-loop.c (fold_left_reduction_fn): New function.
18018 (needs_fold_left_reduction_p): New function, split out from...
18019 (vect_is_simple_reduction): ...here. Accept reductions that
18020 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
18021 (vect_force_simple_reduction): Also store the reduction type in
18022 the assignment's STMT_VINFO_REDUC_TYPE.
18023 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
18024 (merge_with_identity): New function.
18025 (vect_expand_fold_left): Likewise.
18026 (vectorize_fold_left_reduction): Likewise.
18027 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
18028 scalar phi in place for it. Check for target support and reject
18029 cases that would reassociate the operation. Defer the transform
18030 phase to vectorize_fold_left_reduction.
18031 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
18032 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
18033 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
18035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18037 * tree-if-conv.c (predicate_mem_writes): Remove redundant
18038 call to ifc_temp_var.
18040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18041 Alan Hayward <alan.hayward@arm.com>
18042 David Sherwood <david.sherwood@arm.com>
18044 * target.def (legitimize_address_displacement): Take the original
18045 offset as a poly_int.
18046 * targhooks.h (default_legitimize_address_displacement): Update
18048 * targhooks.c (default_legitimize_address_displacement): Likewise.
18049 * doc/tm.texi: Regenerate.
18050 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
18051 as an argument, moving assert of ad->disp == ad->disp_term to...
18052 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
18053 Try calling targetm.legitimize_address_displacement before expanding
18054 the address rather than afterwards, and adjust for the new interface.
18055 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
18056 Match the new hook interface. Handle SVE addresses.
18057 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
18058 new hook interface.
18060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18062 * Makefile.in (OBJS): Add early-remat.o.
18063 * target.def (select_early_remat_modes): New hook.
18064 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
18065 * doc/tm.texi: Regenerate.
18066 * targhooks.h (default_select_early_remat_modes): Declare.
18067 * targhooks.c (default_select_early_remat_modes): New function.
18068 * timevar.def (TV_EARLY_REMAT): New timevar.
18069 * passes.def (pass_early_remat): New pass.
18070 * tree-pass.h (make_pass_early_remat): Declare.
18071 * early-remat.c: New file.
18072 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
18074 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
18076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18077 Alan Hayward <alan.hayward@arm.com>
18078 David Sherwood <david.sherwood@arm.com>
18080 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
18081 vfm1 with a bound_epilog parameter.
18082 (vect_do_peeling): Update calls accordingly, and move the prologue
18083 call earlier in the function. Treat the base bound_epilog as 0 for
18084 fully-masked loops and retain vf - 1 for other loops. Add 1 to
18085 this base when peeling for gaps.
18086 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
18087 with fully-masked loops.
18088 (vect_estimate_min_profitable_iters): Handle the single peeled
18089 iteration in that case.
18091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18092 Alan Hayward <alan.hayward@arm.com>
18093 David Sherwood <david.sherwood@arm.com>
18095 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
18096 single-element interleaving even if the size is not a power of 2.
18097 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
18098 accesses for single-element interleaving if the group size is
18101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18102 Alan Hayward <alan.hayward@arm.com>
18103 David Sherwood <david.sherwood@arm.com>
18105 * doc/md.texi (fold_extract_last_@var{m}): Document.
18106 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
18107 * optabs.def (fold_extract_last_optab): New optab.
18108 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
18109 * internal-fn.c (fold_extract_direct): New macro.
18110 (expand_fold_extract_optab_fn): Likewise.
18111 (direct_fold_extract_optab_supported_p): Likewise.
18112 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
18113 * tree-vect-loop.c (vect_model_reduction_cost): Handle
18114 EXTRACT_LAST_REDUCTION.
18115 (get_initial_def_for_reduction): Do not create an initial vector
18116 for EXTRACT_LAST_REDUCTION reductions.
18117 (vectorizable_reduction): Leave the scalar phi in place for
18118 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
18119 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
18120 epilogue code for EXTRACT_LAST_REDUCTION and defer the
18121 transform phase to vectorizable_condition.
18122 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
18124 (vect_finish_stmt_generation): ...here.
18125 (vect_finish_replace_stmt): New function.
18126 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
18127 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
18129 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
18131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18132 Alan Hayward <alan.hayward@arm.com>
18133 David Sherwood <david.sherwood@arm.com>
18135 * doc/md.texi (extract_last_@var{m}): Document.
18136 * optabs.def (extract_last_optab): New optab.
18137 * internal-fn.def (EXTRACT_LAST): New internal function.
18138 * internal-fn.c (cond_unary_direct): New macro.
18139 (expand_cond_unary_optab_fn): Likewise.
18140 (direct_cond_unary_optab_supported_p): Likewise.
18141 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
18142 loops using EXTRACT_LAST.
18143 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
18144 (extract_last_<mode>): ...this optab.
18145 (vec_extract<mode><Vel>): Update accordingly.
18147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18148 Alan Hayward <alan.hayward@arm.com>
18149 David Sherwood <david.sherwood@arm.com>
18151 * target.def (empty_mask_is_expensive): New hook.
18152 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
18153 * doc/tm.texi: Regenerate.
18154 * targhooks.h (default_empty_mask_is_expensive): Declare.
18155 * targhooks.c (default_empty_mask_is_expensive): New function.
18156 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
18157 if the target says that empty masks are expensive.
18158 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
18160 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
18162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18163 Alan Hayward <alan.hayward@arm.com>
18164 David Sherwood <david.sherwood@arm.com>
18166 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
18167 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
18168 (vect_use_loop_mask_for_alignment_p): New function.
18169 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
18170 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
18171 niters_skip argument. Make sure that the first niters_skip elements
18172 of the first iteration are inactive.
18173 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
18174 Update call to vect_set_loop_masks_directly.
18175 (get_misalign_in_elems): New function, split out from...
18176 (vect_gen_prolog_loop_niters): ...here.
18177 (vect_update_init_of_dr): Take a code argument that specifies whether
18178 the adjustment should be added or subtracted.
18179 (vect_update_init_of_drs): Likewise.
18180 (vect_prepare_for_masked_peels): New function.
18181 (vect_do_peeling): Skip prologue peeling if we're using a mask
18182 instead. Update call to vect_update_inits_of_drs.
18183 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18185 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
18186 alignment. Do not include the number of peeled iterations in
18187 the minimum threshold in that case.
18188 (vectorizable_induction): Adjust the start value down by
18189 LOOP_VINFO_MASK_SKIP_NITERS iterations.
18190 (vect_transform_loop): Call vect_prepare_for_masked_peels.
18191 Take the number of skipped iterations into account when calculating
18193 * tree-vect-stmts.c (vect_gen_while_not): New function.
18195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18196 Alan Hayward <alan.hayward@arm.com>
18197 David Sherwood <david.sherwood@arm.com>
18199 * doc/sourcebuild.texi (vect_fully_masked): Document.
18200 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
18201 default value to 0.
18202 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
18204 (vect_analyze_loop_2): ...here. Don't check the vectorization
18205 factor against the number of loop iterations if the loop is
18208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18209 Alan Hayward <alan.hayward@arm.com>
18210 David Sherwood <david.sherwood@arm.com>
18212 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
18213 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
18214 (dump_groups): Update accordingly.
18215 (iv_use::mem_type): New member variable.
18216 (address_p): New function.
18217 (record_use): Add a mem_type argument and initialize the new
18219 (record_group_use): Add a mem_type argument. Use address_p.
18220 Remove obsolete null checks of base_object. Update call to record_use.
18221 (find_interesting_uses_op): Update call to record_group_use.
18222 (find_interesting_uses_cond): Likewise.
18223 (find_interesting_uses_address): Likewise.
18224 (get_mem_type_for_internal_fn): New function.
18225 (find_address_like_use): Likewise.
18226 (find_interesting_uses_stmt): Try find_address_like_use before
18227 calling find_interesting_uses_op.
18228 (addr_offset_valid_p): Use the iv mem_type field as the type
18229 of the addressed memory.
18230 (add_autoinc_candidates): Likewise.
18231 (get_address_cost): Likewise.
18232 (split_small_address_groups_p): Use address_p.
18233 (split_address_groups): Likewise.
18234 (add_iv_candidate_for_use): Likewise.
18235 (autoinc_possible_for_pair): Likewise.
18236 (rewrite_groups): Likewise.
18237 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
18238 (determine_group_iv_cost): Update after split of USE_ADDRESS.
18239 (get_alias_ptr_type_for_ptr_address): New function.
18240 (rewrite_use_address): Rewrite address uses in calls that were
18241 identified by find_address_like_use.
18243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18244 Alan Hayward <alan.hayward@arm.com>
18245 David Sherwood <david.sherwood@arm.com>
18247 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
18249 * gimple-expr.h (is_gimple_addressable: Likewise.
18250 * gimple-expr.c (is_gimple_address): Likewise.
18251 * internal-fn.c (expand_call_mem_ref): New function.
18252 (expand_mask_load_optab_fn): Use it.
18253 (expand_mask_store_optab_fn): Likewise.
18255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18256 Alan Hayward <alan.hayward@arm.com>
18257 David Sherwood <david.sherwood@arm.com>
18259 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
18260 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
18261 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
18262 (cond_umax@var{mode}): Document.
18263 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
18264 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
18265 (cond_umin_optab, cond_umax_optab): New optabs.
18266 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
18267 (COND_IOR, COND_XOR): New internal functions.
18268 * internal-fn.h (get_conditional_internal_fn): Declare.
18269 * internal-fn.c (cond_binary_direct): New macro.
18270 (expand_cond_binary_optab_fn): Likewise.
18271 (direct_cond_binary_optab_supported_p): Likewise.
18272 (get_conditional_internal_fn): New function.
18273 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
18274 Cope with reduction statements that are vectorized as calls rather
18276 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
18277 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
18278 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
18279 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18280 (UNSPEC_COND_EOR): New unspecs.
18281 (optab): Add mappings for them.
18282 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
18283 (sve_int_op, sve_fp_op): New int attributes.
18285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18286 Alan Hayward <alan.hayward@arm.com>
18287 David Sherwood <david.sherwood@arm.com>
18289 * optabs.def (while_ult_optab): New optab.
18290 * doc/md.texi (while_ult@var{m}@var{n}): Document.
18291 * internal-fn.def (WHILE_ULT): New internal function.
18292 * internal-fn.h (direct_internal_fn_supported_p): New override
18293 that takes two types as argument.
18294 * internal-fn.c (while_direct): New macro.
18295 (expand_while_optab_fn): New function.
18296 (convert_optab_supported_p): Likewise.
18297 (direct_while_optab_supported_p): New macro.
18298 * wide-int.h (wi::udiv_ceil): New function.
18299 * tree-vectorizer.h (rgroup_masks): New structure.
18300 (vec_loop_masks): New typedef.
18301 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
18302 and fully_masked_p.
18303 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
18304 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
18305 (vect_max_vf): New function.
18306 (slpeel_make_loop_iterate_ntimes): Delete.
18307 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
18308 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
18309 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
18310 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
18311 internal-fn.h, stor-layout.h and optabs-query.h.
18312 (vect_set_loop_mask): New function.
18313 (add_preheader_seq): Likewise.
18314 (add_header_seq): Likewise.
18315 (interleave_supported_p): Likewise.
18316 (vect_maybe_permute_loop_masks): Likewise.
18317 (vect_set_loop_masks_directly): Likewise.
18318 (vect_set_loop_condition_masked): Likewise.
18319 (vect_set_loop_condition_unmasked): New function, split out from
18320 slpeel_make_loop_iterate_ntimes.
18321 (slpeel_make_loop_iterate_ntimes): Rename to..
18322 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
18323 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
18324 (vect_do_peeling): Update call accordingly.
18325 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
18327 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18328 mask_compare_type, can_fully_mask_p and fully_masked_p.
18329 (release_vec_loop_masks): New function.
18330 (_loop_vec_info): Use it to free the loop masks.
18331 (can_produce_all_loop_masks_p): New function.
18332 (vect_get_max_nscalars_per_iter): Likewise.
18333 (vect_verify_full_masking): Likewise.
18334 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
18335 retries, and free the mask rgroups before retrying. Check loop-wide
18336 reasons for disallowing fully-masked loops. Make the final decision
18337 about whether use a fully-masked loop or not.
18338 (vect_estimate_min_profitable_iters): Do not assume that peeling
18339 for the number of iterations will be needed for fully-masked loops.
18340 (vectorizable_reduction): Disable fully-masked loops.
18341 (vectorizable_live_operation): Likewise.
18342 (vect_halve_mask_nunits): New function.
18343 (vect_double_mask_nunits): Likewise.
18344 (vect_record_loop_mask): Likewise.
18345 (vect_get_loop_mask): Likewise.
18346 (vect_transform_loop): Handle the case in which the final loop
18347 iteration might handle a partial vector. Call vect_set_loop_condition
18348 instead of slpeel_make_loop_iterate_ntimes.
18349 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
18350 (check_load_store_masking): New function.
18351 (prepare_load_store_mask): Likewise.
18352 (vectorizable_store): Handle fully-masked loops.
18353 (vectorizable_load): Likewise.
18354 (supportable_widening_operation): Use vect_halve_mask_nunits for
18356 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
18357 (vect_gen_while): New function.
18358 * config/aarch64/aarch64.md (umax<mode>3): New expander.
18359 (aarch64_uqdec<mode>): New insn.
18361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18362 Alan Hayward <alan.hayward@arm.com>
18363 David Sherwood <david.sherwood@arm.com>
18365 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
18366 (reduc_xor_scal_optab): New optabs.
18367 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
18368 (reduc_xor_scal_@var{m}): Document.
18369 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
18370 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
18371 internal functions.
18372 * fold-const-call.c (fold_const_call): Handle them.
18373 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
18374 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
18375 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
18376 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
18377 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
18378 (UNSPEC_XORV): New unspecs.
18379 (optab): Add entries for them.
18380 (BITWISEV): New int iterator.
18381 (bit_reduc_op): New int attributes.
18383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18384 Alan Hayward <alan.hayward@arm.com>
18385 David Sherwood <david.sherwood@arm.com>
18387 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
18388 * internal-fn.def (VEC_SHL_INSERT): New internal function.
18389 * optabs.def (vec_shl_insert_optab): New optab.
18390 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
18391 (duplicate_and_interleave): Likewise.
18392 * tree-vect-loop.c: Include internal-fn.h.
18393 (neutral_op_for_slp_reduction): New function, split out from
18394 get_initial_defs_for_reduction.
18395 (get_initial_def_for_reduction): Handle option 2 for variable-length
18396 vectors by loading the neutral value into a vector and then shifting
18397 the initial value into element 0.
18398 (get_initial_defs_for_reduction): Replace the code argument with
18399 the neutral value calculated by neutral_op_for_slp_reduction.
18400 Use gimple_build_vector for constant-length vectors.
18401 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
18402 but the first group_size elements have a neutral value.
18403 Use duplicate_and_interleave otherwise.
18404 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
18405 Update call to get_initial_defs_for_reduction. Handle SLP
18406 reductions for variable-length vectors by creating one vector
18407 result for each scalar result, with the elements associated
18408 with other scalar results stubbed out with the neutral value.
18409 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
18410 Require IFN_VEC_SHL_INSERT for double reductions on
18411 variable-length vectors, or SLP reductions that have
18412 a neutral value. Require can_duplicate_and_interleave_p
18413 support for variable-length unchained SLP reductions if there
18414 is no neutral value, such as for MIN/MAX reductions. Also require
18415 the number of vector elements to be a multiple of the number of
18416 SLP statements when doing variable-length unchained SLP reductions.
18417 Update call to vect_create_epilog_for_reduction.
18418 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
18419 and remove initial values.
18420 (duplicate_and_interleave): Make public.
18421 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
18422 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
18424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18425 Alan Hayward <alan.hayward@arm.com>
18426 David Sherwood <david.sherwood@arm.com>
18428 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
18429 (can_duplicate_and_interleave_p): New function.
18430 (vect_get_and_check_slp_defs): Take the vector of statements
18431 rather than just the current one. Remove excess parentheses.
18432 Restriction rejectinon of vect_constant_def and vect_external_def
18433 for variable-length vectors to boolean types, or types for which
18434 can_duplicate_and_interleave_p is false.
18435 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
18436 (duplicate_and_interleave): New function.
18437 (vect_get_constant_vectors): Use gimple_build_vector for
18438 constant-length vectors and suitable variable-length constant
18439 vectors. Use duplicate_and_interleave for other variable-length
18440 vectors. Don't defer the update when inserting new statements.
18442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18443 Alan Hayward <alan.hayward@arm.com>
18444 David Sherwood <david.sherwood@arm.com>
18446 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
18447 min_profitable_iters doesn't go negative.
18449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18450 Alan Hayward <alan.hayward@arm.com>
18451 David Sherwood <david.sherwood@arm.com>
18453 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
18454 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
18455 * optabs.def (vec_mask_load_lanes_optab): New optab.
18456 (vec_mask_store_lanes_optab): Likewise.
18457 * internal-fn.def (MASK_LOAD_LANES): New internal function.
18458 (MASK_STORE_LANES): Likewise.
18459 * internal-fn.c (mask_load_lanes_direct): New macro.
18460 (mask_store_lanes_direct): Likewise.
18461 (expand_mask_load_optab_fn): Handle masked operations.
18462 (expand_mask_load_lanes_optab_fn): New macro.
18463 (expand_mask_store_optab_fn): Handle masked operations.
18464 (expand_mask_store_lanes_optab_fn): New macro.
18465 (direct_mask_load_lanes_optab_supported_p): Likewise.
18466 (direct_mask_store_lanes_optab_supported_p): Likewise.
18467 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
18469 (vect_load_lanes_supported): Likewise.
18470 * tree-vect-data-refs.c (strip_conversion): New function.
18471 (can_group_stmts_p): Likewise.
18472 (vect_analyze_data_ref_accesses): Use it instead of checking
18473 for a pair of assignments.
18474 (vect_store_lanes_supported): Take a masked_p parameter.
18475 (vect_load_lanes_supported): Likewise.
18476 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
18477 vect_store_lanes_supported and vect_load_lanes_supported.
18478 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
18479 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
18480 parameter. Don't allow gaps for masked accesses.
18481 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
18482 and vect_load_lanes_supported.
18483 (get_load_store_type): Take a masked_p parameter and update
18484 call to get_group_load_store_type.
18485 (vectorizable_store): Update call to get_load_store_type.
18486 Handle IFN_MASK_STORE_LANES.
18487 (vectorizable_load): Update call to get_load_store_type.
18488 Handle IFN_MASK_LOAD_LANES.
18490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18491 Alan Hayward <alan.hayward@arm.com>
18492 David Sherwood <david.sherwood@arm.com>
18494 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
18496 * config/aarch64/aarch64-protos.h
18497 (aarch64_sve_struct_memory_operand_p): Declare.
18498 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
18499 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
18500 (VPRED, vpred): Handle SVE structure modes.
18501 * config/aarch64/constraints.md (Utx): New constraint.
18502 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
18503 (aarch64_sve_struct_nonimmediate_operand): New predicates.
18504 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
18505 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
18506 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
18507 structure modes. Split into pieces after RA.
18508 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
18509 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
18511 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
18512 SVE structure modes.
18513 (aarch64_classify_address): Likewise.
18514 (sizetochar): Move earlier in file.
18515 (aarch64_print_operand): Handle SVE register lists.
18516 (aarch64_array_mode): New function.
18517 (aarch64_sve_struct_memory_operand_p): Likewise.
18518 (TARGET_ARRAY_MODE): Redefine.
18520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18521 Alan Hayward <alan.hayward@arm.com>
18522 David Sherwood <david.sherwood@arm.com>
18524 * target.def (array_mode): New target hook.
18525 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
18526 * doc/tm.texi: Regenerate.
18527 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
18528 * hooks.c (hook_optmode_mode_uhwi_none): New function.
18529 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
18530 targetm.array_mode.
18531 * stor-layout.c (mode_for_array): Likewise. Support polynomial
18534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18535 Alan Hayward <alan.hayward@arm.com>
18536 David Sherwood <david.sherwood@arm.com>
18538 * fold-const.c (fold_binary_loc): Check the argument types
18539 rather than the result type when testing for a vector operation.
18541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18543 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
18544 * doc/tm.texi: Regenerate.
18546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18547 Alan Hayward <alan.hayward@arm.com>
18548 David Sherwood <david.sherwood@arm.com>
18550 * doc/invoke.texi (-msve-vector-bits=): Document new option.
18551 (sve): Document new AArch64 extension.
18552 * doc/md.texi (w): Extend the description of the AArch64
18553 constraint to include SVE vectors.
18554 (Upl, Upa): Document new AArch64 predicate constraints.
18555 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
18557 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
18558 (msve-vector-bits=): New option.
18559 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
18560 SVE when these are disabled.
18561 (sve): New extension.
18562 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
18563 modes. Adjust their number of units based on aarch64_sve_vg.
18564 (MAX_BITSIZE_MODE_ANY_MODE): Define.
18565 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
18566 aarch64_addr_query_type.
18567 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
18568 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
18569 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
18570 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
18571 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
18572 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
18573 (aarch64_simd_imm_zero_p): Delete.
18574 (aarch64_check_zero_based_sve_index_immediate): Declare.
18575 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18576 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18577 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18578 (aarch64_sve_float_mul_immediate_p): Likewise.
18579 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18580 rather than an rtx.
18581 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
18582 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
18583 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
18584 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
18585 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
18586 (aarch64_regmode_natural_size): Likewise.
18587 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
18588 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
18590 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
18591 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
18592 for VG and the SVE predicate registers.
18593 (V_ALIASES): Add a "z"-prefixed alias.
18594 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
18595 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
18596 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
18597 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
18598 (REG_CLASS_NAMES): Add entries for them.
18599 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
18600 and the predicate registers.
18601 (aarch64_sve_vg): Declare.
18602 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
18603 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
18604 (REGMODE_NATURAL_SIZE): Define.
18605 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
18607 * config/aarch64/aarch64.c: Include cfgrtl.h.
18608 (simd_immediate_info): Add a constructor for series vectors,
18609 and an associated step field.
18610 (aarch64_sve_vg): New variable.
18611 (aarch64_dbx_register_number): Handle VG and the predicate registers.
18612 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
18613 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
18614 (VEC_ANY_DATA, VEC_STRUCT): New constants.
18615 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
18616 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
18617 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
18618 (aarch64_get_mask_mode): New functions.
18619 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
18620 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
18621 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
18622 predicate modes and predicate registers. Explicitly restrict
18623 GPRs to modes of 16 bytes or smaller. Only allow FP registers
18624 to store a vector mode if it is recognized by
18625 aarch64_classify_vector_mode.
18626 (aarch64_regmode_natural_size): New function.
18627 (aarch64_hard_regno_caller_save_mode): Return the original mode
18629 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
18630 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
18631 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
18632 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
18634 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
18635 does not overlap dest if the function is frame-related. Handle
18637 (aarch64_split_add_offset): New function.
18638 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
18639 them aarch64_add_offset.
18640 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
18641 and update call to aarch64_sub_sp.
18642 (aarch64_add_cfa_expression): New function.
18643 (aarch64_expand_prologue): Pass extra temporary registers to the
18644 functions above. Handle the case in which we need to emit new
18645 DW_CFA_expressions for registers that were originally saved
18646 relative to the stack pointer, but now have to be expressed
18647 relative to the frame pointer.
18648 (aarch64_output_mi_thunk): Pass extra temporary registers to the
18650 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
18651 IP0 and IP1 values for SVE frames.
18652 (aarch64_expand_vec_series): New function.
18653 (aarch64_expand_sve_widened_duplicate): Likewise.
18654 (aarch64_expand_sve_const_vector): Likewise.
18655 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
18656 Handle SVE constants. Use emit_move_insn to move a force_const_mem
18657 into the register, rather than emitting a SET directly.
18658 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
18659 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
18660 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
18661 (offset_9bit_signed_scaled_p): New functions.
18662 (aarch64_replicate_bitmask_imm): New function.
18663 (aarch64_bitmask_imm): Use it.
18664 (aarch64_cannot_force_const_mem): Reject expressions involving
18665 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
18666 (aarch64_classify_index): Handle SVE indices, by requiring
18667 a plain register index with a scale that matches the element size.
18668 (aarch64_classify_address): Handle SVE addresses. Assert that
18669 the mode of the address is VOIDmode or an integer mode.
18670 Update call to aarch64_classify_symbol.
18671 (aarch64_classify_symbolic_expression): Update call to
18672 aarch64_classify_symbol.
18673 (aarch64_const_vec_all_in_range_p): New function.
18674 (aarch64_print_vector_float_operand): Likewise.
18675 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
18676 "vN" for FP registers with SVE modes. Handle (const ...) vectors
18677 and the FP immediates 1.0 and 0.5.
18678 (aarch64_print_address_internal): Handle SVE addresses.
18679 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
18680 (aarch64_regno_regclass): Handle predicate registers.
18681 (aarch64_secondary_reload): Handle big-endian reloads of SVE
18683 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
18684 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
18685 (aarch64_convert_sve_vector_bits): New function.
18686 (aarch64_override_options): Use it to handle -msve-vector-bits=.
18687 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18688 rather than an rtx.
18689 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
18690 Handle SVE vector and predicate modes. Accept VL-based constants
18691 that need only one temporary register, and VL offsets that require
18692 no temporary registers.
18693 (aarch64_conditional_register_usage): Mark the predicate registers
18694 as fixed if SVE isn't available.
18695 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
18696 Return true for SVE vector and predicate modes.
18697 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
18698 rather than an unsigned int. Handle SVE modes.
18699 (aarch64_preferred_simd_mode): Update call accordingly. Handle
18701 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
18703 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18704 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18705 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18706 (aarch64_sve_float_mul_immediate_p): New functions.
18707 (aarch64_sve_valid_immediate): New function.
18708 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
18709 Explicitly reject structure modes. Check for INDEX constants.
18710 Handle PTRUE and PFALSE constants.
18711 (aarch64_check_zero_based_sve_index_immediate): New function.
18712 (aarch64_simd_imm_zero_p): Delete.
18713 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
18714 vector modes. Accept constants in the range of CNT[BHWD].
18715 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
18716 ask for an Advanced SIMD mode.
18717 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
18718 (aarch64_simd_vector_alignment): Handle SVE predicates.
18719 (aarch64_vectorize_preferred_vector_alignment): New function.
18720 (aarch64_simd_vector_alignment_reachable): Use it instead of
18722 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
18723 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
18725 (MAX_VECT_LEN): Delete.
18726 (expand_vec_perm_d): Add a vec_flags field.
18727 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
18728 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
18729 (aarch64_evpc_ext): Don't apply a big-endian lane correction
18731 (aarch64_evpc_rev): Rename to...
18732 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
18733 (aarch64_evpc_rev_global): New function.
18734 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
18735 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
18737 (aarch64_evpc_sve_tbl): New function.
18738 (aarch64_expand_vec_perm_const_1): Update after rename of
18739 aarch64_evpc_rev. Handle SVE permutes too, trying
18740 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
18741 than aarch64_evpc_tbl.
18742 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
18743 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
18744 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
18745 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
18746 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
18747 (aarch64_expand_sve_vcond): New functions.
18748 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
18749 of aarch64_vector_mode_p.
18750 (aarch64_dwarf_poly_indeterminate_value): New function.
18751 (aarch64_compute_pressure_classes): Likewise.
18752 (aarch64_can_change_mode_class): Likewise.
18753 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
18754 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
18755 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
18756 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
18757 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
18758 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
18759 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
18760 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
18762 (Dn, Dl, Dr): Accept const as well as const_vector.
18763 (Dz): Likewise. Compare against CONST0_RTX.
18764 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
18765 of "vector" where appropriate.
18766 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18767 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18768 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18769 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18770 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18771 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18772 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18773 (v_int_equiv): Extend to SVE modes.
18774 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18776 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18777 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18778 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18779 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18780 (SVE_COND_FP_CMP): New int iterators.
18781 (perm_hilo): Handle the new unpack unspecs.
18782 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18784 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18785 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18786 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18787 (aarch64_equality_operator, aarch64_constant_vector_operand)
18788 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18789 (aarch64_sve_nonimmediate_operand): Likewise.
18790 (aarch64_sve_general_operand): Likewise.
18791 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18792 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18793 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18794 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18795 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18796 (aarch64_sve_float_arith_immediate): Likewise.
18797 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18798 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18799 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18800 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18801 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18802 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18803 (aarch64_sve_float_arith_operand): Likewise.
18804 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18805 (aarch64_sve_float_mul_operand): Likewise.
18806 (aarch64_sve_vec_perm_operand): Likewise.
18807 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18808 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18809 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18810 as well as const_vector.
18811 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18812 in file. Use CONST0_RTX and CONSTM1_RTX.
18813 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18814 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18815 Use aarch64_simd_imm_zero.
18816 * config/aarch64/aarch64-sve.md: New file.
18817 * config/aarch64/aarch64.md: Include it.
18818 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18819 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18820 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18821 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18822 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18823 (sve): New attribute.
18824 (enabled): Disable instructions with the sve attribute unless
18826 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18827 aarch64_expand_mov_immediate.
18828 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18829 CNT[BHSD] immediates.
18830 (movti): Split CONST_POLY_INT moves into two halves.
18831 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18832 Split additions that need a temporary here if the destination
18833 is the stack pointer.
18834 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18835 (*add<mode>3_poly_1): New instruction.
18836 (set_clobber_cc): New expander.
18838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18840 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18841 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18842 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18843 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18844 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18845 Change innermode from fixed_mode_size to machine_mode.
18846 (simplify_subreg): Update call accordingly. Handle a constant-sized
18847 subreg of a variable-length CONST_VECTOR.
18849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18850 Alan Hayward <alan.hayward@arm.com>
18851 David Sherwood <david.sherwood@arm.com>
18853 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18854 (add_offset_to_base): New function, split out from...
18855 (create_mem_ref): ...here. When handling a scale other than 1,
18856 check first whether the address is valid without the offset.
18857 Add it into the base if so, leaving the index and scale as-is.
18859 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18862 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18863 fold_for_warn before checking if arg2 is INTEGER_CST.
18865 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18867 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18868 (store_multiple_operation): Delete.
18869 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18870 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18871 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18872 guarded by TARGET_STRING.
18873 (rs6000_output_load_multiple): Delete.
18874 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18875 OPTION_MASK_STRING / TARGET_STRING handling.
18876 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18877 (const rs6000_opt_masks) <"string">: Change mask to 0.
18878 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18879 (MASK_STRING): Delete.
18880 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18882 (load_multiple): Delete.
18889 (store_multiple): Delete.
18896 (movmemsi_8reg): Delete.
18897 (corresponding unnamed define_insn): Delete.
18898 (movmemsi_6reg): Delete.
18899 (corresponding unnamed define_insn): Delete.
18900 (movmemsi_4reg): Delete.
18901 (corresponding unnamed define_insn): Delete.
18902 (movmemsi_2reg): Delete.
18903 (corresponding unnamed define_insn): Delete.
18904 (movmemsi_1reg): Delete.
18905 (corresponding unnamed define_insn): Delete.
18906 * config/rs6000/rs6000.opt (mno-string): New.
18907 (mstring): Replace by deprecation warning stub.
18908 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18910 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18912 * regrename.c (regrename_do_replace): If replacing the same
18913 reg multiple times, try to reuse last created gen_raw_REG.
18916 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18917 main to workaround a bug in GDB.
18919 2018-01-12 Tom de Vries <tom@codesourcery.com>
18922 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18924 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18926 PR rtl-optimization/80481
18927 * ira-color.c (get_cap_member): New function.
18928 (allocnos_conflict_by_live_ranges_p): Use it.
18929 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18930 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18932 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18935 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18936 (*saddl_se_1): Ditto.
18937 (*ssubsi_1): Ditto.
18938 (*ssubl_se_1): Ditto.
18940 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18942 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18943 rather than wi::to_widest for DR_INITs.
18944 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18945 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18946 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18948 (vect_analyze_group_access_1): Note that here.
18950 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18952 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18953 polynomial type sizes.
18955 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18957 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18958 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18959 (gimple_add_tmp_var): Likewise.
18961 2018-01-12 Martin Liska <mliska@suse.cz>
18963 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18964 (gimple_alloc_sizes): Likewise.
18965 (dump_gimple_statistics): Use PRIu64 in printf format.
18966 * gimple.h: Change uint64_t to int.
18968 2018-01-12 Martin Liska <mliska@suse.cz>
18970 * tree-core.h: Use uint64_t instead of int.
18971 * tree.c (tree_node_counts): Likewise.
18972 (tree_node_sizes): Likewise.
18973 (dump_tree_statistics): Use PRIu64 in printf format.
18975 2018-01-12 Martin Liska <mliska@suse.cz>
18977 * Makefile.in: As qsort_chk is implemented in vec.c, add
18978 vec.o to linkage of gencfn-macros.
18979 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18980 passing the info to record_node_allocation_statistics.
18981 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18983 * ggc-common.c (struct ggc_usage): Add operator== and use
18984 it in operator< and compare function.
18985 * mem-stats.h (struct mem_usage): Likewise.
18986 * vec.c (struct vec_usage): Remove operator< and compare
18987 function. Can be simply inherited.
18989 2018-01-12 Martin Jambor <mjambor@suse.cz>
18992 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18993 * tree-ssa-math-opts.c: Include domwalk.h.
18994 (convert_mult_to_fma_1): New function.
18995 (fma_transformation_info): New type.
18996 (fma_deferring_state): Likewise.
18997 (cancel_fma_deferring): New function.
18998 (result_of_phi): Likewise.
18999 (last_fma_candidate_feeds_initial_phi): Likewise.
19000 (convert_mult_to_fma): Added deferring logic, split actual
19001 transformation to convert_mult_to_fma_1.
19002 (math_opts_dom_walker): New type.
19003 (math_opts_dom_walker::after_dom_children): New method, body moved
19004 here from pass_optimize_widening_mul::execute, added deferring logic
19006 (pass_optimize_widening_mul::execute): Moved most of code to
19007 math_opts_dom_walker::after_dom_children.
19008 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
19009 * config/i386/i386.c (ix86_option_override_internal): Added
19010 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
19012 2018-01-12 Richard Biener <rguenther@suse.de>
19015 * dwarf2out.c (gen_variable_die): Do not reset old_die for
19016 inline instance vars.
19018 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
19021 * config/rx/rx.c (rx_is_restricted_memory_address):
19022 Handle SUBREG case.
19024 2018-01-12 Richard Biener <rguenther@suse.de>
19026 PR tree-optimization/80846
19027 * target.def (split_reduction): New target hook.
19028 * targhooks.c (default_split_reduction): New function.
19029 * targhooks.h (default_split_reduction): Declare.
19030 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
19031 target requests first reduce vectors by combining low and high
19033 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
19034 (get_vectype_for_scalar_type_and_size): Export.
19035 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
19036 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
19037 * doc/tm.texi: Regenerate.
19038 * config/i386/i386.c (ix86_split_reduction): Implement
19039 TARGET_VECTORIZE_SPLIT_REDUCTION.
19041 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19044 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
19045 in PIC mode except for TARGET_VXWORKS_RTP.
19046 * config/sparc/sparc.c: Include cfgrtl.h.
19047 (TARGET_INIT_PIC_REG): Define.
19048 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
19049 (sparc_pic_register_p): New predicate.
19050 (sparc_legitimate_address_p): Use it.
19051 (sparc_legitimize_pic_address): Likewise.
19052 (sparc_delegitimize_address): Likewise.
19053 (sparc_mode_dependent_address_p): Likewise.
19054 (gen_load_pcrel_sym): Remove 4th parameter.
19055 (load_got_register): Adjust call to above. Remove obsolete stuff.
19056 (sparc_expand_prologue): Do not call load_got_register here.
19057 (sparc_flat_expand_prologue): Likewise.
19058 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
19059 (sparc_use_pseudo_pic_reg): New function.
19060 (sparc_init_pic_reg): Likewise.
19061 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19062 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
19064 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
19066 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
19067 Add item for branch_cost.
19069 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19071 PR rtl-optimization/83565
19072 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
19073 not extend the result to a larger mode for rotate operations.
19074 (num_sign_bit_copies1): Likewise.
19076 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19079 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
19081 Use values-Xc.o for -pedantic.
19082 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
19084 2018-01-12 Martin Liska <mliska@suse.cz>
19087 * ipa-devirt.c (final_warning_record::grow_type_warnings):
19089 (possible_polymorphic_call_targets): Use it.
19090 (ipa_devirt): Likewise.
19092 2018-01-12 Martin Liska <mliska@suse.cz>
19094 * profile-count.h (enum profile_quality): Use 0 as invalid
19095 enum value of profile_quality.
19097 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
19099 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
19100 -mext-string options.
19102 2018-01-12 Richard Biener <rguenther@suse.de>
19104 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
19105 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
19106 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
19108 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
19110 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
19112 * configure.ac (--with-long-double-format): Add support for the
19113 configuration option to change the default long double format on
19115 * config.gcc (powerpc*-linux*-*): Likewise.
19116 * configure: Regenerate.
19117 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
19118 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
19119 used without modification.
19121 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19123 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
19124 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
19125 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
19126 MISC_BUILTIN_SPEC_BARRIER.
19127 (rs6000_init_builtins): Likewise.
19128 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
19130 (speculation_barrier): New define_insn.
19131 * doc/extend.texi: Document __builtin_speculation_barrier.
19133 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19136 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
19137 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
19138 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
19140 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
19141 integral modes instead of "ss" and "sd".
19142 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
19143 vectors with 32-bit and 64-bit elements.
19144 (vecdupssescalarmodesuffix): New mode attribute.
19145 (vec_dup<mode>): Use it.
19147 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
19150 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
19151 frame if argument is passed on stack.
19153 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19156 * ree.c (combine_reaching_defs): Optimize also
19157 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
19158 reg2=any_extend(exp); reg1=reg2;, formatting fix.
19160 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19162 PR middle-end/83189
19163 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
19165 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19167 PR middle-end/83718
19168 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
19169 after they are computed.
19171 2018-01-11 Bin Cheng <bin.cheng@arm.com>
19173 PR tree-optimization/83695
19174 * gimple-loop-linterchange.cc
19175 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
19176 reset cached scev information after interchange.
19177 (pass_linterchange::execute): Remove call to scev_reset_htab.
19179 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19181 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
19182 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
19183 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
19184 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
19185 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
19186 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
19187 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
19188 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
19189 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
19190 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
19191 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
19192 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
19193 (V_lane_reg): Likewise.
19194 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
19196 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
19197 (vfmal_lane_low<mode>_intrinsic,
19198 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
19199 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
19200 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
19201 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
19202 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
19203 vfmsl_lane_high<mode>_intrinsic): New define_insns.
19205 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19207 * config/arm/arm-cpus.in (fp16fml): New feature.
19208 (ALL_SIMD): Add fp16fml.
19209 (armv8.2-a): Add fp16fml as an option.
19210 (armv8.3-a): Likewise.
19211 (armv8.4-a): Add fp16fml as part of fp16.
19212 * config/arm/arm.h (TARGET_FP16FML): Define.
19213 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
19215 * config/arm/arm-modes.def (V2HF): Define.
19216 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
19217 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
19218 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
19219 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
19220 vfmsl_low, vfmsl_high): New set of builtins.
19221 * config/arm/iterators.md (PLUSMINUS): New code iterator.
19222 (vfml_op): New code attribute.
19223 (VFMLHALVES): New int iterator.
19224 (VFML, VFMLSEL): New mode attributes.
19225 (V_reg): Define mapping for V2HF.
19226 (V_hi, V_lo): New mode attributes.
19227 (VF_constraint): Likewise.
19228 (vfml_half, vfml_half_selector): New int attributes.
19229 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
19231 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
19232 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
19234 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
19235 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
19236 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
19237 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
19239 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
19240 Document new effective target and option set.
19242 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19244 * config/arm/arm-cpus.in (armv8_4): New feature.
19245 (ARMv8_4a): New fgroup.
19246 (armv8.4-a): New arch.
19247 * config/arm/arm-tables.opt: Regenerate.
19248 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
19249 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
19250 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
19251 Add matching rules for -march=armv8.4-a and extensions.
19252 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
19254 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
19257 * config/rx/rx.md (BW): New mode attribute.
19258 (sync_lock_test_and_setsi): Add mode suffix to insn output.
19260 2018-01-11 Richard Biener <rguenther@suse.de>
19262 PR tree-optimization/83435
19263 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
19264 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
19265 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
19267 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19268 Alan Hayward <alan.hayward@arm.com>
19269 David Sherwood <david.sherwood@arm.com>
19271 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
19273 (aarch64_classify_address): Initialize it. Track polynomial offsets.
19274 (aarch64_print_address_internal): Use it to check for a zero offset.
19276 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19277 Alan Hayward <alan.hayward@arm.com>
19278 David Sherwood <david.sherwood@arm.com>
19280 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
19281 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
19282 Return a poly_int64 rather than a HOST_WIDE_INT.
19283 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
19284 rather than a HOST_WIDE_INT.
19285 * config/aarch64/aarch64.h (aarch64_frame): Protect with
19286 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
19287 hard_fp_offset, frame_size, initial_adjust, callee_offset and
19288 final_offset from HOST_WIDE_INT to poly_int64.
19289 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
19290 to_constant when getting the number of units in an Advanced SIMD
19292 (aarch64_builtin_vectorized_function): Check for a constant number
19294 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
19296 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
19297 attribute instead of GET_MODE_NUNITS.
19298 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
19299 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
19300 GET_MODE_SIZE for fixed-size registers.
19301 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
19302 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
19303 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
19304 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
19305 (aarch64_print_operand, aarch64_print_address_internal)
19306 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
19307 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
19308 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
19309 Handle polynomial GET_MODE_SIZE.
19310 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
19311 wider than SImode without modification.
19312 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
19313 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
19314 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
19315 passing and returning SVE modes.
19316 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
19317 rather than GEN_INT.
19318 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
19319 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
19320 (aarch64_allocate_and_probe_stack_space): Likewise.
19321 (aarch64_layout_frame): Cope with polynomial offsets.
19322 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
19323 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
19324 polynomial offsets.
19325 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
19326 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
19327 poly_int64 rather than a HOST_WIDE_INT.
19328 (aarch64_get_separate_components, aarch64_process_components)
19329 (aarch64_expand_prologue, aarch64_expand_epilogue)
19330 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
19331 (aarch64_anchor_offset): New function, split out from...
19332 (aarch64_legitimize_address): ...here.
19333 (aarch64_builtin_vectorization_cost): Handle polynomial
19334 TYPE_VECTOR_SUBPARTS.
19335 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
19337 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
19338 number of elements from the PARALLEL rather than the mode.
19339 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
19340 rather than GET_MODE_BITSIZE.
19341 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
19342 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
19343 (aarch64_expand_vec_perm_const_1): Handle polynomial
19344 d->perm.length () and d->perm elements.
19345 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
19346 Apply to_constant to d->perm elements.
19347 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
19348 polynomial CONST_VECTOR_NUNITS.
19349 (aarch64_move_pointer): Take amount as a poly_int64 rather
19351 (aarch64_progress_pointer): Avoid temporary variable.
19352 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
19353 the mode attribute instead of GET_MODE.
19355 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19356 Alan Hayward <alan.hayward@arm.com>
19357 David Sherwood <david.sherwood@arm.com>
19359 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
19360 x exists before using it.
19361 (aarch64_add_constant_internal): Rename to...
19362 (aarch64_add_offset_1): ...this. Replace regnum with separate
19363 src and dest rtxes. Handle the case in which they're different,
19364 including when the offset is zero. Replace scratchreg with an rtx.
19365 Use 2 additions if there is no spare register into which we can
19366 move a 16-bit constant.
19367 (aarch64_add_constant): Delete.
19368 (aarch64_add_offset): Replace reg with separate src and dest
19369 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
19370 Use aarch64_add_offset_1.
19371 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
19372 an rtx rather than an int. Take the delta as a poly_int64
19373 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
19374 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
19375 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
19376 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
19377 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
19378 and aarch64_add_sp.
19379 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
19380 aarch64_add_constant.
19382 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19384 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
19385 Use scalar_float_mode.
19387 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19389 * config/aarch64/aarch64-simd.md
19390 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
19391 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
19392 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
19393 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
19394 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
19395 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
19396 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
19397 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
19398 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
19399 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
19401 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19404 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
19405 targ_options->x_arm_arch_string is non NULL.
19407 2018-01-11 Tamar Christina <tamar.christina@arm.com>
19409 * config/aarch64/aarch64.h
19410 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
19412 2018-01-11 Sudakshina Das <sudi.das@arm.com>
19415 * expmed.c (emit_store_flag_force): Swap if const op0
19416 and change VOIDmode to mode of op0.
19418 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19420 PR rtl-optimization/83761
19421 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
19422 than bytes to mode_for_size.
19424 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19426 PR middle-end/83189
19427 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
19428 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
19431 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19433 PR middle-end/83575
19434 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
19435 when in layout mode.
19436 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
19437 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
19440 2018-01-10 Michael Collison <michael.collison@arm.com>
19442 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
19443 * config/aarch64/aarch64-option-extension.def: Add
19444 AARCH64_OPT_EXTENSION of 'fp16fml'.
19445 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19446 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
19447 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
19448 * config/aarch64/constraints.md (Ui7): New constraint.
19449 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
19450 (VFMLA_SEL_W): Ditto.
19453 (VFMLA16_LOW): New int iterator.
19454 (VFMLA16_HIGH): Ditto.
19455 (UNSPEC_FMLAL): New unspec.
19456 (UNSPEC_FMLSL): Ditto.
19457 (UNSPEC_FMLAL2): Ditto.
19458 (UNSPEC_FMLSL2): Ditto.
19459 (f16mac): New code attribute.
19460 * config/aarch64/aarch64-simd-builtins.def
19461 (aarch64_fmlal_lowv2sf): Ditto.
19462 (aarch64_fmlsl_lowv2sf): Ditto.
19463 (aarch64_fmlalq_lowv4sf): Ditto.
19464 (aarch64_fmlslq_lowv4sf): Ditto.
19465 (aarch64_fmlal_highv2sf): Ditto.
19466 (aarch64_fmlsl_highv2sf): Ditto.
19467 (aarch64_fmlalq_highv4sf): Ditto.
19468 (aarch64_fmlslq_highv4sf): Ditto.
19469 (aarch64_fmlal_lane_lowv2sf): Ditto.
19470 (aarch64_fmlsl_lane_lowv2sf): Ditto.
19471 (aarch64_fmlal_laneq_lowv2sf): Ditto.
19472 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
19473 (aarch64_fmlalq_lane_lowv4sf): Ditto.
19474 (aarch64_fmlsl_lane_lowv4sf): Ditto.
19475 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
19476 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
19477 (aarch64_fmlal_lane_highv2sf): Ditto.
19478 (aarch64_fmlsl_lane_highv2sf): Ditto.
19479 (aarch64_fmlal_laneq_highv2sf): Ditto.
19480 (aarch64_fmlsl_laneq_highv2sf): Ditto.
19481 (aarch64_fmlalq_lane_highv4sf): Ditto.
19482 (aarch64_fmlsl_lane_highv4sf): Ditto.
19483 (aarch64_fmlalq_laneq_highv4sf): Ditto.
19484 (aarch64_fmlsl_laneq_highv4sf): Ditto.
19485 * config/aarch64/aarch64-simd.md:
19486 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
19487 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19488 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
19489 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19490 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
19491 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
19492 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
19493 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
19494 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
19495 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
19496 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
19497 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
19498 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
19499 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
19500 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
19501 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
19502 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
19503 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
19504 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
19505 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
19506 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
19507 (vfmlsl_low_u32): Ditto.
19508 (vfmlalq_low_u32): Ditto.
19509 (vfmlslq_low_u32): Ditto.
19510 (vfmlal_high_u32): Ditto.
19511 (vfmlsl_high_u32): Ditto.
19512 (vfmlalq_high_u32): Ditto.
19513 (vfmlslq_high_u32): Ditto.
19514 (vfmlal_lane_low_u32): Ditto.
19515 (vfmlsl_lane_low_u32): Ditto.
19516 (vfmlal_laneq_low_u32): Ditto.
19517 (vfmlsl_laneq_low_u32): Ditto.
19518 (vfmlalq_lane_low_u32): Ditto.
19519 (vfmlslq_lane_low_u32): Ditto.
19520 (vfmlalq_laneq_low_u32): Ditto.
19521 (vfmlslq_laneq_low_u32): Ditto.
19522 (vfmlal_lane_high_u32): Ditto.
19523 (vfmlsl_lane_high_u32): Ditto.
19524 (vfmlal_laneq_high_u32): Ditto.
19525 (vfmlsl_laneq_high_u32): Ditto.
19526 (vfmlalq_lane_high_u32): Ditto.
19527 (vfmlslq_lane_high_u32): Ditto.
19528 (vfmlalq_laneq_high_u32): Ditto.
19529 (vfmlslq_laneq_high_u32): Ditto.
19530 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
19531 (AARCH64_FL_FOR_ARCH8_4): New.
19532 (AARCH64_ISA_F16FML): New ISA flag.
19533 (TARGET_F16FML): New feature flag for fp16fml.
19534 (doc/invoke.texi): Document new fp16fml option.
19536 2018-01-10 Michael Collison <michael.collison@arm.com>
19538 * config/aarch64/aarch64-builtins.c:
19539 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
19540 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19541 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
19542 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
19543 (AARCH64_ISA_SHA3): New ISA flag.
19544 (TARGET_SHA3): New feature flag for sha3.
19545 * config/aarch64/iterators.md (sha512_op): New int attribute.
19546 (CRYPTO_SHA512): New int iterator.
19547 (UNSPEC_SHA512H): New unspec.
19548 (UNSPEC_SHA512H2): Ditto.
19549 (UNSPEC_SHA512SU0): Ditto.
19550 (UNSPEC_SHA512SU1): Ditto.
19551 * config/aarch64/aarch64-simd-builtins.def
19552 (aarch64_crypto_sha512hqv2di): New builtin.
19553 (aarch64_crypto_sha512h2qv2di): Ditto.
19554 (aarch64_crypto_sha512su0qv2di): Ditto.
19555 (aarch64_crypto_sha512su1qv2di): Ditto.
19556 (aarch64_eor3qv8hi): Ditto.
19557 (aarch64_rax1qv2di): Ditto.
19558 (aarch64_xarqv2di): Ditto.
19559 (aarch64_bcaxqv8hi): Ditto.
19560 * config/aarch64/aarch64-simd.md:
19561 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
19562 (aarch64_crypto_sha512su0qv2di): Ditto.
19563 (aarch64_crypto_sha512su1qv2di): Ditto.
19564 (aarch64_eor3qv8hi): Ditto.
19565 (aarch64_rax1qv2di): Ditto.
19566 (aarch64_xarqv2di): Ditto.
19567 (aarch64_bcaxqv8hi): Ditto.
19568 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
19569 (vsha512h2q_u64): Ditto.
19570 (vsha512su0q_u64): Ditto.
19571 (vsha512su1q_u64): Ditto.
19572 (veor3q_u16): Ditto.
19573 (vrax1q_u64): Ditto.
19574 (vxarq_u64): Ditto.
19575 (vbcaxq_u16): Ditto.
19576 * config/arm/types.md (crypto_sha512): New type attribute.
19577 (crypto_sha3): Ditto.
19578 (doc/invoke.texi): Document new sha3 option.
19580 2018-01-10 Michael Collison <michael.collison@arm.com>
19582 * config/aarch64/aarch64-builtins.c:
19583 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
19584 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19585 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
19586 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
19587 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
19588 (AARCH64_ISA_SM4): New ISA flag.
19589 (TARGET_SM4): New feature flag for sm4.
19590 * config/aarch64/aarch64-simd-builtins.def
19591 (aarch64_sm3ss1qv4si): Ditto.
19592 (aarch64_sm3tt1aq4si): Ditto.
19593 (aarch64_sm3tt1bq4si): Ditto.
19594 (aarch64_sm3tt2aq4si): Ditto.
19595 (aarch64_sm3tt2bq4si): Ditto.
19596 (aarch64_sm3partw1qv4si): Ditto.
19597 (aarch64_sm3partw2qv4si): Ditto.
19598 (aarch64_sm4eqv4si): Ditto.
19599 (aarch64_sm4ekeyqv4si): Ditto.
19600 * config/aarch64/aarch64-simd.md:
19601 (aarch64_sm3ss1qv4si): Ditto.
19602 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
19603 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
19604 (aarch64_sm4eqv4si): Ditto.
19605 (aarch64_sm4ekeyqv4si): Ditto.
19606 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
19607 (sm3part_op): Ditto.
19608 (CRYPTO_SM3TT): Ditto.
19609 (CRYPTO_SM3PART): Ditto.
19610 (UNSPEC_SM3SS1): New unspec.
19611 (UNSPEC_SM3TT1A): Ditto.
19612 (UNSPEC_SM3TT1B): Ditto.
19613 (UNSPEC_SM3TT2A): Ditto.
19614 (UNSPEC_SM3TT2B): Ditto.
19615 (UNSPEC_SM3PARTW1): Ditto.
19616 (UNSPEC_SM3PARTW2): Ditto.
19617 (UNSPEC_SM4E): Ditto.
19618 (UNSPEC_SM4EKEY): Ditto.
19619 * config/aarch64/constraints.md (Ui2): New constraint.
19620 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
19621 * config/arm/types.md (crypto_sm3): New type attribute.
19622 (crypto_sm4): Ditto.
19623 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
19624 (vsm3tt1aq_u32): Ditto.
19625 (vsm3tt1bq_u32): Ditto.
19626 (vsm3tt2aq_u32): Ditto.
19627 (vsm3tt2bq_u32): Ditto.
19628 (vsm3partw1q_u32): Ditto.
19629 (vsm3partw2q_u32): Ditto.
19630 (vsm4eq_u32): Ditto.
19631 (vsm4ekeyq_u32): Ditto.
19632 (doc/invoke.texi): Document new sm4 option.
19634 2018-01-10 Michael Collison <michael.collison@arm.com>
19636 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
19637 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
19638 (AARCH64_FL_FOR_ARCH8_4): New.
19639 (AARCH64_FL_V8_4): New flag.
19640 (doc/invoke.texi): Document new armv8.4-a option.
19642 2018-01-10 Michael Collison <michael.collison@arm.com>
19644 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19645 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
19646 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
19647 * config/aarch64/aarch64-option-extension.def: Add
19648 AARCH64_OPT_EXTENSION of 'sha2'.
19649 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
19650 (crypto): Disable sha2 and aes if crypto disabled.
19651 (crypto): Enable aes and sha2 if enabled.
19652 (simd): Disable sha2 and aes if simd disabled.
19653 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
19655 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
19656 (TARGET_SHA2): New feature flag for sha2.
19657 (TARGET_AES): New feature flag for aes.
19658 * config/aarch64/aarch64-simd.md:
19659 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
19660 conditional on TARGET_AES.
19661 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
19662 (aarch64_crypto_sha1hsi): Make pattern conditional
19664 (aarch64_crypto_sha1hv4si): Ditto.
19665 (aarch64_be_crypto_sha1hv4si): Ditto.
19666 (aarch64_crypto_sha1su1v4si): Ditto.
19667 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
19668 (aarch64_crypto_sha1su0v4si): Ditto.
19669 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
19670 (aarch64_crypto_sha256su0v4si): Ditto.
19671 (aarch64_crypto_sha256su1v4si): Ditto.
19672 (doc/invoke.texi): Document new aes and sha2 options.
19674 2018-01-10 Martin Sebor <msebor@redhat.com>
19676 PR tree-optimization/83781
19677 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
19680 2018-01-11 Martin Sebor <msebor@gmail.com>
19681 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19683 PR tree-optimization/83501
19684 PR tree-optimization/81703
19686 * tree-ssa-strlen.c (get_string_cst): Rename...
19687 (get_string_len): ...to this. Handle global constants.
19688 (handle_char_store): Adjust.
19690 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
19691 Jim Wilson <jimw@sifive.com>
19693 * config/riscv/riscv-protos.h (riscv_output_return): New.
19694 * config/riscv/riscv.c (struct machine_function): New naked_p field.
19695 (riscv_attribute_table, riscv_output_return),
19696 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
19697 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
19698 (riscv_compute_frame_info): Only compute frame->mask if not a naked
19700 (riscv_expand_prologue): Add early return for naked function.
19701 (riscv_expand_epilogue): Likewise.
19702 (riscv_function_ok_for_sibcall): Return false for naked function.
19703 (riscv_set_current_function): New.
19704 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
19705 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
19706 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
19707 * doc/extend.texi (RISC-V Function Attributes): New.
19709 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19711 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
19712 check for 128-bit long double before checking TCmode.
19713 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
19714 128-bit long doubles before checking TFmode or TCmode.
19715 (FLOAT128_IBM_P): Likewise.
19717 2018-01-10 Martin Sebor <msebor@redhat.com>
19719 PR tree-optimization/83671
19720 * builtins.c (c_strlen): Unconditionally return zero for the empty
19722 Use -Warray-bounds for warnings.
19723 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
19724 for non-constant array indices with COMPONENT_REF, arrays of
19725 arrays, and pointers to arrays.
19726 (gimple_fold_builtin_strlen): Determine and set length range for
19727 non-constant character arrays.
19729 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
19731 PR middle-end/81897
19732 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
19735 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
19737 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
19739 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19742 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
19743 VECTOR_MEM_ALTIVEC_OR_VSX_P.
19744 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
19745 indexed_or_indirect_operand predicate.
19746 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
19747 (*vsx_le_perm_load_v8hi): Likewise.
19748 (*vsx_le_perm_load_v16qi): Likewise.
19749 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
19750 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
19751 (*vsx_le_perm_store_v8hi): Likewise.
19752 (*vsx_le_perm_store_v16qi): Likewise.
19753 (eight unnamed splitters): Likewise.
19755 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19757 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
19758 * config/rs6000/emmintrin.h: Likewise.
19759 * config/rs6000/mmintrin.h: Likewise.
19760 * config/rs6000/xmmintrin.h: Likewise.
19762 2018-01-10 David Malcolm <dmalcolm@redhat.com>
19765 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19767 * tree.c (tree_nop_conversion): Return true for location wrapper
19769 (maybe_wrap_with_location): New function.
19770 (selftest::check_strip_nops): New function.
19771 (selftest::test_location_wrappers): New function.
19772 (selftest::tree_c_tests): Call it.
19773 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19774 (maybe_wrap_with_location): New decl.
19775 (EXPR_LOCATION_WRAPPER_P): New macro.
19776 (location_wrapper_p): New inline function.
19777 (tree_strip_any_location_wrapper): New inline function.
19779 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19782 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19783 stack_realign_offset for the largest alignment of stack slot
19785 (ix86_find_max_used_stack_alignment): New function.
19786 (ix86_finalize_stack_frame_flags): Use it. Set
19787 max_used_stack_alignment if we don't realign stack.
19788 * config/i386/i386.h (machine_function): Add
19789 max_used_stack_alignment.
19791 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19793 * config/arm/arm.opt (-mbranch-cost): New option.
19794 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19797 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19800 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19801 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19803 2018-01-10 Richard Biener <rguenther@suse.de>
19806 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19807 early out so it also covers the case where we have a non-NULL
19810 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19812 PR tree-optimization/83753
19813 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19814 for non-strided grouped accesses if the number of elements is 1.
19816 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19819 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19820 * i386.h (TARGET_USE_GATHER): Define.
19821 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19823 2018-01-10 Martin Liska <mliska@suse.cz>
19826 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19827 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19829 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19830 CLEANUP_NO_PARTITIONING is not set.
19832 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19834 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19835 for vectors, as a partial revert of r254296.
19836 * rtl.h (const_vec_p): Delete.
19837 (const_vec_duplicate_p): Don't test for vector CONSTs.
19838 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19839 * expmed.c (make_tree): Likewise.
19842 * common.md (E, F): Use CONSTANT_P instead of checking for
19844 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19845 checking for CONST_VECTOR.
19847 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19849 PR middle-end/83575
19850 * predict.c (force_edge_cold): Handle in more sane way edges
19851 with no prediction.
19853 2018-01-09 Carl Love <cel@us.ibm.com>
19855 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19857 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19858 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19859 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19860 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19861 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19862 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19863 * config/rs6000/rs6000-protos.h: Add extern defition for
19864 rs6000_generate_float2_double_code.
19865 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19867 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19868 (float2_v2df): Add define_expand.
19870 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19873 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19874 op_mode in the force_to_mode call.
19876 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19878 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19879 instead of checking each element individually.
19880 (aarch64_evpc_uzp): Likewise.
19881 (aarch64_evpc_zip): Likewise.
19882 (aarch64_evpc_ext): Likewise.
19883 (aarch64_evpc_rev): Likewise.
19884 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19885 instead of checking each element individually. Return true without
19887 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19888 whether all selected elements come from the same input, instead of
19889 checking each element individually. Remove calls to gen_rtx_REG,
19890 start_sequence and end_sequence and instead assert that no rtl is
19893 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19895 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19896 order of HIGH and CONST checks.
19898 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19900 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19901 if the destination isn't an SSA_NAME.
19903 2018-01-09 Richard Biener <rguenther@suse.de>
19905 PR tree-optimization/83668
19906 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19908 (canonicalize_loop_form): ... here, renamed from ...
19909 (canonicalize_loop_closed_ssa_form): ... this and amended to
19910 swap successor edges for loop exit blocks to make us use
19911 the RPO order we need for initial schedule generation.
19913 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19915 PR tree-optimization/64811
19916 * match.pd: When optimizing comparisons with Inf, avoid
19917 introducing or losing exceptions from comparisons with NaN.
19919 2018-01-09 Martin Liska <mliska@suse.cz>
19922 * asan.c (shadow_mem_size): Add gcc_assert.
19924 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19926 Don't save registers in main().
19929 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19930 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19931 * config/avr/avr.c (avr_set_current_function): Don't error if
19932 naked, OS_task or OS_main are specified at the same time.
19933 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19935 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19937 * common/config/avr/avr-common.c (avr_option_optimization_table):
19938 Switch on -mmain-is-OS_task for optimizing compilations.
19940 2018-01-09 Richard Biener <rguenther@suse.de>
19942 PR tree-optimization/83572
19943 * graphite.c: Include cfganal.h.
19944 (graphite_transform_loops): Connect infinite loops to exit
19945 and remove fake edges at the end.
19947 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19949 * ipa-inline.c (edge_badness): Revert accidental checkin.
19951 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19954 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19955 symbols; not inline clones.
19957 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19960 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19961 hard registers. Formatting fixes.
19963 PR preprocessor/83722
19964 * gcc.c (try_generate_repro): Pass
19965 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19966 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19969 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19970 Kito Cheng <kito.cheng@gmail.com>
19972 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19973 (riscv_leaf_function_p): Delete.
19974 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19976 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19978 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19980 (do_ifelse): New function.
19981 (do_isel): New function.
19982 (do_sub3): New function.
19983 (do_add3): New function.
19984 (do_load_mask_compare): New function.
19985 (do_overlap_load_compare): New function.
19986 (expand_compare_loop): New function.
19987 (expand_block_compare): Call expand_compare_loop() when appropriate.
19988 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19989 option description.
19990 (-mblock-compare-inline-loop-limit): New option.
19992 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19995 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19996 Reverse order of second and third operands in first alternative.
19997 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19998 of first and second elements in UNSPEC_VPERMR vector.
19999 (altivec_expand_vec_perm_le): Likewise.
20001 2018-01-08 Jeff Law <law@redhat.com>
20003 PR rtl-optimizatin/81308
20004 * tree-switch-conversion.c (cfg_altered): New file scoped static.
20005 (process_switch): If group_case_labels makes a change, then set
20007 (pass_convert_switch::execute): If a switch is converted, then
20008 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
20010 PR rtl-optimization/81308
20011 * recog.c (split_all_insns): Conditionally cleanup the CFG after
20014 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
20016 PR target/83663 - Revert r255946
20017 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
20018 generation for cases where splatting a value is not useful.
20019 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
20020 across a vec_duplicate and a paradoxical subreg forming a vector
20021 mode to a vec_concat.
20023 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20025 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
20026 -march=armv8.3-a variants.
20027 * config/arm/t-multilib: Likewise.
20028 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
20030 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20032 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
20034 (cceq_ior_compare_complement): Give it a name so I can use it, and
20035 change boolean_or_operator predicate to boolean_operator so it can
20036 be used to generate a crand.
20037 (eqne): New code iterator.
20038 (bd/bd_neg): New code_attrs.
20039 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
20040 a single define_insn.
20041 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
20042 decrement (bdnzt/bdnzf/bdzt/bdzf).
20043 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
20044 with the new names of the branch decrement patterns, and added the
20045 names of the branch decrement conditional patterns.
20047 2018-01-08 Richard Biener <rguenther@suse.de>
20049 PR tree-optimization/83563
20050 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
20053 2018-01-08 Richard Biener <rguenther@suse.de>
20055 PR middle-end/83713
20056 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
20058 2018-01-08 Richard Biener <rguenther@suse.de>
20060 PR tree-optimization/83685
20061 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
20062 references to abnormals.
20064 2018-01-08 Richard Biener <rguenther@suse.de>
20067 * dwarf2out.c (output_indirect_strings): Handle empty
20068 skeleton_debug_str_hash.
20069 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
20071 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20073 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
20074 (emit_store_direct): Likewise.
20075 (arc_trampoline_adjust_address): Likewise.
20076 (arc_asm_trampoline_template): New function.
20077 (arc_initialize_trampoline): Use asm_trampoline_template.
20078 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
20079 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
20080 * config/arc/arc.md (flush_icache): Delete pattern.
20082 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20084 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
20085 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
20088 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20091 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
20092 by not USED_FOR_TARGET.
20093 (make_pass_resolve_sw_modes): Likewise.
20095 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20097 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
20100 2018-01-08 Richard Biener <rguenther@suse.de>
20102 PR middle-end/83580
20103 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
20105 2018-01-08 Richard Biener <rguenther@suse.de>
20107 PR middle-end/83517
20108 * match.pd ((t * 2) / 2) -> t): Add missing :c.
20110 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
20112 PR middle-end/81897
20113 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
20114 basic blocks with a small number of successors.
20115 (convert_control_dep_chain_into_preds): Improve handling of
20117 (dump_predicates): Split apart into...
20118 (dump_pred_chain): ...here...
20119 (dump_pred_info): ...and here.
20120 (can_one_predicate_be_invalidated_p): Add debugging printfs.
20121 (can_chain_union_be_invalidated_p): Improve check for invalidation
20123 (uninit_uses_cannot_happen): Avoid unnecessary if
20124 convert_control_dep_chain_into_preds yielded nothing.
20126 2018-01-06 Martin Sebor <msebor@redhat.com>
20128 PR tree-optimization/83640
20129 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
20130 subtracting negative offset from size.
20131 (builtin_access::overlap): Adjust offset bounds of the access to fall
20132 within the size of the object if possible.
20134 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
20136 PR rtl-optimization/83699
20137 * expmed.c (extract_bit_field_1): Restrict the vector usage of
20138 extract_bit_field_as_subreg to cases in which the extracted
20139 value is also a vector.
20141 * lra-constraints.c (process_alt_operands): Test for the equivalence
20142 substitutions when detecting a possible reload cycle.
20144 2018-01-06 Jakub Jelinek <jakub@redhat.com>
20147 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
20148 by default if flag_selective_schedling{,2}. Formatting fixes.
20150 PR rtl-optimization/83682
20151 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
20152 if it has non-VECTOR_MODE element mode.
20153 (vec_duplicate_p): Likewise.
20155 PR middle-end/83694
20156 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
20157 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
20159 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20162 * config/i386/i386-builtin.def
20163 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
20164 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
20165 Require also OPTION_MASK_ISA_AVX512F in addition to
20166 OPTION_MASK_ISA_GFNI.
20167 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
20168 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
20169 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
20170 to OPTION_MASK_ISA_GFNI.
20171 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
20172 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
20173 OPTION_MASK_ISA_AVX512BW.
20174 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
20175 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
20176 addition to OPTION_MASK_ISA_GFNI.
20177 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
20178 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
20179 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
20180 to OPTION_MASK_ISA_GFNI.
20181 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
20182 a requirement for all ISAs rather than any of them with a few
20184 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
20186 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
20187 bitmasks to be enabled with 3 exceptions, instead of requiring any
20188 enabled ISA with lots of exceptions.
20189 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
20190 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
20191 Change avx512bw in isa attribute to avx512f.
20192 * config/i386/sgxintrin.h: Add license boilerplate.
20193 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
20194 to __AVX512F__ and __AVX512VL to __AVX512VL__.
20195 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
20196 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
20198 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
20199 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
20200 temporarily sse2 rather than sse if not enabled already.
20203 * config/i386/sse.md (VI248_VLBW): Rename to ...
20204 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
20205 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
20206 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
20207 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
20208 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
20209 mode iterator instead of VI248_VLBW.
20211 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
20213 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
20214 (record_modified): Skip clobbers; add debug output.
20215 (param_change_prob): Use sreal frequencies.
20217 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20219 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
20220 punt for user-aligned variables.
20222 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20224 * tree-chrec.c (chrec_contains_symbols): Return true for
20227 2018-01-05 Sudakshina Das <sudi.das@arm.com>
20230 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
20231 of (x|y) == x for BICS pattern.
20233 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20235 PR tree-optimization/83605
20236 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
20237 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
20240 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
20242 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
20243 * config/epiphany/rtems.h: New file.
20245 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20246 Uros Bizjak <ubizjak@gmail.com>
20249 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
20250 QIreg_operand instead of register_operand predicate.
20251 * config/i386/i386.c (ix86_rop_should_change_byte_p,
20252 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
20253 comments instead of -fmitigate[-_]rop.
20255 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20258 * cgraphunit.c (symbol_table::compile): Switch to text_section
20259 before calling assembly_start debug hook.
20260 * run-rtl-passes.c (run_rtl_passes): Likewise.
20263 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20265 * tree-vrp.c (extract_range_from_binary_expr_1): Check
20266 range_int_cst_p rather than !symbolic_range_p before calling
20267 extract_range_from_multiplicative_op_1.
20269 2018-01-04 Jeff Law <law@redhat.com>
20271 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
20272 redundant test in assertion.
20274 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20276 * doc/rtl.texi: Document machine_mode wrapper classes.
20278 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20280 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
20281 using tree_to_uhwi.
20283 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20285 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
20286 the VEC_PERM_EXPR fold to fail.
20288 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20291 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
20292 to switched_sections.
20294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20297 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
20298 test for d.testing.
20300 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
20303 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
20304 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
20306 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20309 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
20310 is BLKmode and bitpos not zero or mode change is needed.
20312 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20315 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
20318 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
20321 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
20322 instead of MULT rtx. Update all corresponding splitters.
20323 (*saddl_se): Ditto.
20324 (*ssub<modesuffix>): Ditto.
20325 (*ssubl_se): Ditto.
20326 (*cmp_sadd_di): Update split patterns.
20327 (*cmp_sadd_si): Ditto.
20328 (*cmp_sadd_sidi): Ditto.
20329 (*cmp_ssub_di): Ditto.
20330 (*cmp_ssub_si): Ditto.
20331 (*cmp_ssub_sidi): Ditto.
20332 * config/alpha/predicates.md (const23_operand): New predicate.
20333 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
20334 Look for ASHIFT, not MULT inner operand.
20335 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
20337 2018-01-04 Martin Liska <mliska@suse.cz>
20339 PR gcov-profile/83669
20340 * gcov.c (output_intermediate_file): Add version to intermediate
20342 * doc/gcov.texi: Document new field 'version' in intermediate
20343 file format. Fix location of '-k' option of gcov command.
20345 2018-01-04 Martin Liska <mliska@suse.cz>
20348 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
20350 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20352 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
20354 2018-01-03 Martin Sebor <msebor@redhat.com>
20356 PR tree-optimization/83655
20357 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
20358 checking calls with invalid arguments.
20360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20362 * tree-vect-stmts.c (vect_get_store_rhs): New function.
20363 (vectorizable_mask_load_store): Delete.
20364 (vectorizable_call): Return false for masked loads and stores.
20365 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
20366 instead of gimple_assign_rhs1.
20367 (vectorizable_load): Handle IFN_MASK_LOAD.
20368 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
20370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20372 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
20374 (vectorizable_mask_load_store): ...here.
20375 (vectorizable_load): ...and here.
20377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20379 * tree-vect-stmts.c (vect_build_all_ones_mask)
20380 (vect_build_zero_merge_argument): New functions, split out from...
20381 (vectorizable_load): ...here.
20383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20385 * tree-vect-stmts.c (vect_check_store_rhs): New function,
20387 (vectorizable_mask_load_store): ...here.
20388 (vectorizable_store): ...and here.
20390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20392 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
20394 (vectorizable_mask_load_store): ...here.
20396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20398 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
20399 (vect_model_store_cost): Take a vec_load_store_type instead of a
20401 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
20402 (vect_model_store_cost): Take a vec_load_store_type instead of a
20404 (vectorizable_mask_load_store): Update accordingly.
20405 (vectorizable_store): Likewise.
20406 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
20408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20410 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
20411 IFN_MASK_LOAD calls here rather than...
20412 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
20414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20415 Alan Hayward <alan.hayward@arm.com>
20416 David Sherwood <david.sherwood@arm.com>
20418 * expmed.c (extract_bit_field_1): For vector extracts,
20419 fall back to extract_bit_field_as_subreg if vec_extract
20422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20423 Alan Hayward <alan.hayward@arm.com>
20424 David Sherwood <david.sherwood@arm.com>
20426 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
20427 they are variable or constant sized.
20428 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
20429 slots for constant-sized data.
20431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20432 Alan Hayward <alan.hayward@arm.com>
20433 David Sherwood <david.sherwood@arm.com>
20435 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
20436 handling COND_EXPRs with boolean comparisons, try to find a better
20437 basis for the mask type than the boolean itself.
20439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20441 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
20442 is calculated and how it can be overridden.
20443 * genmodes.c (max_bitsize_mode_any_mode): New variable.
20444 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
20446 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
20449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20450 Alan Hayward <alan.hayward@arm.com>
20451 David Sherwood <david.sherwood@arm.com>
20453 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
20454 Remove the mode argument.
20455 (aarch64_simd_valid_immediate): Remove the mode and inverse
20457 * config/aarch64/iterators.md (bitsize): New iterator.
20458 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
20459 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
20460 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
20461 aarch64_simd_valid_immediate.
20462 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
20463 (aarch64_reg_or_bic_imm): Likewise.
20464 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
20465 with an insn_type enum and msl with a modifier_type enum.
20466 Replace element_width with a scalar_mode. Change the shift
20467 to unsigned int. Add constructors for scalar_float_mode and
20468 scalar_int_mode elements.
20469 (aarch64_vect_float_const_representable_p): Delete.
20470 (aarch64_can_const_movi_rtx_p)
20471 (aarch64_simd_scalar_immediate_valid_for_move)
20472 (aarch64_simd_make_constant): Update call to
20473 aarch64_simd_valid_immediate.
20474 (aarch64_advsimd_valid_immediate_hs): New function.
20475 (aarch64_advsimd_valid_immediate): Likewise.
20476 (aarch64_simd_valid_immediate): Remove mode and inverse
20477 arguments. Rewrite to use the above. Use const_vec_duplicate_p
20478 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
20479 and aarch64_float_const_representable_p on the result.
20480 (aarch64_output_simd_mov_immediate): Remove mode argument.
20481 Update call to aarch64_simd_valid_immediate and use of
20482 simd_immediate_info.
20483 (aarch64_output_scalar_simd_mov_immediate): Update call
20486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20487 Alan Hayward <alan.hayward@arm.com>
20488 David Sherwood <david.sherwood@arm.com>
20490 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
20491 (mode_nunits): Likewise CONST_MODE_NUNITS.
20492 * machmode.def (ADJUST_NUNITS): Document.
20493 * genmodes.c (mode_data::need_nunits_adj): New field.
20494 (blank_mode): Update accordingly.
20495 (adj_nunits): New variable.
20496 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
20498 (emit_mode_size_inline): Set need_bytesize_adj for all modes
20499 listed in adj_nunits.
20500 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
20501 listed in adj_nunits. Don't emit case statements for such modes.
20502 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
20503 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
20504 nothing if adj_nunits is nonnull.
20505 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
20506 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
20507 (emit_mode_fbit): Update use of print_maybe_const_decl.
20508 (emit_move_size): Likewise. Treat the array as non-const
20510 (emit_mode_adjustments): Handle adj_nunits.
20512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20514 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
20515 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
20516 (VECTOR_MODES): Use it.
20517 (make_vector_modes): Take the prefix as an argument.
20519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20520 Alan Hayward <alan.hayward@arm.com>
20521 David Sherwood <david.sherwood@arm.com>
20523 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
20524 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
20525 for MODE_VECTOR_BOOL.
20526 * machmode.def (VECTOR_BOOL_MODE): Document.
20527 * genmodes.c (VECTOR_BOOL_MODE): New macro.
20528 (make_vector_bool_mode): New function.
20529 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
20531 * lto-streamer-in.c (lto_input_mode_table): Likewise.
20532 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
20534 * stor-layout.c (int_mode_for_mode): Likewise.
20535 * tree.c (build_vector_type_for_mode): Likewise.
20536 * varasm.c (output_constant_pool_2): Likewise.
20537 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
20538 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
20539 for MODE_VECTOR_BOOL.
20540 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
20541 of mode class checks.
20542 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
20543 instead of a list of mode class checks.
20544 (expand_vector_scalar_condition): Likewise.
20545 (type_for_widest_vector_mode): Handle BImode as an inner mode.
20547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20548 Alan Hayward <alan.hayward@arm.com>
20549 David Sherwood <david.sherwood@arm.com>
20551 * machmode.h (mode_size): Change from unsigned short to
20553 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
20554 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20555 or if measurement_type is not polynomial.
20556 (fixed_size_mode::includes_p): Check for constant-sized modes.
20557 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
20558 return a poly_uint16 rather than an unsigned short.
20559 (emit_mode_size): Change the type of mode_size from unsigned short
20560 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
20561 (emit_mode_adjustments): Cope with polynomial vector sizes.
20562 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20564 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20566 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
20567 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
20568 * caller-save.c (setup_save_areas): Likewise.
20569 (replace_reg_with_saved_mem): Likewise.
20570 * calls.c (emit_library_call_value_1): Likewise.
20571 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
20572 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
20573 (gen_lowpart_for_combine): Likewise.
20574 * convert.c (convert_to_integer_1): Likewise.
20575 * cse.c (equiv_constant, cse_insn): Likewise.
20576 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
20577 (cselib_subst_to_values): Likewise.
20578 * dce.c (word_dce_process_block): Likewise.
20579 * df-problems.c (df_word_lr_mark_ref): Likewise.
20580 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
20581 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
20582 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
20583 (rtl_for_decl_location): Likewise.
20584 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
20585 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
20586 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
20587 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
20588 (expand_expr_real_1): Likewise.
20589 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
20590 (pad_below): Likewise.
20591 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20592 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
20593 * ira.c (get_subreg_tracking_sizes): Likewise.
20594 * ira-build.c (ira_create_allocno_objects): Likewise.
20595 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
20596 (ira_sort_regnos_for_alter_reg): Likewise.
20597 * ira-costs.c (record_operand_costs): Likewise.
20598 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
20599 (resolve_simple_move): Likewise.
20600 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
20601 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
20602 (lra_constraints): Likewise.
20603 (CONST_POOL_OK_P): Reject variable-sized modes.
20604 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
20605 (add_pseudo_to_slot, lra_spill): Likewise.
20606 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20607 * optabs-query.c (get_best_extraction_insn): Likewise.
20608 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20609 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
20610 (expand_mult_highpart, valid_multiword_target_p): Likewise.
20611 * recog.c (offsettable_address_addr_space_p): Likewise.
20612 * regcprop.c (maybe_mode_change): Likewise.
20613 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
20614 * regrename.c (build_def_use): Likewise.
20615 * regstat.c (dump_reg_info): Likewise.
20616 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
20617 (find_reloads, find_reloads_subreg_address): Likewise.
20618 * reload1.c (eliminate_regs_1): Likewise.
20619 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
20620 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
20621 (simplify_binary_operation_1, simplify_subreg): Likewise.
20622 * targhooks.c (default_function_arg_padding): Likewise.
20623 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
20624 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
20625 (verify_gimple_assign_ternary): Likewise.
20626 * tree-inline.c (estimate_move_cost): Likewise.
20627 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20628 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
20629 (get_address_cost_ainc): Likewise.
20630 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
20631 (vect_supportable_dr_alignment): Likewise.
20632 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20633 (vectorizable_reduction): Likewise.
20634 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
20635 (vectorizable_operation, vectorizable_load): Likewise.
20636 * tree.c (build_same_sized_truth_vector_type): Likewise.
20637 * valtrack.c (cleanup_auto_inc_dec): Likewise.
20638 * var-tracking.c (emit_note_insn_var_location): Likewise.
20639 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
20640 (ADDR_VEC_ALIGN): Likewise.
20642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20643 Alan Hayward <alan.hayward@arm.com>
20644 David Sherwood <david.sherwood@arm.com>
20646 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
20648 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20649 or if measurement_type is polynomial.
20650 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
20651 * combine.c (make_extraction): Likewise.
20652 * dse.c (find_shift_sequence): Likewise.
20653 * dwarf2out.c (mem_loc_descriptor): Likewise.
20654 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
20655 (extract_bit_field, extract_low_bits): Likewise.
20656 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
20657 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
20658 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
20659 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
20660 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20661 * reload.c (find_reloads): Likewise.
20662 * reload1.c (alter_reg): Likewise.
20663 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
20664 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
20665 * tree-if-conv.c (predicate_mem_writes): Likewise.
20666 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
20667 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
20668 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
20669 * valtrack.c (dead_debug_insert_temp): Likewise.
20670 * varasm.c (mergeable_constant_section): Likewise.
20671 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
20673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20674 Alan Hayward <alan.hayward@arm.com>
20675 David Sherwood <david.sherwood@arm.com>
20677 * expr.c (expand_assignment): Cope with polynomial mode sizes
20678 when assigning to a CONCAT.
20680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20681 Alan Hayward <alan.hayward@arm.com>
20682 David Sherwood <david.sherwood@arm.com>
20684 * machmode.h (mode_precision): Change from unsigned short to
20686 (mode_to_precision): Return a poly_uint16 rather than an unsigned
20688 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
20689 or if measurement_type is not polynomial.
20690 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
20691 in which the mode is already known to be a scalar_int_mode.
20692 * genmodes.c (emit_mode_precision): Change the type of mode_precision
20693 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
20695 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20696 for GET_MODE_PRECISION.
20697 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20698 for GET_MODE_PRECISION.
20699 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
20701 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
20702 (expand_field_assignment, make_extraction): Likewise.
20703 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
20704 (get_last_value): Likewise.
20705 * convert.c (convert_to_integer_1): Likewise.
20706 * cse.c (cse_insn): Likewise.
20707 * expr.c (expand_expr_real_1): Likewise.
20708 * lra-constraints.c (simplify_operand_subreg): Likewise.
20709 * optabs-query.c (can_atomic_load_p): Likewise.
20710 * optabs.c (expand_atomic_load): Likewise.
20711 (expand_atomic_store): Likewise.
20712 * ree.c (combine_reaching_defs): Likewise.
20713 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
20714 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
20715 * tree.h (type_has_mode_precision_p): Likewise.
20716 * ubsan.c (instrument_si_overflow): Likewise.
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 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
20723 polynomial numbers of units.
20724 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
20725 (valid_vector_subparts_p): New function.
20726 (build_vector_type): Remove temporary shim and take the number
20727 of units as a poly_uint64 rather than an int.
20728 (build_opaque_vector_type): Take the number of units as a
20729 poly_uint64 rather than an int.
20730 * tree.c (build_vector_from_ctor): Handle polynomial
20731 TYPE_VECTOR_SUBPARTS.
20732 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
20733 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
20734 (build_vector_from_val): If the number of units is variable,
20735 use build_vec_duplicate_cst for constant operands and
20736 VEC_DUPLICATE_EXPR otherwise.
20737 (make_vector_type): Remove temporary is_constant ().
20738 (build_vector_type, build_opaque_vector_type): Take the number of
20739 units as a poly_uint64 rather than an int.
20740 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
20742 * cfgexpand.c (expand_debug_expr): Likewise.
20743 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
20744 (store_constructor, expand_expr_real_1): Likewise.
20745 (const_scalar_mask_from_tree): Likewise.
20746 * fold-const-call.c (fold_const_reduction): Likewise.
20747 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
20748 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
20749 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
20750 (fold_relational_const): Likewise.
20751 (native_interpret_vector): Likewise. Change the size from an
20752 int to an unsigned int.
20753 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
20754 TYPE_VECTOR_SUBPARTS.
20755 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
20756 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
20757 duplicating a non-constant operand into a variable-length vector.
20758 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
20759 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
20760 * ipa-icf.c (sem_variable::equals): Likewise.
20761 * match.pd: Likewise.
20762 * omp-simd-clone.c (simd_clone_subparts): Likewise.
20763 * print-tree.c (print_node): Likewise.
20764 * stor-layout.c (layout_type): Likewise.
20765 * targhooks.c (default_builtin_vectorization_cost): Likewise.
20766 * tree-cfg.c (verify_gimple_comparison): Likewise.
20767 (verify_gimple_assign_binary): Likewise.
20768 (verify_gimple_assign_ternary): Likewise.
20769 (verify_gimple_assign_single): Likewise.
20770 * tree-pretty-print.c (dump_generic_node): Likewise.
20771 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20772 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20773 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20774 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20775 (vect_shift_permute_load_chain): Likewise.
20776 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20777 (expand_vector_condition, optimize_vector_constructor): Likewise.
20778 (lower_vec_perm, get_compute_type): Likewise.
20779 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20780 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20781 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20782 (vect_recog_mask_conversion_pattern): Likewise.
20783 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20784 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20785 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20786 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20787 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20788 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20789 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20790 (supportable_widening_operation): Likewise.
20791 (supportable_narrowing_operation): Likewise.
20792 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20794 * varasm.c (output_constant): Likewise.
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-data-refs.c (vect_permute_store_chain): Reorganize
20801 so that both the length == 3 and length != 3 cases set up their
20802 own permute vectors. Add comments explaining why we know the
20803 number of elements is constant.
20804 (vect_permute_load_chain): Likewise.
20806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20807 Alan Hayward <alan.hayward@arm.com>
20808 David Sherwood <david.sherwood@arm.com>
20810 * machmode.h (mode_nunits): Change from unsigned char to
20812 (ONLY_FIXED_SIZE_MODES): New macro.
20813 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20814 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20815 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20817 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20818 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20819 or if measurement_type is not polynomial.
20820 * genmodes.c (ZERO_COEFFS): New macro.
20821 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20823 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20824 Use ZERO_COEFFS when emitting initializers.
20825 * data-streamer.h (bp_pack_poly_value): New function.
20826 (bp_unpack_poly_value): Likewise.
20827 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20828 for GET_MODE_NUNITS.
20829 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20830 for GET_MODE_NUNITS.
20831 * tree.c (make_vector_type): Remove temporary shim and make
20832 the real function take the number of units as a poly_uint64
20833 rather than an int.
20834 (build_vector_type_for_mode): Handle polynomial nunits.
20835 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20836 * emit-rtl.c (const_vec_series_p_1): Likewise.
20837 (gen_rtx_CONST_VECTOR): Likewise.
20838 * fold-const.c (test_vec_duplicate_folding): Likewise.
20839 * genrecog.c (validate_pattern): Likewise.
20840 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20841 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20842 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20843 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20844 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20845 * rtlanal.c (subreg_get_info): Likewise.
20846 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20847 (vect_grouped_load_supported): Likewise.
20848 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20849 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20850 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20851 (simplify_const_unary_operation, simplify_binary_operation_1)
20852 (simplify_const_binary_operation, simplify_ternary_operation)
20853 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20854 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20855 instead of CONST_VECTOR_NUNITS.
20856 * varasm.c (output_constant_pool_2): Likewise.
20857 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20858 explicit-encoded elements in the XVEC for variable-length vectors.
20860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20862 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
20864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20865 Alan Hayward <alan.hayward@arm.com>
20866 David Sherwood <david.sherwood@arm.com>
20868 * coretypes.h (fixed_size_mode): Declare.
20869 (fixed_size_mode_pod): New typedef.
20870 * builtins.h (target_builtins::x_apply_args_mode)
20871 (target_builtins::x_apply_result_mode): Change type to
20872 fixed_size_mode_pod.
20873 * builtins.c (apply_args_size, apply_result_size, result_vector)
20874 (expand_builtin_apply_args_1, expand_builtin_apply)
20875 (expand_builtin_return): Update accordingly.
20877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20879 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20880 * cselib.c (cselib_hash_rtx): Likewise.
20881 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20882 CONST_VECTOR encoding.
20884 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20885 Jeff Law <law@redhat.com>
20888 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20889 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20890 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20891 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20894 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20895 explicitly probe *sp in a noreturn function if there were any callee
20896 register saves or frame pointer is needed.
20898 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20901 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20902 BLKmode for ternary, binary or unary expressions.
20905 * var-tracking.c (delete_vta_debug_insn): New inline function.
20906 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20907 insns from get_insns () to NULL instead of each bb separately.
20908 Use delete_vta_debug_insn. No longer static.
20909 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20910 delete_vta_debug_insns callers.
20911 * rtl.h (delete_vta_debug_insns): Declare.
20912 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20913 instead of variable_tracking_main.
20915 2018-01-03 Martin Sebor <msebor@redhat.com>
20917 PR tree-optimization/83603
20918 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20919 arguments past the endof the argument list in functions declared
20920 without a prototype.
20921 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20922 Avoid checking when arguments are null.
20924 2018-01-03 Martin Sebor <msebor@redhat.com>
20927 * doc/extend.texi (attribute const): Fix a typo.
20928 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20929 issuing -Wsuggest-attribute for void functions.
20931 2018-01-03 Martin Sebor <msebor@redhat.com>
20933 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20934 offset_int::from instead of wide_int::to_shwi.
20935 (maybe_diag_overlap): Remove assertion.
20936 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20937 * gimple-ssa-sprintf.c (format_directive): Same.
20938 (parse_directive): Same.
20939 (sprintf_dom_walker::compute_format_length): Same.
20940 (try_substitute_return_value): Same.
20942 2018-01-03 Jeff Law <law@redhat.com>
20944 PR middle-end/83654
20945 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20946 non-constant residual for zero at runtime and avoid probing in
20947 that case. Reorganize code for trailing problem to mirror handling
20950 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20952 PR tree-optimization/83501
20953 * tree-ssa-strlen.c (get_string_cst): New.
20954 (handle_char_store): Call get_string_cst.
20956 2018-01-03 Martin Liska <mliska@suse.cz>
20958 PR tree-optimization/83593
20959 * tree-ssa-strlen.c: Include tree-cfg.h.
20960 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20961 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20962 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20964 (strlen_dom_walker::before_dom_children): Call
20965 gimple_purge_dead_eh_edges. Dump tranformation with details
20967 (strlen_dom_walker::before_dom_children): Update call by adding
20968 new argument cleanup_eh.
20969 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20971 2018-01-03 Martin Liska <mliska@suse.cz>
20974 * cif-code.def (VARIADIC_THUNK): New enum value.
20975 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20978 2018-01-03 Jan Beulich <jbeulich@suse.com>
20980 * sse.md (mov<mode>_internal): Tighten condition for when to use
20981 vmovdqu<ssescalarsize> for TI and OI modes.
20983 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20985 Update copyright years.
20987 2018-01-03 Martin Liska <mliska@suse.cz>
20990 * ipa-visibility.c (function_and_variable_visibility): Skip
20991 functions with noipa attribure.
20993 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20995 * gcc.c (process_command): Update copyright notice dates.
20996 * gcov-dump.c (print_version): Ditto.
20997 * gcov.c (print_version): Ditto.
20998 * gcov-tool.c (print_version): Ditto.
20999 * gengtype.c (create_file): Ditto.
21000 * doc/cpp.texi: Bump @copying's copyright year.
21001 * doc/cppinternals.texi: Ditto.
21002 * doc/gcc.texi: Ditto.
21003 * doc/gccint.texi: Ditto.
21004 * doc/gcov.texi: Ditto.
21005 * doc/install.texi: Ditto.
21006 * doc/invoke.texi: Ditto.
21008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21010 * vector-builder.h (vector_builder::m_full_nelts): Change from
21011 unsigned int to poly_uint64.
21012 (vector_builder::full_nelts): Update prototype accordingly.
21013 (vector_builder::new_vector): Likewise.
21014 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
21015 (vector_builder::operator ==): Likewise.
21016 (vector_builder::finalize): Likewise.
21017 * int-vector-builder.h (int_vector_builder::int_vector_builder):
21018 Take the number of elements as a poly_uint64 rather than an
21020 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
21021 from unsigned int to poly_uint64.
21022 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
21023 (vec_perm_indices::new_vector): Likewise.
21024 (vec_perm_indices::length): Likewise.
21025 (vec_perm_indices::nelts_per_input): Likewise.
21026 (vec_perm_indices::input_nelts): Likewise.
21027 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
21028 number of elements per input as a poly_uint64 rather than an
21029 unsigned int. Use the original encoding for variable-length
21030 vectors, rather than clamping each individual element.
21031 For the second and subsequent elements in each pattern,
21032 clamp the step and base before clamping their sum.
21033 (vec_perm_indices::series_p): Handle polynomial element counts.
21034 (vec_perm_indices::all_in_range_p): Likewise.
21035 (vec_perm_indices_to_tree): Likewise.
21036 (vec_perm_indices_to_rtx): Likewise.
21037 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
21038 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
21039 (tree_vector_builder::new_binary_operation): Handle polynomial
21040 element counts. Return false if we need to know the number
21041 of elements at compile time.
21042 * fold-const.c (fold_vec_perm): Punt if the number of elements
21043 isn't known at compile time.
21045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21047 * vec-perm-indices.h (vec_perm_builder): Change element type
21048 from HOST_WIDE_INT to poly_int64.
21049 (vec_perm_indices::element_type): Update accordingly.
21050 (vec_perm_indices::clamp): Handle polynomial element_types.
21051 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21052 (vec_perm_indices::all_in_range_p): Likewise.
21053 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
21055 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
21056 polynomial vec_perm_indices element types.
21057 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
21058 * fold-const.c (fold_vec_perm): Likewise.
21059 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
21060 * tree-vect-generic.c (lower_vec_perm): Likewise.
21061 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21062 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
21063 element type to HOST_WIDE_INT.
21065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21066 Alan Hayward <alan.hayward@arm.com>
21067 David Sherwood <david.sherwood@arm.com>
21069 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
21070 rather than an int. Use plus_constant.
21071 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
21072 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
21074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21075 Alan Hayward <alan.hayward@arm.com>
21076 David Sherwood <david.sherwood@arm.com>
21078 * calls.c (emit_call_1, expand_call): Change struct_value_size from
21079 a HOST_WIDE_INT to a poly_int64.
21081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21082 Alan Hayward <alan.hayward@arm.com>
21083 David Sherwood <david.sherwood@arm.com>
21085 * calls.c (load_register_parameters): Cope with polynomial
21086 mode sizes. Require a constant size for BLKmode parameters
21087 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
21088 forces a parameter to be padded at the lsb end in order to
21089 fill a complete number of words, require the parameter size
21090 to be ordered wrt UNITS_PER_WORD.
21092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21093 Alan Hayward <alan.hayward@arm.com>
21094 David Sherwood <david.sherwood@arm.com>
21096 * reload1.c (spill_stack_slot_width): Change element type
21097 from unsigned int to poly_uint64_pod.
21098 (alter_reg): Treat mode sizes as polynomial.
21100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21101 Alan Hayward <alan.hayward@arm.com>
21102 David Sherwood <david.sherwood@arm.com>
21104 * reload.c (complex_word_subreg_p): New function.
21105 (reload_inner_reg_of_subreg, push_reload): Use it.
21107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21108 Alan Hayward <alan.hayward@arm.com>
21109 David Sherwood <david.sherwood@arm.com>
21111 * lra-constraints.c (process_alt_operands): Reject matched
21112 operands whose sizes aren't ordered.
21113 (match_reload): Refer to this check here.
21115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21116 Alan Hayward <alan.hayward@arm.com>
21117 David Sherwood <david.sherwood@arm.com>
21119 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
21120 that the mode size is in the set {1, 2, 4, 8, 16}.
21122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21123 Alan Hayward <alan.hayward@arm.com>
21124 David Sherwood <david.sherwood@arm.com>
21126 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
21127 Use plus_constant instead of gen_rtx_PLUS.
21129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21130 Alan Hayward <alan.hayward@arm.com>
21131 David Sherwood <david.sherwood@arm.com>
21133 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
21134 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
21135 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
21136 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
21137 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
21138 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
21139 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
21140 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
21141 * config/i386/i386.c (ix86_push_rounding): ...this new function.
21142 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
21144 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
21145 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
21146 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
21147 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
21148 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
21149 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
21150 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
21151 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
21152 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
21153 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
21155 * expr.c (emit_move_resolve_push): Treat the input and result
21156 of PUSH_ROUNDING as a poly_int64.
21157 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
21158 (emit_push_insn): Likewise.
21159 * lra-eliminations.c (mark_not_eliminable): Likewise.
21160 * recog.c (push_operand): Likewise.
21161 * reload1.c (elimination_effects): Likewise.
21162 * rtlanal.c (nonzero_bits1): Likewise.
21163 * calls.c (store_one_arg): Likewise. Require the padding to be
21164 known at compile time.
21166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21167 Alan Hayward <alan.hayward@arm.com>
21168 David Sherwood <david.sherwood@arm.com>
21170 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
21171 Use plus_constant instead of gen_rtx_PLUS.
21173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21174 Alan Hayward <alan.hayward@arm.com>
21175 David Sherwood <david.sherwood@arm.com>
21177 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
21178 rather than an int.
21180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21181 Alan Hayward <alan.hayward@arm.com>
21182 David Sherwood <david.sherwood@arm.com>
21184 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
21185 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
21186 via stack temporaries. Treat the mode size as polynomial too.
21188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21189 Alan Hayward <alan.hayward@arm.com>
21190 David Sherwood <david.sherwood@arm.com>
21192 * expr.c (expand_expr_real_2): When handling conversions involving
21193 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
21194 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
21195 as a poly_uint64 too.
21197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21198 Alan Hayward <alan.hayward@arm.com>
21199 David Sherwood <david.sherwood@arm.com>
21201 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
21203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21204 Alan Hayward <alan.hayward@arm.com>
21205 David Sherwood <david.sherwood@arm.com>
21207 * combine.c (can_change_dest_mode): Handle polynomial
21208 REGMODE_NATURAL_SIZE.
21209 * expmed.c (store_bit_field_1): Likewise.
21210 * expr.c (store_constructor): Likewise.
21211 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
21212 and polynomial REGMODE_NATURAL_SIZE.
21213 (gen_lowpart_common): Likewise.
21214 * reginfo.c (record_subregs_of_mode): Likewise.
21215 * rtlanal.c (read_modify_subreg_p): Likewise.
21217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21218 Alan Hayward <alan.hayward@arm.com>
21219 David Sherwood <david.sherwood@arm.com>
21221 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
21222 numbers of elements.
21224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21225 Alan Hayward <alan.hayward@arm.com>
21226 David Sherwood <david.sherwood@arm.com>
21228 * match.pd: Cope with polynomial numbers of vector elements.
21230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21231 Alan Hayward <alan.hayward@arm.com>
21232 David Sherwood <david.sherwood@arm.com>
21234 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
21235 in a POINTER_PLUS_EXPR.
21237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21238 Alan Hayward <alan.hayward@arm.com>
21239 David Sherwood <david.sherwood@arm.com>
21241 * omp-simd-clone.c (simd_clone_subparts): New function.
21242 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
21243 (ipa_simd_modify_function_body): Likewise.
21245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21246 Alan Hayward <alan.hayward@arm.com>
21247 David Sherwood <david.sherwood@arm.com>
21249 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
21250 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
21251 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
21252 (expand_vector_condition, vector_element): Likewise.
21253 (subparts_gt): New function.
21254 (get_compute_type): Use subparts_gt.
21255 (count_type_subparts): Delete.
21256 (expand_vector_operations_1): Use subparts_gt instead of
21257 count_type_subparts.
21259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21260 Alan Hayward <alan.hayward@arm.com>
21261 David Sherwood <david.sherwood@arm.com>
21263 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
21264 (vect_compile_time_alias): ...this new function. Do the calculation
21265 on poly_ints rather than trees.
21266 (vect_prune_runtime_alias_test_list): Update call accordingly.
21268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21269 Alan Hayward <alan.hayward@arm.com>
21270 David Sherwood <david.sherwood@arm.com>
21272 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
21274 (vect_schedule_slp_instance): Likewise.
21276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21277 Alan Hayward <alan.hayward@arm.com>
21278 David Sherwood <david.sherwood@arm.com>
21280 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
21281 constant and extern definitions for variable-length vectors.
21282 (vect_get_constant_vectors): Note that the number of units
21283 is known to be constant.
21285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21286 Alan Hayward <alan.hayward@arm.com>
21287 David Sherwood <david.sherwood@arm.com>
21289 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
21290 of units as polynomial. Choose between WIDE and NARROW based
21293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21294 Alan Hayward <alan.hayward@arm.com>
21295 David Sherwood <david.sherwood@arm.com>
21297 * tree-vect-stmts.c (simd_clone_subparts): New function.
21298 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
21300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21301 Alan Hayward <alan.hayward@arm.com>
21302 David Sherwood <david.sherwood@arm.com>
21304 * tree-vect-stmts.c (vectorizable_call): Treat the number of
21305 vectors as polynomial. Use build_index_vector for
21306 IFN_GOMP_SIMD_LANE.
21308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21309 Alan Hayward <alan.hayward@arm.com>
21310 David Sherwood <david.sherwood@arm.com>
21312 * tree-vect-stmts.c (get_load_store_type): Treat the number of
21313 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
21314 for variable-length vectors.
21315 (vectorizable_mask_load_store): Treat the number of units as
21316 polynomial, asserting that it is constant if the condition has
21317 already been enforced.
21318 (vectorizable_store, vectorizable_load): Likewise.
21320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21321 Alan Hayward <alan.hayward@arm.com>
21322 David Sherwood <david.sherwood@arm.com>
21324 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
21325 of units as polynomial. Punt if we can't tell at compile time
21326 which vector contains the final result.
21328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21329 Alan Hayward <alan.hayward@arm.com>
21330 David Sherwood <david.sherwood@arm.com>
21332 * tree-vect-loop.c (vectorizable_induction): Treat the number
21333 of units as polynomial. Punt on SLP inductions. Use an integer
21334 VEC_SERIES_EXPR for variable-length integer reductions. Use a
21335 cast of such a series for variable-length floating-point
21338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21339 Alan Hayward <alan.hayward@arm.com>
21340 David Sherwood <david.sherwood@arm.com>
21342 * tree.h (build_index_vector): Declare.
21343 * tree.c (build_index_vector): New function.
21344 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
21345 of units as polynomial, forcibly converting it to a constant if
21346 vectorizable_reduction has already enforced the condition.
21347 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
21348 to create a {1,2,3,...} vector.
21349 (vectorizable_reduction): Treat the number of units as polynomial.
21350 Choose vectype_in based on the largest scalar element size rather
21351 than the smallest number of units. Enforce the restrictions
21354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21355 Alan Hayward <alan.hayward@arm.com>
21356 David Sherwood <david.sherwood@arm.com>
21358 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
21359 number of units as polynomial.
21361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21362 Alan Hayward <alan.hayward@arm.com>
21363 David Sherwood <david.sherwood@arm.com>
21365 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
21366 * target.def (autovectorize_vector_sizes): Return the vector sizes
21367 by pointer, using vector_sizes rather than a bitmask.
21368 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
21369 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
21370 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
21372 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
21373 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
21374 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
21375 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
21376 * omp-general.c (omp_max_vf): Likewise.
21377 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21378 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
21379 * tree-vect-loop.c (vect_analyze_loop): Likewise.
21380 * tree-vect-slp.c (vect_slp_bb): Likewise.
21381 * doc/tm.texi: Regenerate.
21382 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
21384 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
21385 the vector size as a poly_uint64 rather than an unsigned int.
21386 (current_vector_size): Change from an unsigned int to a poly_uint64.
21387 (get_vectype_for_scalar_type): Update accordingly.
21388 * tree.h (build_truth_vector_type): Take the size and number of
21389 units as a poly_uint64 rather than an unsigned int.
21390 (build_vector_type): Add a temporary overload that takes
21391 the number of units as a poly_uint64 rather than an unsigned int.
21392 * tree.c (make_vector_type): Likewise.
21393 (build_truth_vector_type): Take the number of units as a poly_uint64
21394 rather than an unsigned int.
21396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21397 Alan Hayward <alan.hayward@arm.com>
21398 David Sherwood <david.sherwood@arm.com>
21400 * target.def (get_mask_mode): Take the number of units and length
21401 as poly_uint64s rather than unsigned ints.
21402 * targhooks.h (default_get_mask_mode): Update accordingly.
21403 * targhooks.c (default_get_mask_mode): Likewise.
21404 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
21405 * doc/tm.texi: Regenerate.
21407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21408 Alan Hayward <alan.hayward@arm.com>
21409 David Sherwood <david.sherwood@arm.com>
21411 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
21412 * omp-general.c (omp_max_vf): Likewise.
21413 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
21414 (expand_omp_simd): Handle polynomial safelen.
21415 * omp-low.c (omplow_simd_context): Add a default constructor.
21416 (omplow_simd_context::max_vf): Change from int to poly_uint64.
21417 (lower_rec_simd_input_clauses): Update accordingly.
21418 (lower_rec_input_clauses): Likewise.
21420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21421 Alan Hayward <alan.hayward@arm.com>
21422 David Sherwood <david.sherwood@arm.com>
21424 * tree-vectorizer.h (vect_nunits_for_cost): New function.
21425 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
21426 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
21427 (vect_analyze_slp_cost): Likewise.
21428 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
21429 (vect_model_load_cost): Likewise.
21431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21432 Alan Hayward <alan.hayward@arm.com>
21433 David Sherwood <david.sherwood@arm.com>
21435 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
21436 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
21437 from an unsigned int * to a poly_uint64_pod *.
21438 (calculate_unrolling_factor): New function.
21439 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
21441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21442 Alan Hayward <alan.hayward@arm.com>
21443 David Sherwood <david.sherwood@arm.com>
21445 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
21446 from an unsigned int to a poly_uint64.
21447 (_loop_vec_info::slp_unrolling_factor): Likewise.
21448 (_loop_vec_info::vectorization_factor): Change from an int
21450 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
21451 (vect_get_num_vectors): New function.
21452 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
21453 (vect_get_num_copies): Use vect_get_num_vectors.
21454 (vect_analyze_data_ref_dependences): Change max_vf from an int *
21455 to an unsigned int *.
21456 (vect_analyze_data_refs): Change min_vf from an int * to a
21458 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21459 than an unsigned HOST_WIDE_INT.
21460 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
21461 (vect_analyze_data_ref_dependence): Change max_vf from an int *
21462 to an unsigned int *.
21463 (vect_analyze_data_ref_dependences): Likewise.
21464 (vect_compute_data_ref_alignment): Handle polynomial vf.
21465 (vect_enhance_data_refs_alignment): Likewise.
21466 (vect_prune_runtime_alias_test_list): Likewise.
21467 (vect_shift_permute_load_chain): Likewise.
21468 (vect_supportable_dr_alignment): Likewise.
21469 (dependence_distance_ge_vf): Take the vectorization factor as a
21470 poly_uint64 rather than an unsigned HOST_WIDE_INT.
21471 (vect_analyze_data_refs): Change min_vf from an int * to a
21473 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
21474 vfm1 as a poly_uint64 rather than an int. Make the same change
21475 for the returned bound_scalar.
21476 (vect_gen_vector_loop_niters): Handle polynomial vf.
21477 (vect_do_peeling): Likewise. Update call to
21478 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
21479 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
21481 * tree-vect-loop.c (vect_determine_vectorization_factor)
21482 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
21483 (vect_get_known_peeling_cost): Likewise.
21484 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
21485 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
21486 (vect_transform_loop): Likewise. Use the lowest possible VF when
21487 updating the upper bounds of the loop.
21488 (vect_min_worthwhile_factor): Make static. Return an unsigned int
21489 rather than an int.
21490 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
21491 polynomial unroll factors.
21492 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
21493 (vect_make_slp_decision): Likewise.
21494 (vect_supported_load_permutation_p): Likewise, and polynomial
21496 (vect_analyze_slp_cost): Handle polynomial vf.
21497 (vect_slp_analyze_node_operations): Likewise.
21498 (vect_slp_analyze_bb_1): Likewise.
21499 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21500 than an unsigned HOST_WIDE_INT.
21501 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
21502 (vectorizable_load): Handle polynomial vf.
21503 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
21505 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
21507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21508 Alan Hayward <alan.hayward@arm.com>
21509 David Sherwood <david.sherwood@arm.com>
21511 * match.pd: Handle bit operations involving three constants
21512 and try to fold one pair.
21514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21516 * tree-vect-loop-manip.c: Include gimple-fold.h.
21517 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
21518 niters_maybe_zero parameters. Handle other cases besides a step of 1.
21519 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
21520 Add a path that uses a step of VF instead of 1, but disable it
21522 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
21523 and niters_no_overflow parameters. Update calls to
21524 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
21525 Create a new SSA name if the latter choses to use a ste other
21526 than zero, and return it via niters_vector_mult_vf_var.
21527 * tree-vect-loop.c (vect_transform_loop): Update calls to
21528 vect_do_peeling, vect_gen_vector_loop_niters and
21529 slpeel_make_loop_iterate_ntimes.
21530 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
21531 (vect_gen_vector_loop_niters): Update declarations after above changes.
21533 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
21535 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
21536 128-bit round to integer instructions.
21537 (ceil<mode>2): Likewise.
21538 (btrunc<mode>2): Likewise.
21539 (round<mode>2): Likewise.
21541 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21543 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
21544 unaligned VSX load/store on P8/P9.
21545 (expand_block_clear): Allow the use of unaligned VSX
21546 load/store on P8/P9.
21548 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21550 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
21552 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
21553 swap associated with both a load and a store.
21555 2018-01-02 Andrew Waterman <andrew@sifive.com>
21557 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
21558 * config/riscv/riscv.md (clear_cache): Use it.
21560 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
21562 * web.c: Remove out-of-date comment.
21564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21566 * expr.c (fixup_args_size_notes): Check that any existing
21567 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
21568 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
21569 (emit_single_push_insn): ...here.
21571 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21573 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
21574 (const_vector_encoded_nelts): New function.
21575 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
21576 (const_vector_int_elt, const_vector_elt): Declare.
21577 * emit-rtl.c (const_vector_int_elt_1): New function.
21578 (const_vector_elt): Likewise.
21579 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
21580 of CONST_VECTOR_ELT.
21582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21584 * expr.c: Include rtx-vector-builder.h.
21585 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
21586 directly on the tree encoding.
21587 (const_vector_from_tree): Likewise.
21588 * optabs.c: Include rtx-vector-builder.h.
21589 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
21590 sequence of "u" values.
21591 * vec-perm-indices.c: Include rtx-vector-builder.h.
21592 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
21593 directly on the vec_perm_indices encoding.
21595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21597 * doc/rtl.texi (const_vector): Describe new encoding scheme.
21598 * Makefile.in (OBJS): Add rtx-vector-builder.o.
21599 * rtx-vector-builder.h: New file.
21600 * rtx-vector-builder.c: Likewise.
21601 * rtl.h (rtx_def::u2): Add a const_vector field.
21602 (CONST_VECTOR_NPATTERNS): New macro.
21603 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
21604 (CONST_VECTOR_DUPLICATE_P): Likewise.
21605 (CONST_VECTOR_STEPPED_P): Likewise.
21606 (CONST_VECTOR_ENCODED_ELT): Likewise.
21607 (const_vec_duplicate_p): Check for a duplicated vector encoding.
21608 (unwrap_const_vec_duplicate): Likewise.
21609 (const_vec_series_p): Check for a non-duplicated vector encoding.
21610 Say that the function only returns true for integer vectors.
21611 * emit-rtl.c: Include rtx-vector-builder.h.
21612 (gen_const_vec_duplicate_1): Delete.
21613 (gen_const_vector): Call gen_const_vec_duplicate instead of
21614 gen_const_vec_duplicate_1.
21615 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
21616 (gen_const_vec_duplicate): Use rtx_vector_builder.
21617 (gen_const_vec_series): Likewise.
21618 (gen_rtx_CONST_VECTOR): Likewise.
21619 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
21620 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21621 Build a new vector rather than modifying a CONST_VECTOR in-place.
21622 (handle_special_swappables): Update call accordingly.
21623 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
21624 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21625 Build a new vector rather than modifying a CONST_VECTOR in-place.
21626 (handle_special_swappables): Update call accordingly.
21628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21630 * simplify-rtx.c (simplify_const_binary_operation): Use
21631 CONST_VECTOR_ELT instead of XVECEXP.
21633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21635 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
21636 the selector elements to be different from the data elements
21637 if the selector is a VECTOR_CST.
21638 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
21639 ssizetype for the selector.
21641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21643 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
21644 before testing each element individually.
21645 * tree-vect-generic.c (lower_vec_perm): Likewise.
21647 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21649 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
21650 * selftest-run-tests.c (selftest::run_tests): Call it.
21651 * vector-builder.h (vector_builder::operator ==): New function.
21652 (vector_builder::operator !=): Likewise.
21653 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
21654 (vec_perm_indices::all_from_input_p): New function.
21655 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21656 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
21657 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
21658 instead of reading the VECTOR_CST directly. Detect whether both
21659 vector inputs are the same before constructing the vec_perm_indices,
21660 and update the number of inputs argument accordingly. Use the
21661 utility functions added above. Only construct sel2 if we need to.
21663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21665 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
21666 the broadcast of the low byte.
21667 (expand_mult_highpart): Use an explicit encoding for the permutes.
21668 * optabs-query.c (can_mult_highpart_p): Likewise.
21669 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
21670 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21671 (vectorizable_bswap): Likewise.
21672 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
21673 explicit encoding for the power-of-2 permutes.
21674 (vect_permute_store_chain): Likewise.
21675 (vect_grouped_load_supported): Likewise.
21676 (vect_permute_load_chain): Likewise.
21678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21680 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
21681 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
21682 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
21683 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21684 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
21685 (vect_gen_perm_mask_any): Likewise.
21687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21689 * int-vector-builder.h: New file.
21690 * vec-perm-indices.h: Include int-vector-builder.h.
21691 (vec_perm_indices): Redefine as an int_vector_builder.
21692 (auto_vec_perm_indices): Delete.
21693 (vec_perm_builder): Redefine as a stand-alone class.
21694 (vec_perm_indices::vec_perm_indices): New function.
21695 (vec_perm_indices::clamp): Likewise.
21696 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
21697 (vec_perm_indices::new_vector): New function.
21698 (vec_perm_indices::new_expanded_vector): Update for new
21699 vec_perm_indices class.
21700 (vec_perm_indices::rotate_inputs): New function.
21701 (vec_perm_indices::all_in_range_p): Operate directly on the
21702 encoded form, without computing elided elements.
21703 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
21704 encoding. Update for new vec_perm_indices class.
21705 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
21706 the given vec_perm_builder.
21707 (expand_vec_perm_var): Update vec_perm_builder constructor.
21708 (expand_mult_highpart): Use vec_perm_builder instead of
21709 auto_vec_perm_indices.
21710 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
21711 vec_perm_indices instead of auto_vec_perm_indices. Use a single
21712 or double series encoding as appropriate.
21713 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
21714 vec_perm_indices instead of auto_vec_perm_indices.
21715 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21716 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21717 (vect_permute_store_chain): Likewise.
21718 (vect_grouped_load_supported): Likewise.
21719 (vect_permute_load_chain): Likewise.
21720 (vect_shift_permute_load_chain): Likewise.
21721 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21722 (vect_transform_slp_perm_load): Likewise.
21723 (vect_schedule_slp_instance): Likewise.
21724 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21725 (vectorizable_mask_load_store): Likewise.
21726 (vectorizable_bswap): Likewise.
21727 (vectorizable_store): Likewise.
21728 (vectorizable_load): Likewise.
21729 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
21730 vec_perm_indices instead of auto_vec_perm_indices. Use
21731 tree_to_vec_perm_builder to read the vector from a tree.
21732 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
21733 vec_perm_builder instead of a vec_perm_indices.
21734 (have_whole_vector_shift): Use vec_perm_builder and
21735 vec_perm_indices instead of auto_vec_perm_indices. Leave the
21736 truncation to calc_vec_perm_mask_for_shift.
21737 (vect_create_epilog_for_reduction): Likewise.
21738 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
21739 from auto_vec_perm_indices to vec_perm_indices.
21740 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21741 instead of changing individual elements.
21742 (aarch64_vectorize_vec_perm_const): Use new_vector to install
21743 the vector in d.perm.
21744 * config/arm/arm.c (expand_vec_perm_d::perm): Change
21745 from auto_vec_perm_indices to vec_perm_indices.
21746 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21747 instead of changing individual elements.
21748 (arm_vectorize_vec_perm_const): Use new_vector to install
21749 the vector in d.perm.
21750 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
21751 Update vec_perm_builder constructor.
21752 (rs6000_expand_interleave): Likewise.
21753 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
21754 (rs6000_expand_interleave): Likewise.
21756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21758 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
21759 to qimode could truncate the indices.
21760 * optabs.c (expand_vec_perm_var): Likewise.
21762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21764 * Makefile.in (OBJS): Add vec-perm-indices.o.
21765 * vec-perm-indices.h: New file.
21766 * vec-perm-indices.c: Likewise.
21767 * target.h (vec_perm_indices): Replace with a forward class
21769 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21770 * optabs.h: Include vec-perm-indices.h.
21771 (expand_vec_perm): Delete.
21772 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21773 (expand_vec_perm_const): Declare.
21774 * target.def (vec_perm_const_ok): Replace with...
21775 (vec_perm_const): ...this new hook.
21776 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21777 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21778 * doc/tm.texi: Regenerate.
21779 * optabs.def (vec_perm_const): Delete.
21780 * doc/md.texi (vec_perm_const): Likewise.
21781 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21782 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21783 expand_vec_perm for constant permutation vectors. Assert that
21784 the mode of variable permutation vectors is the integer equivalent
21785 of the mode that is being permuted.
21786 * optabs-query.h (selector_fits_mode_p): Declare.
21787 * optabs-query.c: Include vec-perm-indices.h.
21788 (selector_fits_mode_p): New function.
21789 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21790 is defined, instead of checking whether the vec_perm_const_optab
21791 exists. Use targetm.vectorize.vec_perm_const instead of
21792 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21793 fit in the vector mode before using a variable permute.
21794 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21795 vec_perm_indices instead of an rtx.
21796 (expand_vec_perm): Replace with...
21797 (expand_vec_perm_const): ...this new function. Take the selector
21798 as a vec_perm_indices rather than an rtx. Also take the mode of
21799 the selector. Update call to shift_amt_for_vec_perm_mask.
21800 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21801 Use vec_perm_indices::new_expanded_vector to expand the original
21802 selector into bytes. Check whether the indices fit in the vector
21803 mode before using a variable permute.
21804 (expand_vec_perm_var): Make global.
21805 (expand_mult_highpart): Use expand_vec_perm_const.
21806 * fold-const.c: Includes vec-perm-indices.h.
21807 * tree-ssa-forwprop.c: Likewise.
21808 * tree-vect-data-refs.c: Likewise.
21809 * tree-vect-generic.c: Likewise.
21810 * tree-vect-loop.c: Likewise.
21811 * tree-vect-slp.c: Likewise.
21812 * tree-vect-stmts.c: Likewise.
21813 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21815 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21816 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21817 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21818 (aarch64_vectorize_vec_perm_const): ...this new function.
21819 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21820 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21821 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21822 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21823 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21824 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21825 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21827 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21828 check for NEON modes.
21829 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21830 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21831 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21832 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21834 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21835 the old VEC_PERM_CONST conditions.
21836 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21837 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21838 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21839 (ia64_vectorize_vec_perm_const_ok): Merge into...
21840 (ia64_vectorize_vec_perm_const): ...this new function.
21841 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21842 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21843 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21844 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21845 * config/mips/mips.c (mips_expand_vec_perm_const)
21846 (mips_vectorize_vec_perm_const_ok): Merge into...
21847 (mips_vectorize_vec_perm_const): ...this new function.
21848 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21849 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21850 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21851 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21852 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21853 (rs6000_expand_vec_perm_const): Delete.
21854 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21856 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21857 (altivec_expand_vec_perm_const_le): Take each operand individually.
21858 Operate on constant selectors rather than rtxes.
21859 (altivec_expand_vec_perm_const): Likewise. Update call to
21860 altivec_expand_vec_perm_const_le.
21861 (rs6000_expand_vec_perm_const): Delete.
21862 (rs6000_vectorize_vec_perm_const_ok): Delete.
21863 (rs6000_vectorize_vec_perm_const): New function.
21864 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21865 an element count and rtx array.
21866 (rs6000_expand_extract_even): Update call accordingly.
21867 (rs6000_expand_interleave): Likewise.
21868 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21869 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21870 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21871 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21872 (rs6000_expand_vec_perm_const): Delete.
21873 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21874 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21875 (altivec_expand_vec_perm_const_le): Take each operand individually.
21876 Operate on constant selectors rather than rtxes.
21877 (altivec_expand_vec_perm_const): Likewise. Update call to
21878 altivec_expand_vec_perm_const_le.
21879 (rs6000_expand_vec_perm_const): Delete.
21880 (rs6000_vectorize_vec_perm_const_ok): Delete.
21881 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21882 reference to the SPE evmerge intructions.
21883 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21884 an element count and rtx array.
21885 (rs6000_expand_extract_even): Update call accordingly.
21886 (rs6000_expand_interleave): Likewise.
21887 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21888 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21890 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21892 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21894 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21895 vector mode and that that mode matches the mode of the data
21897 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21898 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21899 directly using expand_vec_perm_1 when forcing selectors into
21901 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21905 * optabs-query.h (can_vec_perm_p): Delete.
21906 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21907 * optabs-query.c (can_vec_perm_p): Split into...
21908 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21909 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21910 particular selector is valid.
21911 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21912 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21913 (vect_grouped_load_supported): Likewise.
21914 (vect_shift_permute_load_chain): Likewise.
21915 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21916 (vect_transform_slp_perm_load): Likewise.
21917 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21918 (vectorizable_bswap): Likewise.
21919 (vect_gen_perm_mask_checked): Likewise.
21920 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21921 implementations of variable permutation vectors into account
21922 when deciding which selector to use.
21923 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21924 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21925 with a false third argument.
21926 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21927 to test whether the constant selector is valid and can_vec_perm_var_p
21928 to test whether a variable selector is valid.
21930 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21932 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21933 * optabs-query.c (can_vec_perm_p): Likewise.
21934 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21935 instead of vec_perm_indices.
21936 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21937 (vect_gen_perm_mask_checked): Likewise,
21938 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21939 (vect_gen_perm_mask_checked): Likewise,
21941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21943 * optabs-query.h (qimode_for_vec_perm): Declare.
21944 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21945 (qimode_for_vec_perm): ...this new function.
21946 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21948 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21950 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21951 does not have a conditional at the top.
21953 2018-01-02 Richard Biener <rguenther@suse.de>
21955 * ipa-inline.c (big_speedup_p): Fix expression.
21957 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21960 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21963 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21967 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21968 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21969 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21970 cond_taken_branch_cost 3->4.
21972 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21974 PR tree-optimization/83581
21975 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21976 TODO_cleanup_cfg if any changes have been made.
21978 PR middle-end/83608
21979 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21980 convert_modes if target mode has the right side, but different mode
21983 PR middle-end/83609
21984 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21985 last argument when extracting from CONCAT. If either from_real or
21986 from_imag is NULL, use expansion through memory. If result is not
21987 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21988 the parts directly to inner mode, if even that fails, use expansion
21991 PR middle-end/83623
21992 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21993 check for bswap in mode rather than HImode and use that in expand_unop
21996 Copyright (C) 2018 Free Software Foundation, Inc.
21998 Copying and distribution of this file, with or without modification,
21999 are permitted in any medium without royalty provided the copyright
22000 notice and this notice are preserved.