1 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3 * genemit.c (print_overload_test): Fix typo.
5 2018-08-02 Richard Biener <rguenther@suse.de>
8 * class.c (layout_class_type): Copy TYPE_TYPELESS_STORAGE
9 to the CLASSTYPE_AS_BASE.
11 2018-08-02 Richard Biener <rguenther@suse.de>
13 PR tree-optimization/86816
14 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
15 which checks for value availability before querying it.
16 (gvn_uses_equal): Use it.
17 (same_succ_hash): Likewise.
18 (gimple_equal_p): Likewise.
20 2018-08-02 Nick Clifton <nickc@redhat.com>
23 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
24 Define to speculation_safe_value_not_needed.
27 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
28 Define to speculation_safe_value_not_needed.
31 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
32 Define to speculation_safe_value_not_needed.
35 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
36 Define to speculation_safe_value_not_needed.
39 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
40 speculation_safe_value_not_needed.
43 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
44 to speculation_safe_value_not_needed.
46 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
48 * doc/md.texi: Expand the documentation of instruction names
49 to mention port-local uses. Document '@' in pattern names.
50 * read-md.h (overloaded_instance, overloaded_name): New structs.
52 (md_reader::handle_overloaded_name): New member function.
53 (md_reader::get_overloads): Likewise.
54 (md_reader::m_first_overload): New member variable.
55 (md_reader::m_next_overload_ptr): Likewise.
56 (md_reader::m_overloads_htab): Likewise.
57 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
58 m_next_overload_ptr and m_overloads_htab.
59 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
60 (get_mode_token, get_code_token, get_int_token): New functions.
61 (map_attr_string): Add an optional argument that passes back
62 the associated iterator.
63 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
64 (md_reader::handle_overloaded_name, add_overload_instance): New
66 (apply_iterators): Handle '@' names. Report an error if '@'
67 is used without iterators.
68 (initialize_iterators): Initialize the new iterator_group fields.
69 * genopinit.c (handle_overloaded_code_for)
70 (handle_overloaded_gen): New functions.
71 (main): Use them to print declarations of maybe_code_for_* and
72 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
73 * genemit.c (print_overload_arguments, print_overload_test)
74 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
75 (main): Use it to print definitions of maybe_code_for_* and
76 maybe_gen_* functions.
77 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
78 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
79 instead of explicit mode checks.
80 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
81 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
82 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
83 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
84 (aarch64_expand_compare_and_swap): Likewise
85 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
86 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
87 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
88 (aarch64_constant_pool_reload_icode): Delete.
89 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
90 instead of aarch64_constant_pool_reload_icode. Use
91 code_for_aarch64_reload_mov instead of explicit mode checks.
92 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
93 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
94 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
95 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
96 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
97 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
98 (aarch64_atomic_load_op_code): Delete.
99 (aarch64_emit_atomic_load_op): Likewise.
100 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
101 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
102 instead of aarch64_emit_atomic_load_op.
103 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
104 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
105 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
106 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
107 character before the pattern name.
108 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
109 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
110 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
111 (aarch64_frecps<mode>): Likewise.
112 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
113 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
114 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
115 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
116 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
118 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
120 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
121 Allow HFmode constants if TARGET_FP_F16INST.
123 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
126 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
127 No longer check last store for clobber of address register.
129 2018-08-02 Martin Liska <mliska@suse.cz>
131 PR gcov-profile/86817
132 * gcov.c (process_all_functions): New function.
134 (process_file): Move functions processing to
135 process_all_functions.
137 2018-08-02 David Malcolm <dmalcolm@redhat.com>
139 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
140 "const" to the "gimple *" and "rtx_insn *" parameters.
141 * dumpfile.h (dump_user_location_t::dump_user_location_t):
143 (dump_location_t::dump_location_t): Likewise.
145 2018-08-01 Martin Sebor <msebor@redhat.com>
147 PR tree-optimization/86650
148 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
149 rather than a "gcall *". Directly pass the data of interest
150 to percent_K_format, rather than building a temporary CALL_EXPR
152 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
153 (gimple_fold_builtin_strncat): Adjust.
154 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
155 gcall* argument with gimple*.
156 * gimple-ssa-warn-restrict.c (check_call): Same.
157 (wrestrict_dom_walker::before_dom_children): Same.
158 (builtin_access::builtin_access): Same.
159 (check_bounds_or_overlap): Same
160 (maybe_diag_overlap): Same.
161 (maybe_diag_offset_bounds): Same.
162 * tree-diagnostic.c (default_tree_printer): Move usage of
163 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
165 * tree-pretty-print.c (percent_K_format): Add argument.
166 * tree-pretty-print.h: Add argument.
167 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
169 (maybe_diag_stxncpy_trunc): Same.
170 (handle_builtin_stxncpy): Same.
171 (handle_builtin_strcat): Same.
173 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
175 * match.pd: Optimise pointer range checks.
177 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
179 PR tree-optimization/86758
180 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
181 to remove pattern statements.
183 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
185 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
186 result of dfs_enumerate_from when constructing stmt_vec_infos,
187 instead of additionally calling get_loop_body.
189 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
191 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
193 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
194 When creating an iv, assert that the step is not known to be zero.
195 (vect_setup_realignment): Update call accordingly.
196 * tree-vect-stmts.c (vectorizable_store): Likewise.
197 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
199 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
201 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
202 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
203 (vectorizable_reduction): Likewise.
204 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
205 (vect_detect_hybrid_slp_stmts): Likewise.
206 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
208 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
210 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
211 (wide_int_set_zero_nonzero_bits): ...this.
212 (zero_nonzero_bits_from_vr): Rename to...
213 (vrp_set_zero_nonzero_bits): ...this.
214 (extract_range_from_multiplicative_op_1): Abstract wide int
216 (wide_int_range_multiplicative_op): ...here.
217 (extract_range_from_binary_expr_1): Extract wide int binary
218 operations into their own functions.
219 (wide_int_range_lshift): New.
220 (wide_int_range_can_optimize_bit_op): New.
221 (wide_int_range_shift_undefined_p): New.
222 (wide_int_range_bit_xor): New.
223 (wide_int_range_bit_ior): New.
224 (wide_int_range_bit_and): New.
225 (wide_int_range_trunc_mod): New.
226 (extract_range_into_wide_ints): New.
227 (vrp_shift_undefined_p): New.
228 (extract_range_from_multiplicative_op): New.
229 (vrp_can_optimize_bit_op): New.
230 * tree-vrp.h (value_range::dump): New.
231 (wide_int_range_multiplicative_op): New.
232 (wide_int_range_lshift):New.
233 (wide_int_range_shift_undefined_p): New.
234 (wide_int_range_bit_xor): New.
235 (wide_int_range_bit_ior): New.
236 (wide_int_range_bit_and): New.
237 (wide_int_range_trunc_mod): New.
238 (zero_nonzero_bits_from_bounds): Rename to...
239 (wide_int_set_zero_nonzero_bits): ...this.
240 (zero_nonzero_bits_from_vr): Rename to...
241 (vrp_set_zero_nonzero_bits): ...this.
242 (range_easy_mask_min_max): Rename to...
243 (wide_int_range_can_optimize_bit_op): this.
244 * vr-values.c (simplify_bit_ops_using_ranges): Rename
245 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
247 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
249 * tree-vectorizer.h (vect_orig_stmt): New function.
250 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
251 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
252 (vect_create_epilog_for_reduction): Likewise.
253 (vectorizable_live_operation): Likewise.
254 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
255 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
256 * tree-vect-stmts.c (vectorizable_call): Likewise.
257 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
259 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
261 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
263 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
264 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
265 (vect_transform_loop): Likewise.
266 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
268 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
270 * tree-vectorizer.h (vect_schedule_slp): Return void.
271 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
272 (vect_schedule_slp): Likewise.
274 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
276 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
278 (vect_transform_loop): Update calls accordingly. Schedule SLP
279 instances before the main loop, if any exist.
281 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
283 PR tree-optimization/86749
284 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
285 If the lhs is used in a COND_EXPR, check that it is being used
286 as the "then" or "else" value.
288 2018-08-01 Tom de Vries <tdevries@suse.de>
291 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
292 speculation_safe_value_not_needed.
294 2018-08-01 Richard Biener <rguenther@suse.de>
296 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
299 2018-08-01 Martin Liska <mliska@suse.cz>
301 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
302 format how successful transformation is dumped.
303 (gimple_mod_pow2_value_transform): Likewise.
304 (gimple_mod_subtract_transform): Likewise.
305 (gimple_stringops_transform): Likewise.
307 2018-08-01 Martin Liska <mliska@suse.cz>
310 * value-prof.c (interesting_stringop_to_profile_p):
311 Simplify the code and add BUILT_IN_MEMMOVE.
312 (gimple_stringops_transform): Likewise.
314 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
316 * config/aarch64/aarch64-simd.md
317 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
318 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
319 use GPI iterator instead of SI mode.
321 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
323 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
324 rs6000_speculation_barrier.
325 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
326 new barrier pattern name.
328 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
330 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
331 (speculation_barrier): New insn.
333 2018-08-01 Richard Biener <rguenther@suse.de>
336 * graphite.h: Include isl/id.h and isl/space.h to allow build
339 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
342 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
343 mode for COFF targets.
344 * defaults.h (TARGET_COFF): Define.
345 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
346 TARGET_COFF): Define.
347 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
348 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
349 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
351 2018-07-31 Alexandre Oliva <oliva@adacore.com>
353 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
354 * gimple-streamer-out.c (output_bb): Save it.
355 * lto-streamer-in.c (input_struct_function_base): Restore
356 instance discriminator if available. Create map on demand.
357 * lto-streamer-out.c (output_struct_function_base): Save it if
359 * final.c (decl_to_instance_map): Document LTO strategy.
361 2018-07-31 Alexandre Oliva <oliva@adacore.com>
362 Olivier Hainque <hainque@adacore.com>
364 * debug.h (decl_to_instance_map_t): New type.
365 (decl_to_instance_map): Declare.
366 (maybe_create_decl_to_instance_map): New inline function.
367 * final.c (bb_discriminator, last_bb_discriminator): New statics,
368 to track basic block discriminators.
369 (final_start_function_1): Initialize them.
370 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
372 (decl_to_instance_map): New variable.
373 (map_decl_to_instance, maybe_set_discriminator): New functions.
374 (notice_source_line): Set discriminator.
376 2018-07-31 Ian Lance Taylor <iant@golang.org>
378 * targhooks.c (default_have_speculation_safe_value): Add
381 2018-07-31 David Malcolm <dmalcolm@redhat.com>
383 * dump-context.h: Include "pretty-print.h".
384 (dump_context::refresh_dumps_are_enabled): New decl.
385 (dump_context::emit_item): New decl.
386 (class dump_context): Add fields "m_test_pp" and
388 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
389 (temp_dump_context::get_dumped_text): New decl.
390 (class temp_dump_context): Add field "m_pp".
391 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
392 (dump_context::refresh_dumps_are_enabled): ...and add a test for
394 (set_dump_file): Update for above change.
395 (set_alt_dump_file): Likewise.
396 (dump_loc): New overload, taking a pretty_printer *.
397 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
398 to any test pretty-printer.
399 (make_item_for_dump_gimple_stmt): New function, adapted from
400 optinfo::add_gimple_stmt.
401 (dump_context::dump_gimple_stmt): Call it, and use the result,
402 eliminating the direct usage of dump_file and alt_dump_file in
403 favor of indirectly using them via emit_item.
404 (make_item_for_dump_gimple_expr): New function, adapted from
405 optinfo::add_gimple_expr.
406 (dump_context::dump_gimple_expr): Call it, and use the result,
407 eliminating the direct usage of dump_file and alt_dump_file in
408 favor of indirectly using them via emit_item.
409 (make_item_for_dump_generic_expr): New function, adapted from
411 (dump_context::dump_generic_expr): Call it, and use the result,
412 eliminating the direct usage of dump_file and alt_dump_file in
413 favor of indirectly using them via emit_item.
414 (make_item_for_dump_printf_va): New function, adapted from
415 optinfo::add_printf_va.
416 (make_item_for_dump_printf): New function.
417 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
418 and use the result, eliminating the direct usage of dump_file and
419 alt_dump_file in favor of indirectly using them via emit_item.
420 (make_item_for_dump_dec): New function.
421 (dump_context::dump_dec): Call it, and use the result,
422 eliminating the direct usage of dump_file and alt_dump_file in
423 favor of indirectly using them via emit_item.
424 (make_item_for_dump_symtab_node): New function, adapted from
425 optinfo::add_symtab_node.
426 (dump_context::dump_symtab_node): Call it, and use the result,
427 eliminating the direct usage of dump_file and alt_dump_file in
428 favor of indirectly using them via emit_item.
429 (dump_context::begin_scope): Reimplement, avoiding direct usage
430 of dump_file and alt_dump_file in favor of indirectly using them
432 (dump_context::emit_item): New member function.
433 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
434 Set up test pretty-printer on the underlying context. Call
435 refresh_dumps_are_enabled.
436 (temp_dump_context::~temp_dump_context): Call
437 refresh_dumps_are_enabled.
438 (temp_dump_context::get_dumped_text): New member function.
439 (selftest::verify_dumped_text): New function.
440 (ASSERT_DUMPED_TEXT_EQ): New macro.
441 (selftest::test_capture_of_dump_calls): Run all tests twice, with
442 and then without optinfo enabled. Add uses of
443 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
444 * dumpfile.h: Update comment for the dump_* API.
445 * optinfo-emit-json.cc
446 (selftest::test_building_json_from_dump_calls): Update for new
447 param for temp_dump_context ctor.
448 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
450 (optinfo_item::~optinfo_item): Likewise.
451 (optinfo::add_item): New member function.
452 (optinfo::emit): Update comment.
453 (optinfo::add_string): Delete.
454 (optinfo::add_printf): Delete.
455 (optinfo::add_printf_va): Delete.
456 (optinfo::add_gimple_stmt): Delete.
457 (optinfo::add_gimple_expr): Delete.
458 (optinfo::add_tree): Delete.
459 (optinfo::add_symtab_node): Delete.
460 (optinfo::add_dec): Delete.
461 * optinfo.h (class dump_context): New forward decl.
462 (optinfo::add_item): New decl.
463 (optinfo::add_string): Delete.
464 (optinfo::add_printf): Delete.
465 (optinfo::add_printf_va): Delete.
466 (optinfo::add_gimple_stmt): Delete.
467 (optinfo::add_gimple_expr): Delete.
468 (optinfo::add_tree): Delete.
469 (optinfo::add_symtab_node): Delete.
470 (optinfo::add_dec): Delete.
471 (optinfo::add_poly_int): Delete.
472 (optinfo_item::optinfo_item): Remove "owned" param.
473 (class optinfo_item): Remove field "m_owned".
475 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
478 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
479 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
480 requested variable alignment.
481 (expand_one_ssa_partition): Likewise.
482 (expand_one_var): Likewise.
484 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
486 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
487 to speculation_safe_value_not_needed.
489 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
491 * targhooks.h (speculation_safe_value_not_needed): New prototype.
492 * targhooks.c (speculation_safe_value_not_needed): New function.
493 * target.def (have_speculation_safe_value): Update documentation.
494 * doc/tm.texi: Regenerated.
496 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
498 * config/aarch64/iterators.md (ALLI_TI): New iterator.
499 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
501 (despeculate_copy<ALLI:mode>_insn): New insn.
502 (despeculate_copyti_insn): New insn.
503 (despeculate_simple<ALLI:mode>): New insn
504 (despeculate_simpleti): New insn.
505 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
507 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
508 aarch64_speculation_safe_value.
509 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
511 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
513 * config/aarch64/aarch64-speculation.cc: New file.
514 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
515 before pass_reorder_blocks.
516 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
518 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
519 X14 and X15 when tracking speculation.
520 * config/aarch64/aarch64.md (register name constants): Add
521 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
522 (unspec): Add UNSPEC_SPECULATION_TRACKER.
523 (speculation_barrier): New insn attribute.
524 (cmp<mode>): Allow SP in comparisons.
525 (speculation_tracker): New insn.
526 (speculation_barrier): Add speculation_barrier attribute.
527 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
528 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
529 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
531 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
533 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
534 aarch64_track_speculation is true.
535 (tb<optab><mode>1): Likewise.
536 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
537 generate CB[N]Z when tracking speculation.
538 (aarch64_split_compare_and_swap): Likewise.
539 (aarch64_split_atomic_op): Likewise.
541 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
543 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
545 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
547 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
548 (speculation_barrier): New insn.
550 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
552 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
553 * config/arm/arm.md (speculation_barrier): New expand.
554 (speculation_barrier_insn): New pattern.
556 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
558 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
559 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
560 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
561 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
563 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
564 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
565 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
566 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
567 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
568 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
569 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
570 * builtins.c (expand_speculation_safe_value): New function.
571 (expand_builtin): Call it.
572 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
573 * doc/extend.texi: Document __builtin_speculation_safe_value.
574 * doc/md.texi: Document "speculation_barrier" pattern.
575 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
576 TARGET_HAVE_SPECULATION_SAFE_VALUE.
577 * doc/tm.texi: Regenerated.
578 * target.def (have_speculation_safe_value, speculation_safe_value): New
580 * targhooks.c (default_have_speculation_safe_value): New function.
581 (default_speculation_safe_value): New function.
582 * targhooks.h (default_have_speculation_safe_value): Add prototype.
583 (default_speculation_safe_value): Add prototype.
585 2018-07-31 David Malcolm <dmalcolm@redhat.com>
587 * dump-context.h (dump_context::dump_loc): New decl.
588 * dumpfile.c (dump_context::dump_loc): New member function.
589 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
590 and dump_gimple_stmt.
591 (dump_context::dump_gimple_expr_loc): Likewise, using
593 (dump_context::dump_generic_expr_loc): Likewise, using
595 (dump_context::dump_printf_loc_va): Likewise, using
597 (dump_context::begin_scope): Explicitly using the global function
598 "dump_loc", rather than the member function.
600 2018-07-31 Martin Sebor <msebor@redhat.com>
602 PR tree-optimization/86741
603 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
605 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
607 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
608 depend on whether prefetch instructions will be emitted or not.
609 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
610 will be emitted or not.
611 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
612 (TARGET_SETMEM_PFD): New macros.
614 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
616 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
617 (NULL_STMT_VEC_INFO): Delete.
618 (stmt_vec_info::operator*): Likewise.
619 (stmt_vec_info::operator gimple *): Likewise.
620 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
621 of NULL_STMT_VEC_INFO.
622 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
623 (vect_reassociating_reduction_p): Likewise.
624 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
625 (vectorizable_store): Likewise.
626 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
627 (vec_info::free_stmt_vec_infos): Likewise.
629 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
631 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
632 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
634 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
636 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
637 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
638 (vec_info::free_stmt_vec_info): New private member functions.
639 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
640 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
641 * tree-parloops.c (gather_scalar_reductions): Remove calls to
642 set_stmt_vec_info_vec and free_stmt_vec_infos.
643 * tree-vect-loop.c (_loop_vec_info): Remove call to
644 set_stmt_vec_info_vec.
645 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
646 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
647 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
648 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
649 (vec_info::free_stmt_vec_info): ...these new functions. Remove
650 assignments in {vec_info::,}new_stmt_vec_info that are redundant
651 with the clearing in the xcalloc.
652 (stmt_vec_info_vec): Delete.
653 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
654 (vectorize_loops): Likewise.
655 (vec_info::~vec_info): Remove argument from call to
657 (vec_info::add_stmt): Remove vinfo argument from call to
660 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
662 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
663 rather than a gimple stmt.
664 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
665 information for pattern statements when passed the original
666 statement; instead wait to be passed the pattern statement itself.
667 Don't call set_vinfo_for_stmt here.
668 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
669 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
671 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
672 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
673 stmt_vec_infos entry.
675 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
677 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
678 * tree-vectorizer.c (vec_info::replace_stmt): New function.
679 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
680 * tree-vect-stmts.c (vectorizable_call): Likewise.
681 (vectorizable_simd_clone_call): Likewise.
683 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
685 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
686 * tree-vectorizer.c (vec_info::remove_stmt): New function.
687 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
688 * tree-vect-loop.c (vect_transform_loop): Likewise.
689 * tree-vect-slp.c (vect_schedule_slp): Likewise.
690 * tree-vect-stmts.c (vect_remove_stores): Likewise.
692 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
694 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
695 (vect_dr_stmt): Delete.
696 * tree-vectorizer.c (vec_info::lookup_dr): New function.
697 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
699 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
700 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
701 (vect_verify_datarefs_alignment, vect_peeling_supportable)
702 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
703 (vect_analyze_data_refs): Likewise.
704 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
706 (vect_find_same_alignment_drs): Likewise.
707 (vect_slp_analyze_node_dependences): Update calls accordingly.
708 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
709 instead of DR_VECT_AUX.
710 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
711 of a vector data references. Use vec_info::lookup_dr instead of
713 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
714 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
715 instead of DR_VECT_AUX.
717 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
719 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
721 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
723 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
724 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
725 (vect_gen_prolog_loop_niters): Likewise.
727 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
729 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
730 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
731 (vect_known_alignment_in_bytes, vect_dr_behavior)
732 (vect_get_scalar_dr_size): Take references as dr_vec_infos
733 instead of data_references. Update calls to other routines for
734 which the same change has been made.
735 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
736 dr_vec_infos instead of stmt_vec_infos.
737 (vect_analyze_data_ref_dependence): Update call accordingly.
738 (vect_slp_analyze_data_ref_dependence)
739 (vect_record_base_alignments): Use DR_VECT_AUX.
740 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
741 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
742 (vector_alignment_reachable_p, vect_get_data_access_cost)
743 (vect_peeling_supportable, vect_analyze_group_access_1)
744 (vect_analyze_group_access, vect_analyze_data_ref_access)
745 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
746 (vect_compile_time_alias, vect_small_gap_p)
747 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
748 (vect_supportable_dr_alignment): Take references as dr_vec_infos
749 instead of data_references. Update calls to other routines for
750 which the same change has been made.
751 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
752 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
753 (vect_slp_analyze_and_verify_node_alignment)
754 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
755 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
756 (vect_setup_realignment): Use dr_vec_infos. Update calls after
758 (_vect_peel_info::dr): Replace with...
759 (_vect_peel_info::dr_info): ...this new field.
760 (vect_peeling_hash_get_most_frequent)
761 (vect_peeling_hash_choose_best_peeling): Update accordingly.
762 (vect_peeling_hash_get_lowest_cost):
763 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
764 routines for which the same change has been made.
765 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
767 * tree-vect-loop-manip.c (get_misalign_in_elems)
768 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
770 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
771 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
772 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
773 (get_group_load_store_type, get_negative_load_store_type)
774 (vect_get_data_ptr_increment, vectorizable_store)
775 (vectorizable_load): Likewise.
776 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
777 Update calls to other routines for which the same change has been made.
779 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
781 * tree-vectorizer.h (vec_info::move_dr): New member function.
782 (dataref_aux): Rename to...
783 (dr_vec_info): ...this and add "dr" and "stmt" fields.
784 (_stmt_vec_info::dr_aux): Update accordingly.
785 (_stmt_vec_info::data_ref_info): Delete.
786 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
787 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
788 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
789 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
791 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
792 (STMT_VINFO_DR_INFO): New macro.
793 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
794 (set_dr_misalignment): Update after rename of dataref_aux.
795 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
796 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
797 initialization of STMT_VINFO_DATA_REF.
798 * tree-vectorizer.c (vec_info::move_dr): New function.
799 * tree-vect-patterns.c (vect_recog_bool_pattern)
800 (vect_recog_mask_conversion_pattern)
801 (vect_recog_gather_scatter_pattern): Use it.
802 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
803 the "dr" and "stmt" fields of dr_vec_info instead of
806 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
808 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
809 (is_pattern_stmt_p): Use it.
810 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
811 on pattern statements.
813 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
815 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
816 original stmt as a stmt_vec_info rather than a gimple stmt.
817 (vect_pattern_recog_1): Take the statement directly as a
818 stmt_vec_info, rather than via a gimple_stmt_iterator.
819 Update call to vect_mark_pattern_stmts.
820 (vect_pattern_recog): Update calls accordingly.
822 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
824 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
825 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
826 a vect_def_type for the first argument.
827 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
828 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
829 operand if it isn't defined by a vectorized statement.
830 (vect_build_gather_load_calls): Remove the mask_dt argument and
831 update calls to vect_get_vec_def_for_stmt_copy.
832 (vectorizable_bswap): Likewise the dt argument.
833 (vectorizable_call): Update calls to vectorizable_bswap and
834 vect_get_vec_def_for_stmt_copy.
835 (vectorizable_simd_clone_call, vectorizable_assignment)
836 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
837 (vectorizable_comparison): Update calls to
838 vect_get_vec_def_for_stmt_copy.
839 (vectorizable_store): Likewise. Remove now-unnecessary calls to
841 (vect_get_loop_based_defs): Remove dt argument and update call
842 to vect_get_vec_def_for_stmt_copy.
843 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
844 and vect_get_vec_def_for_stmt_copy.
845 (vectorizable_load): Update calls to vect_build_gather_load_calls
846 and vect_get_vec_def_for_stmt_copy.
847 * tree-vect-loop.c (vect_create_epilog_for_reduction)
848 (vectorizable_reduction, vectorizable_live_operation): Update calls
849 to vect_get_vec_def_for_stmt_copy.
851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
853 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
854 and gimple stmt arguments with a stmt_vec_info.
855 (vect_record_base_alignments): Update calls accordingly.
856 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
857 and gimple stmt arguments with a stmt_vec_info.
858 (vect_build_slp_tree_1): Remove vinfo argument and update call
859 to vect_record_max_nunits.
860 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
861 and vect_record_max_nunits.
863 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
865 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
866 file and take a stmt_vec_info instead of a gimple stmt.
867 (supportable_widening_operation, vect_finish_replace_stmt)
868 (vect_finish_stmt_generation, vect_get_store_rhs)
869 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
870 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
871 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
872 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
873 (vect_create_data_ref_ptr, bump_vector_ptr)
874 (vect_permute_store_chain, vect_setup_realignment)
875 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
876 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
877 (vectorizable_reduction, vectorizable_induction)
878 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
879 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
880 than gimple stmts as arguments.
881 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
882 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
883 (can_group_stmts_p, vect_check_gather_scatter)
884 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
885 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
886 (vect_permute_load_chain, vect_shift_permute_load_chain)
887 (vect_transform_grouped_load)
888 (vect_record_grouped_load_vectors): Likewise.
889 * tree-vect-loop.c (vect_fixup_reduc_chain)
890 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
891 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
892 (vectorizable_reduction, vectorizable_induction)
893 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
894 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
895 (vect_get_load_store_mask): Likewise.
896 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
897 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
898 * tree-vect-stmts.c (vect_mark_relevant)
899 (is_simple_and_all_uses_invariant)
900 (exist_non_indexing_operands_for_use_p, process_use)
901 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
902 (vect_get_vec_def_for_operand, vect_get_vec_defs)
903 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
904 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
905 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
906 (get_negative_load_store_type, get_load_store_type)
907 (vect_check_load_store_mask, vect_check_store_rhs)
908 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
909 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
910 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
911 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
912 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
913 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
914 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
915 (supportable_widening_operation): Likewise.
917 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
919 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
920 a stmt_vec_info instead of a gcall.
921 (vect_check_gather_scatter): Update call accordingly.
922 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
924 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
925 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
926 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
927 instead of a gimple stmt.
928 (vect_transform_loop): Update calls accordingly.
929 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
930 stmt_vec_infos instead of gimple stmts.
931 (vect_analyze_slp_instance): Update use accordingly.
932 * tree-vect-stmts.c (read_vector_array, write_vector_array)
933 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
934 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
935 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
936 (vect_gen_widened_results_half, vect_get_loop_based_defs)
937 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
938 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
941 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
943 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
944 of the worklist from a vector of gimple stmts to a vector of
946 * tree-vect-stmts.c (vect_mark_relevant, process_use)
947 (vect_mark_stmts_to_be_vectorized): Likewise
949 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
951 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
952 statement before passing it to vect_analyze_stmt.
953 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
954 the chain of phi vector definitions. Track the exit phi via its
956 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
957 STMT_VINFO_REDUC_DEF.
958 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
959 stmt_vec_infos to handle the statement chains.
960 (vect_get_slp_defs): Record the first statement in the node
961 using a stmt_vec_info.
962 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
963 statements here and pass their stmt_vec_info down to subroutines.
964 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
965 down to vect_finish_stmt_generation.
966 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
967 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
968 stmt_vec_infos to subroutines.
969 (vect_remove_stores): Use stmt_vec_infos to handle the statement
972 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
974 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
975 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
976 (vect_permute_store_chain, vect_setup_realignment)
977 (vect_permute_load_chain, vect_shift_permute_load_chain)
978 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
979 stmts internally, and when passing values to other vectorizer routines.
980 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
981 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
982 (vect_analyze_loop_operations, get_initial_def_for_reduction)
983 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
984 (vectorizable_reduction, vectorizable_induction)
985 (vectorizable_live_operation, vect_transform_loop_stmt)
986 (vect_transform_loop): Likewise.
987 * tree-vect-patterns.c (vect_reassociating_reduction_p)
988 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
989 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
990 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
991 (vect_slp_analyze_node_operations_1): Likewise.
992 * tree-vect-stmts.c (vect_mark_relevant, process_use)
993 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
994 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
995 (vect_finish_stmt_generation_1, get_group_load_store_type)
996 (get_load_store_type, vect_build_gather_load_calls)
997 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
998 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
999 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1000 (vectorizable_store, vectorizable_load, vectorizable_condition)
1001 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
1002 (supportable_widening_operation): Likewise.
1003 (vect_get_vector_types_for_stmt): Likewise.
1004 * tree-vectorizer.h (vect_dr_behavior): Likewise.
1006 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1008 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1009 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
1010 (vect_permute_store_chain, vect_permute_load_chain)
1011 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
1012 repeated stmt_vec_info lookups.
1013 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
1014 (vect_update_ivs_after_vectorizer): Likewise.
1015 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
1016 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1017 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
1018 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1019 (vect_bb_slp_scalar_cost): Likewise.
1020 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
1022 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1024 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1025 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
1026 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
1027 to get gassigns and gcalls, rather than operating on generc gimple
1029 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
1030 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
1031 (vectorizable_load, vect_analyze_stmt): Likewise.
1032 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
1034 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1036 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
1037 return stmt_vec_infos rather than gimple stmts. Do not accept
1039 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
1041 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1042 Update use of get_later_stmt.
1043 (vect_get_constant_vectors): Update call accordingly.
1044 (vect_schedule_slp_instance): Likewise
1045 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
1046 (vect_slp_analyze_instance_dependence): Likewise.
1047 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
1049 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1051 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
1052 (stmt_info_for_cost::stmt_info): ...this new field.
1053 (add_stmt_costs): Update accordingly.
1054 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1055 (vect_get_known_peeling_cost): Likewise.
1056 (vect_estimate_min_profitable_iters): Likewise.
1057 * tree-vect-stmts.c (record_stmt_cost): Likewise.
1059 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1061 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
1062 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1063 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
1065 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
1067 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1069 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
1070 a gimple stmt to a stmt_vec_info.
1071 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1073 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1075 * tree-vectorizer.h (vec_info::grouped_stores): Change from
1076 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1077 (_loop_vec_info::reduction_chains): Likewise.
1078 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
1080 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1082 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1084 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
1085 a gimple stmt to a stmt_vec_info.
1086 (_stmt_vec_info::next_element): Likewise.
1087 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
1088 (vect_slp_analyze_and_verify_node_alignment)
1089 (vect_analyze_group_access_1, vect_analyze_group_access)
1090 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
1091 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
1092 (vect_supportable_dr_alignment): Update accordingly.
1093 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
1094 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
1095 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
1096 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1097 * tree-vect-slp.c (vect_build_slp_tree_1)
1098 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
1099 (vect_split_slp_store_group, vect_analyze_slp_instance)
1100 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
1101 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
1102 (get_group_load_store_type, get_load_store_type)
1103 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
1104 (vect_transform_stmt, vect_remove_stores): Likewise.
1106 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1108 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
1110 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
1111 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
1112 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
1113 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
1114 (vector_alignment_reachable_p, vect_get_data_access_cost)
1115 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
1116 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
1117 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
1118 (vect_analyze_group_access_1, vect_analyze_group_access)
1119 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
1120 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
1121 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
1122 result of vect_dr_stmt and use the stmt_vec_info instead of
1123 the associated gimple stmt.
1124 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1125 (vect_gen_prolog_loop_niters): Likewise.
1126 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1128 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1130 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
1131 to a vec<stmt_vec_info>.
1132 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
1133 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
1135 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
1136 to a vec<stmt_vec_info>.
1137 (bst_traits::value_type, bst_traits::value_type): Likewise.
1138 (bst_traits::hash): Update accordingly.
1139 (vect_get_and_check_slp_defs): Change the stmts parameter from
1140 a vec<gimple *> to a vec<stmt_vec_info>.
1141 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
1142 (vect_build_slp_tree): Likewise.
1143 (vect_build_slp_tree_2): Likewise. Update uses of
1144 SLP_TREE_SCALAR_STMTS.
1145 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
1146 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
1147 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
1148 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
1149 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
1150 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
1151 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
1152 (vect_get_constant_vectors, vect_get_slp_defs)
1153 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
1154 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
1155 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
1156 instead of gimple stmts.
1157 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
1158 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
1159 (vect_slp_analyze_instance_dependence): Update uses of
1160 SLP_TREE_SCALAR_STMTS.
1161 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1162 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
1163 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
1164 (get_initial_defs_for_reduction): Likewise.
1165 (vect_create_epilog_for_reduction): Likewise.
1166 (vectorize_fold_left_reduction): Likewise.
1167 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
1168 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
1169 (can_vectorize_live_stmts): Likewise.
1171 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1173 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
1174 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
1175 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
1177 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
1179 (gather_scalar_reductions): Update after above interface changes.
1180 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
1181 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
1183 (vect_force_simple_reduction): Likewise.
1184 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
1185 LOOP_VINFO_REDUCTIONS.
1186 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1188 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1190 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
1191 a gimple stmt to a stmt_vec_info.
1192 * tree-vect-loop.c (vect_active_double_reduction_p)
1193 (vect_force_simple_reduction, vectorizable_reduction): Update
1196 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1198 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
1199 vec<gimple *> to a vec<stmt_vec_info>.
1200 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
1201 the reduction_phis argument from a vec<gimple *> to a
1203 (vectorizable_reduction): Likewise the phis local variable that
1204 is passed to vect_create_epilog_for_reduction. Update for new type
1205 of SLP_TREE_VEC_STMTS.
1206 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
1207 (vectorizable_live_operation): Likewise.
1208 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
1209 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
1211 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1213 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
1214 a gimple stmt to a stmt_vec_info.
1215 (vectorizable_condition, vectorizable_live_operation)
1216 (vectorizable_reduction, vectorizable_induction): Pass back the
1217 vectorized statement as a stmt_vec_info.
1218 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
1219 use of STMT_VINFO_VEC_STMT.
1220 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
1221 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
1222 as stmt_vec_infos rather than gimple stmts.
1223 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
1225 (vectorizable_live_operation): Likewise.
1226 (vectorizable_reduction, vectorizable_induction): Likewise,
1227 updating use of STMT_VINFO_VEC_STMT.
1228 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
1229 of STMT_VINFO_VEC_STMT.
1230 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
1231 (vectorizable_simd_clone_call, vectorizable_conversion)
1232 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1233 (vectorizable_store, vectorizable_load, vectorizable_condition)
1234 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
1235 from a gimple stmt to a stmt_vec_info.
1236 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
1237 pointer to a stmt_vec_info to the vectorizable_* routines.
1239 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1241 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
1242 a gimple stmt to a stmt_vec_info.
1243 (is_pattern_stmt_p): Update accordingly.
1244 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
1245 (vect_record_grouped_load_vectors): Likewise.
1246 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1247 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
1248 (vect_model_reduction_cost): Likewise.
1249 (vect_create_epilog_for_reduction): Likewise.
1250 (vectorizable_reduction, vectorizable_induction): Likewise.
1251 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1252 Return the stmt_vec_info for the pattern statement.
1253 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
1254 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
1255 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
1256 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
1257 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1258 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
1259 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
1260 (free_stmt_vec_info, vect_is_simple_use): Likewise.
1262 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1264 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
1265 (vect_finish_stmt_generation): Likewise.
1266 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1267 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
1268 (vect_build_gather_load_calls): Use the return value of the above
1269 functions instead of a separate call to vinfo_for_stmt. Use narrow
1270 scopes for the input gimple stmt and wider scopes for the associated
1271 stmt_vec_info. Use vec_info::lookup_def when setting these
1272 stmt_vec_infos from an SSA_NAME definition.
1273 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
1274 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
1275 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
1276 (vectorizable_store, vectorizable_load, vectorizable_condition)
1277 (vectorizable_comparison): Likewise.
1278 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
1279 (vectorizable_reduction): Likewise.
1281 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1283 * tree-vectorizer.h (vect_is_simple_use): Add an optional
1284 stmt_vec_info * parameter before the optional gimple **.
1285 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1286 (process_use, vect_get_vec_def_for_operand_1): Update callers.
1287 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
1288 * tree-vect-loop.c (vectorizable_reduction): Likewise.
1289 (vectorizable_live_operation): Likewise.
1290 * tree-vect-patterns.c (type_conversion_p): Likewise.
1291 (vect_look_through_possible_promotion): Likewise.
1292 (vect_recog_rotate_pattern): Likewise.
1293 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1297 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
1298 a typedef to a wrapper class.
1299 (NULL_STMT_VEC_INFO): New macro.
1300 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
1301 (stmt_vec_info::operator*): New function.
1302 (stmt_vec_info::operator gimple *): Likewise.
1303 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
1304 (add_stmt_costs): Likewise.
1305 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
1306 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
1307 (vect_get_known_peeling_cost): Likewise.
1308 (vect_estimate_min_profitable_iters): Likewise.
1309 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1310 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
1311 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1312 (vectorizable_store, free_stmt_vec_infos): Likewise.
1313 (new_stmt_vec_info): Change return type of xcalloc to
1316 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1318 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
1319 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
1320 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
1321 a single_imm_use-based sequence.
1322 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
1324 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1326 * tree-vectorizer.h (vec_info::lookup_def): Declare.
1327 * tree-vectorizer.c (vec_info::lookup_def): New function.
1328 * tree-vect-patterns.c (vect_get_internal_def): Use it.
1329 (vect_widened_op_tree): Likewise.
1330 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1331 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
1332 (vectorizable_reduction): Likewise.
1333 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
1335 (vect_is_slp_reduction): Update calls accordingly. Use
1336 vec_info::lookup_def.
1337 (vect_is_simple_reduction): Likewise
1338 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
1340 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1342 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
1343 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
1344 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
1346 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
1347 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
1348 (vect_update_vf_for_slp, vect_analyze_loop_operations)
1349 (vect_is_slp_reduction, vectorizable_induction)
1350 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
1351 * tree-vect-patterns.c (vect_init_pattern_stmt):
1352 (vect_determine_min_output_precision_1, vect_determine_precisions)
1353 (vect_pattern_recog): Likewise.
1354 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
1355 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
1356 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
1357 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
1358 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
1359 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
1360 info field from a loop to a loop_vec_info.
1362 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1364 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
1365 (vec_info::add_stmt): Declare.
1366 * tree-vectorizer.c (vec_info::add_stmt): New function.
1367 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
1368 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
1369 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
1370 (vectorizable_induction): Likewise.
1371 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
1372 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1373 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
1374 (vectorizable_load): Likewise.
1375 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1376 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
1377 (vect_recog_gather_scatter_pattern): Likewise.
1378 (append_pattern_def_seq): Likewise. Remove a check that is
1379 performed by add_stmt itself.
1381 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1383 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
1384 which make_ssa_name was called with new_stmt before new_stmt
1387 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1389 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
1391 (vect_is_slp_reduction): ...here...
1392 (vect_is_simple_reduction): ...and here. Remove repetition of tests
1393 that are already known to be false.
1395 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1397 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
1398 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
1399 STMT_VINFO_NUM_SLP_USES when it's true.
1400 (vect_free_slp_instance): Add a final_p parameter and pass it to
1402 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
1403 (vect_analyze_slp_instance): Likewise.
1404 (vect_slp_analyze_operations): Likewise.
1405 (vect_slp_analyze_bb_1): Likewise.
1406 * tree-vectorizer.c (vec_info): Likewise.
1407 * tree-vect-loop.c (vect_transform_loop): Likewise.
1409 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1411 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1412 function is not called for second and subsequent members of
1415 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1417 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
1418 cases for nested loops from here to ...
1419 (vect_create_epilog_for_reduction): ...here. Only call
1420 vect_is_simple_use for inner-loop reductions.
1422 2018-07-31 Martin Liska <mliska@suse.cz>
1424 PR gcov-profile/85338
1425 PR gcov-profile/85350
1426 PR gcov-profile/85372
1427 * profile.c (struct location_triplet): New.
1428 (struct location_triplet_hash): Likewise.
1429 (output_location): Do not output a BB that
1430 is already recorded for a line.
1431 (branch_prob): Use streamed_locations.
1433 2018-07-31 Martin Liska <mliska@suse.cz>
1435 PR gcov-profile/85370
1436 * coverage.c (coverage_begin_function): Do not mark target
1437 clones as artificial functions.
1439 2018-07-31 Martin Liska <mliska@suse.cz>
1441 PR gcov-profile/83813
1442 PR gcov-profile/84758
1443 PR gcov-profile/85217
1444 PR gcov-profile/85332
1445 * profile.c (branch_prob): Do not record GOTO expressions
1446 for GIMPLE statements which locations are already streamed.
1448 2018-07-31 Olivier Hainque <hainque@adacore.com>
1450 * gcc.c (handle_spec_function): Accept a soft_matched_part
1451 argument, as do_spec_1. Pass it down to ...
1452 (eval_spec_function): Accept a soft_matched_part argument,
1453 and pass it down to ...
1454 (do_spec_2): Accept a soft_matched_part argument, and pass
1455 it down to do_spec_1.
1456 (do_spec_1): Pass soft_matched_part to handle_spec_function.
1457 (handle_braces): Update call to handle_spec_function.
1458 (driver::set_up_specs): Update calls to do_spec_2.
1459 (compare_debug_dump_opt_spec_function): Likewise.
1460 (compare_debug_self_opt_spec_function): Likewise.
1462 2018-07-31 Olivier Hainque <hainque@adacore.com>
1464 * common.opt (nolibc): New option.
1465 * doc/invoke.texi (Link Options): Document it.
1466 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
1467 * config/alpha/linux.h: Likewise.
1468 * config/arc/elf.h: Likewise.
1469 * config/arm/uclinux-elf.h: Likewise.
1470 * config/arm/unknown-elf.h: Likewise.
1471 * config/avr/avrlibc.h: Likewise.
1472 * config/bfin/bfin.h: Likewise.
1473 * config/bfin/linux.h: Likewise.
1474 * config/bfin/uclinux.h: Likewise.
1475 * config/darwin.h: Likewise.
1476 * config/darwin10.h: Likewise.
1477 * config/darwin12.h: Likewise.
1478 * config/gnu-user.h: Likewise.
1479 * config/lm32/uclinux-elf.h: Likewise.
1480 * config/pa/pa-hpux11.h: Likewise.
1481 * config/pa/pa64-hpux.h: Likewise.
1482 * config/sparc/sparc.h: Likewise.
1484 2018-07-31 Olivier Hainque <hainque@adacore.com>
1486 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
1487 undefined variables.
1489 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1492 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
1495 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1497 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1498 terminated string literal.
1500 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1502 PR rtl-optimization/85160
1503 * combine.c (is_just_move): New function.
1504 (try_combine): Allow combining two instructions into two if neither of
1505 the original instructions was a move.
1507 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
1510 * doc/extend.texi (Global Register Variables): Discourage use of type
1512 (Local Register Variables): Likewise.
1514 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
1516 PR tree-optimization/86506
1517 * hwint.h (ceil_log2): Resync with hwint.c implementation.
1519 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
1522 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
1523 hard_regno, make sure no insn between `from` and `to` clobbers it.
1525 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
1526 Tom de Vries <tdevries@suse.de>
1528 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
1529 (PTX_DEFAULT_RUNTIME_DIM): ... this.
1530 (nvptx_goacc_validate_dims): Set default worker and gang dims to
1531 PTX_DEFAULT_RUNTIME_DIM.
1532 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
1534 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
1536 * config/pa/pa.c (pa_output_addr_vec): Align address table.
1537 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
1538 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
1540 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
1542 * config/rs6000/constraints.md (wG constraint): Delete, no longer
1544 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
1545 predicate to reflect toc fusion has been deleted.
1546 (toc_fusion_mem_raw): Delete, no longer used.
1547 (toc_fusion_mem_wrapped): Likewise.
1548 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
1550 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
1551 Delete, no longer used.
1552 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
1553 meant to be used for toc fusion.
1554 (rs6000_debug_print_mode): Delete toc fusion debugging.
1555 (rs6000_debug_reg_global): Likewise.
1556 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
1557 fusion and secondary reload support that were never used.
1558 (rs6000_option_override_internal): Delete TOC fusion, that was only
1559 partially defined, and it did not work unless you also used the
1561 (rs6000_legitimate_address_p): Delete TOC fusion support.
1562 (rs6000_opt_masks): Likewise.
1563 (fusion_wrap_memory_address): Delete function, no longer used.
1564 (fusion_split_address); Delete TOC fusion support.
1565 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
1566 longer used with toc fusion being deleted.
1567 (TARGET_TOC_FUSION_FP): Likewise.
1568 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
1570 (toc fusion spliter): Delete TOC fusion support.
1571 (toc_fusionload_<mode>): Likewise.
1572 (toc_fusionload_di): Likewise.
1573 (fusion_gpr_load_<mode>): Delete generator function, this insn no
1574 longer needs to be named. Rename predicate to delete TOC fusion.
1575 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1576 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1577 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1578 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1579 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
1581 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
1583 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
1584 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
1585 __int128 in built-in function prototypes.
1586 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
1587 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
1589 2018-07-27 Martin Sebor <msebor@redhat.com>
1591 PR tree-optimization/86696
1592 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
1593 types, including enums.
1594 (handle_char_store): Be prepared for the above function to fail.
1596 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
1598 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
1599 when optimization level is lower than 2 or optimize for size.
1601 2018-07-26 Martin Sebor <msebor@redhat.com>
1603 PR tree-optimization/86043
1604 PR tree-optimization/86042
1605 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
1606 (get_string_cst_length): Rename...
1607 (get_min_string_length): ...to this. Add argument.
1608 (handle_char_store): Extend to handle multi-character stores by
1610 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
1611 * tree.h (initializer_zerop): Add argument.
1613 2018-07-26 Jakub Jelinek <jakub@redhat.com>
1616 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
1617 declare target to variables if they have always,{to,from,tofrom} map
1620 2018-07-26 Martin Liska <mliska@suse.cz>
1623 * lto-wrapper.c: Add linker_output as prefix
1624 for ltrans_output_file.
1626 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
1628 PR rtl-optimization/85805
1629 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
1630 value for hard registers if that was written in the same mode.
1632 2018-07-26 Martin Liska <mliska@suse.cz>
1634 PR gcov-profile/86536
1635 * gcov.c (format_gcov): Use printf format %.*f directly
1636 and do not handle special values.
1638 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1640 * common/config/arc/arc-common.c (arc_option_optimization_table):
1641 Update default optimizations for size.
1643 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1645 * config/arc/arc.md (movsf_insn): Add short instruction selection.
1646 * config/arc/constraints.md (CfZ): New constraint.
1647 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
1648 (subsf3_fpu): Likewise.
1649 (cmpsf_fpu): Likewise.
1650 (cmpsf_fpu_uneq): Likewise.
1652 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1654 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
1656 (arc_is_uncached_mem_p): uncached applies to both the variable and
1659 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1661 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
1664 2018-07-25 David Malcolm <dmalcolm@redhat.com>
1666 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
1667 field "m_scopes" from vec to auto_vec.
1669 2018-07-25 Martin Liska <mliska@suse.cz>
1671 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
1674 2018-07-25 Richard Biener <rguenther@suse.de>
1677 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
1678 special wrt context_die late.
1679 (gen_subprogram_die): Re-use DIEs in local scope.
1681 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
1683 PR tree-optimization/86644
1684 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
1686 2018-07-25 Martin Liska <mliska@suse.cz>
1689 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
1690 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
1692 2018-07-25 Martin Liska <mliska@suse.cz>
1695 * params.def: Explain ASan abbreviation and provide
1696 a documentation link.
1698 2018-07-24 Martin Sebor <msebor@redhat.com>
1700 PR tree-optimization/86622
1701 PR tree-optimization/86532
1702 * builtins.h (string_length): Declare.
1703 * builtins.c (c_strlen): Correct handling of non-constant offsets.
1704 (check_access): Be prepared for non-constant length ranges.
1705 (string_length): Make extern.
1706 * expr.c (string_constant): Only handle the minor non-constant
1707 array index. Use string_constant to compute the length of
1708 a generic string constant.
1710 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
1712 PR tree-optimization/86618
1713 * tree-vect-stmts.c (vectorizable_call): Don't take the address
1714 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
1716 2018-07-24 David Malcolm <dmalcolm@redhat.com>
1718 PR tree-optimization/86636
1719 * json.cc (json::object::set): Fix comment. Add assertions.
1720 (json::array::append): Move here from json.h. Add comment and an
1722 (json::string::string): Likewise.
1723 * json.h (json::array::append): Move to json.cc.
1724 (json::string::string): Likewise.
1725 * optinfo-emit-json.cc
1726 (optrecord_json_writer::impl_location_to_json): Assert that we
1727 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
1728 wrapper around it. Expand the location once, rather than three
1730 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
1731 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
1733 (optrecord_json_writer::optinfo_to_json): Likewise, in four
1734 places. Fix some overlong lines.
1736 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
1738 * config/aarch64/aarch64-simd.md
1739 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
1740 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
1741 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
1742 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
1743 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
1744 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
1745 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
1746 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
1747 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
1749 2018-07-24 Jakub Jelinek <jakub@redhat.com>
1752 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
1753 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
1754 and abs_d == d, do the power of two handling if profitable.
1756 2018-07-24 Richard Biener <rguenther@suse.de>
1758 * match.pd: Add BIT_FIELD_REF canonicalizations.
1760 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1763 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
1765 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1767 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
1768 terminated STRING_CST object.
1770 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1772 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
1774 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1776 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
1777 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
1778 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
1779 the elements into a register.
1780 (rs6000_split_v4si_init_di_reg): Delete.
1781 (rs6000_split_v4si_init): Delete.
1782 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
1783 (vsx_init_v4si): Rewrite as a define_expand.
1785 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1787 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
1788 zero_extend argument from memory): New.
1790 2018-07-22 Martin Sebor <msebor@redhat.com>
1793 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
1794 diagnosing calls with unknown arguments unless -Walloca-larger-than
1795 is restricted to less than PTRDIFF_MAX bytes.
1797 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
1799 * doc/gcov.texi (Invoking Gcov): Editorial changes.
1801 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1803 * pretty-print.c (text_info::set_location): Remove redundant
1804 "line_table" parameter from call to rich_location::set_range.
1806 2018-07-20 Martin Sebor <msebor@redhat.com>
1809 * builtins.c (expand_builtin_alloca): Adjust.
1810 * calls.c (alloc_max_size): Simplify.
1811 * cgraphunit.c (cgraph_node::expand): Adjust.
1812 * common.opt (larger_than_size, warn_frame_larger_than): Remove
1814 (frame_larger_than_size): Same.
1815 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
1816 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
1818 * doc/invoke.texi (GCC Command Options): Document option arguments.
1819 Explain byte-size arguments and suffixes.
1820 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
1821 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
1822 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
1823 * doc/options.texi (UInteger): Expand.
1824 (Host_Wide_Int, ByteSize): Document new properties.
1825 * final.c (final_start_function_1): Include sizes in an error message.
1826 * function.c (frame_offset_overflow): Same.
1827 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
1828 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
1829 Diagnose unbounded alloca calls only for limits of less than
1831 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
1832 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
1834 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
1835 only for limits of less than PTRDIFF_MAX.
1836 * langhooks-def.h (lhd_handle_option): Change function argument
1838 * langhooks.c (lhd_handle_option): Same.
1839 * langhooks.h (handle_option): Same.
1840 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
1842 (var_type, var_type_struct): Same.
1843 (var_set): Handle ByteSize flag.
1844 * optc-gen.awk: Add comments to output to ease debugging. Make
1845 use of HOST_WIDE_INT where appropriate.
1846 * opts-gen-save.awk: Use %lx to format unsigned long.
1847 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
1848 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
1849 arguments. Parse bytes-size suffixes.
1850 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
1851 (enum_value_to_arg): Same.
1852 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
1853 (handle_option): Adjust.
1854 (generate_option): Change function argument to HOST_WIDE_INT.
1855 (cmdline_handle_error): Adjust.
1856 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
1857 (set_option): Change function argument to HOST_WIDE_INT.
1858 (option_enabled): Handle cl_host_wide_int.
1859 (get_option_state): Handle CLVC_SIZE.
1860 (control_warning_option): Same.
1861 * opts.c (common_handle_option): Change function argument to
1862 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
1863 OPT_Wvla_larger_than_.
1864 * opts.h (enum cl_var_type): Add an enumerator.
1865 * stor-layout.c (layout_decl): Print a more meaningful warning.
1866 * toplev.c (output_stack_usage): Adjust.
1868 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
1870 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
1871 call to inline_expand_builtin_string_cmp.
1872 (expand_builtin_strcmp): Likewise.
1873 (expand_builtin_strncmp): Likewise.
1874 (inline_string_cmp): Delete the last parameter, change char_type_node
1875 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
1877 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
1878 the inlining expansion on target where the type of the call has same or
1879 narrower precision than unsigned char.
1881 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1883 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
1884 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
1885 * common.opt (fsave-optimization-record): New option.
1886 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
1887 * doc/invoke.texi (-fsave-optimization-record): New option.
1888 * dumpfile.c: Include "optinfo-emit-json.h".
1889 (struct kv_pair): Move to coretypes.h.
1890 (optgroup_options): Make non-static.
1891 (dump_context::end_scope): Call
1892 optimization_records_maybe_pop_dump_scope.
1893 * dumpfile.h (optgroup_options): New decl.
1894 * json.cc: New file.
1896 * optinfo-emit-json.cc: New file.
1897 * optinfo-emit-json.h: New file.
1898 * optinfo.cc: Include "optinfo-emit-json.h".
1899 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
1900 (optinfo_enabled_p): Check optimization_records_enabled_p.
1901 (optinfo_wants_inlining_info_p): Likewise.
1902 * optinfo.h: Update comment.
1903 * profile-count.c (profile_quality_as_string): New function.
1904 * profile-count.h (profile_quality_as_string): New decl.
1905 (profile_count::quality): New accessor.
1906 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
1907 and optinfo_emit_json_cc_tests.
1908 * selftest.h (selftest::json_cc_tests): New decl.
1909 (selftest::optinfo_emit_json_cc_tests): New decl.
1910 * toplev.c: Include "optinfo-emit-json.h".
1911 (compile_file): Call optimization_records_finish.
1912 (do_compile): Call optimization_records_start.
1913 * tree-ssa-live.c: Include optinfo.h.
1914 (remove_unused_scope_block_p): Retain inlining information if
1915 optinfo_wants_inlining_info_p returns true.
1917 2018-07-20 Richard Biener <rguenther@suse.de>
1920 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
1921 to cover -flto-partition=none.
1923 2018-07-20 Martin Liska <mliska@suse.cz>
1925 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
1926 (get_decl_source_range): Remove unused function.
1928 2018-07-20 Richard Biener <rguenther@suse.de>
1930 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
1931 (struct vn_phi_s): Likewise.
1932 (struct vn_reference_s): Likewise.
1933 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
1934 for searching the slot of an entry known to be in the hash itself.
1935 (vn_phi_hasher::equal): Likewise.
1936 (vn_reference_hasher::equal): Likewise.
1937 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
1939 (optimistic_info, current_info): Remove, keeping only valid_info.
1940 (vn_reference_lookup_1): Remove fallback lookup.
1941 (vn_reference_lookup_2): Likewise.
1942 (vn_nary_op_lookup_1): Likewise.
1943 (vn_phi_lookup): Likewise.
1944 (vn_nary_build_or_lookup_1): Make sure to not chain the built
1946 (vn_reference_insert): Adjust, chain the inserted hash element
1947 at last_inserted_ref.
1948 (vn_reference_insert_pieces): Likewise.
1949 (visit_reference_op_call): Likewise.
1950 (vn_nary_op_insert_into): Chain the inserted hash element at
1952 (vn_nary_op_insert_pieces): Adjust.
1953 (vn_nary_op_insert): Likewise.
1954 (vn_nary_op_insert_stmt): Likewise.
1955 (vn_phi_insert): Adjust, chain the inserted hash element at
1957 (process_scc): Remove clearing and copying the optimistic
1958 table. Instead remove elements inserted during an optimistic
1959 iteration from the single table we maintain.
1960 (init_scc_vn): Adjust.
1961 (free_scc_vn): Likewise.
1962 (sccvn_dom_walker::record_cond): Likewise.
1963 (sccvn_dom_walker::after_dom_children): Likewise.
1965 2018-07-19 Martin Sebor <msebor@redhat.com>
1967 PR tree-optimization/84047
1968 PR tree-optimization/83776
1969 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
1970 (check_array_bounds): Call it.
1972 2018-07-19 Martin Sebor <msebor@redhat.com>
1974 * align.h (align_flags): Use member initialization.
1976 2018-07-19 David Malcolm <dmalcolm@redhat.com>
1978 * Makefile.in (OBJS): Add optinfo.o.
1979 * coretypes.h (class symtab_node): New forward decl.
1980 (struct cgraph_node): New forward decl.
1981 (class varpool_node): New forward decl.
1982 * dump-context.h: New file.
1983 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
1985 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
1986 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
1987 (set_alt_dump_file): Likewise.
1988 (dump_context::~dump_context): New dtor.
1989 (dump_gimple_stmt): Move implementation to...
1990 (dump_context::dump_gimple_stmt): ...this new member function.
1991 Add the stmt to any pending optinfo, creating one if need be.
1992 (dump_gimple_stmt_loc): Move implementation to...
1993 (dump_context::dump_gimple_stmt_loc): ...this new member function.
1994 Start a new optinfo and add the stmt to it.
1995 (dump_gimple_expr): Move implementation to...
1996 (dump_context::dump_gimple_expr): ...this new member function.
1997 Add the stmt to any pending optinfo, creating one if need be.
1998 (dump_gimple_expr_loc): Move implementation to...
1999 (dump_context::dump_gimple_expr_loc): ...this new member function.
2000 Start a new optinfo and add the stmt to it.
2001 (dump_generic_expr): Move implementation to...
2002 (dump_context::dump_generic_expr): ...this new member function.
2003 Add the tree to any pending optinfo, creating one if need be.
2004 (dump_generic_expr_loc): Move implementation to...
2005 (dump_context::dump_generic_expr_loc): ...this new member
2006 function. Add the tree to any pending optinfo, creating one if
2008 (dump_printf): Move implementation to...
2009 (dump_context::dump_printf_va): ...this new member function. Add
2010 the text to any pending optinfo, creating one if need be.
2011 (dump_printf_loc): Move implementation to...
2012 (dump_context::dump_printf_loc_va): ...this new member function.
2013 Start a new optinfo and add the stmt to it.
2014 (dump_dec): Move implementation to...
2015 (dump_context::dump_dec): ...this new member function. Add the
2016 value to any pending optinfo, creating one if need be.
2017 (dump_context::dump_symtab_node): New member function.
2018 (dump_context::get_scope_depth): New member function.
2019 (dump_context::begin_scope): New member function.
2020 (dump_context::end_scope): New member function.
2021 (dump_context::ensure_pending_optinfo): New member function.
2022 (dump_context::begin_next_optinfo): New member function.
2023 (dump_context::end_any_optinfo): New member function.
2024 (dump_context::s_current): New global.
2025 (dump_context::s_default): New global.
2026 (dump_scope_depth): Delete global.
2027 (dumpfile_ensure_any_optinfo_are_flushed): New function.
2028 (dump_symtab_node): New function.
2029 (get_dump_scope_depth): Reimplement in terms of dump_context.
2030 (dump_begin_scope): Likewise.
2031 (dump_end_scope): Likewise.
2032 (selftest::temp_dump_context::temp_dump_context): New ctor.
2033 (selftest::temp_dump_context::~temp_dump_context): New dtor.
2034 (selftest::verify_item): New function.
2035 (ASSERT_IS_TEXT): New macro.
2036 (ASSERT_IS_TREE): New macro.
2037 (ASSERT_IS_GIMPLE): New macro.
2038 (selftest::test_capture_of_dump_calls): New test.
2039 (selftest::dumpfile_c_tests): Call it.
2040 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
2041 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
2042 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
2043 descriptive comment.
2044 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
2045 (dump_node, dump_bb): Move these unrelated decls.
2046 (class dump_manager): Add leading comment.
2047 * optinfo.cc: New file.
2048 * optinfo.h: New file.
2050 2018-07-19 Michael Collison <michael.collison@arm.com>
2051 Richard Henderson <rth@redhat.com>
2053 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
2054 (subti): Handle op1 zero.
2055 (subvti4, usub4ti4): New.
2056 (*sub<GPI>3_compare1_imm): New.
2057 (sub<GPI>3_carryinCV): New.
2058 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
2059 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
2061 2018-07-19 Michael Collison <michael.collison@arm.com>
2062 Richard Henderson <rth@redhat.com>
2064 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
2065 (addti3): Create simpler code if low part is already known to be 0.
2066 (addvti4, uaddvti4): New.
2067 (*add<GPI>3_compareC_cconly_imm): New.
2068 (*add<GPI>3_compareC_cconly): New.
2069 (*add<GPI>3_compareC_imm): New.
2070 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
2071 handle constants within this pattern..
2072 (*add<GPI>3_compareV_cconly_imm): New.
2073 (*add<GPI>3_compareV_cconly): New.
2074 (*add<GPI>3_compareV_imm): New.
2075 (add<GPI>3_compareV): New.
2076 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
2077 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
2078 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
2079 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
2080 with 'comparison' operator.
2081 (*add<GPI>3_compareV_cconly_imm): Ditto.
2082 (*add<GPI>3_compareV_cconly): Ditto.
2083 (*add<GPI>3_compareV_imm): Ditto.
2084 (add<GPI>3_compareV): Ditto.
2085 (add<mode>3_carryinC): Ditto.
2086 (*add<mode>3_carryinC_zero): Ditto.
2087 (*add<mode>3_carryinC): Ditto.
2088 (add<mode>3_carryinV): Ditto.
2089 (*add<mode>3_carryinV_zero): Ditto.
2090 (*add<mode>3_carryinV): Ditto.
2092 2018-07-19 Michael Collison <michael.collison@arm.com>
2093 Richard Henderson <rth@redhat.com>
2095 * config/aarch64/aarch64-modes.def (CC_V): New.
2096 * config/aarch64/aarch64-protos.h
2097 (aarch64_addti_scratch_regs): Declare
2098 (aarch64_subvti_scratch_regs): Declare.
2099 (aarch64_expand_subvti): Declare.
2100 (aarch64_gen_unlikely_cbranch): Declare
2101 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
2102 for signed overflow using CC_Vmode.
2103 (aarch64_get_condition_code_1): Handle CC_Vmode.
2104 (aarch64_gen_unlikely_cbranch): New function.
2105 (aarch64_addti_scratch_regs): New function.
2106 (aarch64_subvti_scratch_regs): New function.
2107 (aarch64_expand_subvti): New function.
2109 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2111 * config/aarch64/aarch64-option-extensions.def: New entry for profile
2113 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
2114 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
2117 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2120 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2121 address check not strict.
2123 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
2125 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
2127 (store_pair_lanes<mode>): Likewise.
2128 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
2129 enum value 'ADDR_QUERY_LDP_STP_N'.
2130 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
2131 (aarch64_print_address_internal): Add declaration.
2132 (aarch64_print_ldpstp_address): Remove.
2133 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
2134 (aarch64_print_operand): Change printing of 'y'.
2135 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
2136 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
2137 'true' rather than '1'.
2138 * config/aarch64/constraints.md (Uml): Likewise.
2139 (Uml): Rename to Umn.
2142 2018-07-19 Richard Biener <rguenther@suse.de>
2144 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
2146 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
2147 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
2148 (vn_reference_hasher): Likewise.
2149 (struct vn_tables_s): Remove obstack and alloc-pool members.
2150 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
2151 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
2152 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
2153 (vn_reference_insert_pieces): Likewise.
2154 (alloc_vn_nary_op_noinit): Adjust.
2155 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
2156 (vn_phi_eq): Adjust.
2157 (shared_lookup_phiargs): Remove.
2158 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
2159 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
2160 (visit_reference_op_call): Likewise.
2161 (copy_nary, copy_phi, copy_reference): Remove.
2162 (process_scc): Rewind the obstack when iterating. Do not
2163 copy the elements to valid_info but just move them from one
2164 hashtable to the other.
2165 (allocate_vn_table): Adjust.
2166 (free_vn_table): Likewise.
2167 (init_scc_vn): Likewise.
2168 (free_scc_vn): Likewise.
2170 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
2173 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
2174 indirect_return as function type attribute.
2175 (ix86_attribute_table): Change indirect_return to function
2177 * doc/extend.texi: Update indirect_return attribute.
2179 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
2181 * wide-int.h (widest2_int): New.
2182 * gimple-fold.c (arith_overflowed_p): Use it.
2183 * tree.h (widest2_int_cst): New.
2184 * tree-vrp.c (wide_int_binop_overflow): Rename from
2185 vrp_int_const_binop.
2186 Rewrite to work on trees.
2187 (extract_range_from_multiplicative_op_1): Abstract code to...
2188 (wide_int_range_min_max): ...here.
2189 (wide_int_range_cross_product): ...and here.
2190 (extract_range_from_binary_expr_1): Abstract overflow code to...
2191 (wide_int_range_mult_wrapping): ...here.
2192 * tree-vrp.h (wide_int_range_cross_product): New.
2193 (wide_int_range_mult_wrapping): New.
2195 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
2196 Julia Koval <julia.koval@intel.com>
2198 * config/i386/x86-tune-costs.h (skylake_memcpy,
2199 skylake_memset): Replace rep_prefix with unrolling for size 512.
2201 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
2204 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
2207 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
2209 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
2210 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
2211 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
2212 previously known as "PowerPC AltiVec Built-in Functions". Move
2213 some material to new subsubsections "PowerPC AltiVec Built-in
2214 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
2216 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
2217 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
2218 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
2219 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
2221 2018-07-18 Richard Biener <rguenther@suse.de>
2223 PR tree-optimization/86557
2224 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
2227 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
2229 * config/s390/s390.c (s390_function_profiler): Generate CFI.
2231 2018-07-17 Jeff Law <law@redhat.com>
2233 * config/arm/arm.c (get_label_padding): Update for recent
2234 changes to label_to_alignment.
2236 PR tree-optimization/86010
2237 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
2239 * config/mips/mips.c (vr4130_align_insns): Update for recent
2240 changes to label_to_alignment.
2242 * config/frv/frv.c (frv_label_align): Update for recent changes
2243 to label_to_alignment.
2245 * config/nios2/nios2.c (nios2_label_align): Update for recent
2246 changes which dropped ALIGN_LABELS_LOG.
2248 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
2250 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
2251 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
2253 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
2255 * config/arc/arc.c (arc_label_align): Use align_labels instead of
2256 deprecated align_labels_log.
2258 2018-07-17 Richard Biener <rguenther@suse.de>
2261 * dwarf2out.c (init_sections_and_labels): Always generate
2262 a debug_line_str_section for early LTO debug.
2263 (dwarf2out_finish): Reset debug_line_str_hash output early.
2264 Bump counter for extra dwarf5 .debug_loc labels to not conflict
2266 (dwarf2out_early_finish): Output debug_line_str.
2268 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2270 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
2271 index register on z196 or later.
2273 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
2275 * config/s390/s390.c (s390_default_align): Set default function
2277 (s390_override_options_after_change): Call s390_default align.
2278 (s390_option_override_internal): Call s390_default_align.
2279 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
2281 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2284 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
2285 to _looptemp_ clauses, other than the first two.
2287 2018-07-17 Martin Liska <mliska@suse.cz>
2289 * opts.c: Do not enable OPT_falign_* for -Os.
2291 2018-07-17 Martin Liska <mliska@suse.cz>
2293 * align.h (MAX_CODE_ALIGN): New.
2294 (MAX_CODE_ALIGN_VALUE): New.
2295 * common/config/i386/i386-common.c (ix86_handle_option):
2296 (MAX_CODE_ALIGN): Moved to align.h.
2297 * final.c (MAX_CODE_ALIGN): Likewise.
2298 * opts.c (parse_and_check_align_values):
2299 (MAX_CODE_ALIGN): Likewise.
2300 (MAX_CODE_ALIGN_VALUE): Likewise.
2302 2018-07-17 Martin Liska <mliska@suse.cz>
2304 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
2305 in order to fulfil coding style.
2306 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
2307 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2308 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2309 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
2310 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
2311 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
2312 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
2313 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
2314 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2316 2018-07-17 Martin Liska <mliska@suse.cz>
2318 * align.h: New file.
2319 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
2320 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
2321 align_flags of label_to_alignment.
2322 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
2324 * config/m68k/m68k.c: Do not use removed align_labels_value and
2326 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
2327 (LOOP_ALIGN): Likewise.
2328 (LABEL_ALIGN): Likewise.
2329 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
2330 Remove not used macro.
2331 (rs6000_loop_align): Change return type to align_flags.
2332 (rs6000_loop_align_max_skip): Remove.
2333 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
2334 Change return type to align_flags.
2335 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
2336 Remove not used macro.
2337 (rs6000_loop_align): Change return type to align_flags.
2338 (rs6000_loop_align_max_skip): Remove.
2339 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
2340 * config/rx/rx-protos.h (rx_align_for_label): Make it
2342 * config/rx/rx.c (rx_align_for_label): Change return type
2344 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
2346 into align_flags class.
2347 (LABEL_ALIGN): Likewise.
2348 (LOOP_ALIGN): Likewise.
2349 * config/s390/s390.c (s390_label_align): Use align_flags
2351 (s390_asm_output_function_label): Likewise.
2352 * config/sh/sh.c (sh_override_options_after_change):
2353 Use align_flags class directly without macros.
2354 (find_barrier): Likewise.
2355 (barrier_align): Likewise.
2356 (sh_loop_align): Likewise.
2357 * config/spu/spu.c (spu_option_override):
2358 Use align_flags_tuple::get_value instead of removed macros.
2359 (spu_sched_init): Likewise.
2360 * config/spu/spu.h (GTY): Likewise.
2361 * config/visium/visium.c (visium_option_override):
2362 Set "8" as default secondary alignment.
2363 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
2364 in order to guarantee secondary alignment of 8.
2365 * coretypes.h: Include align.h header file.
2366 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
2367 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
2368 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
2369 * doc/tm.texi.in: Likewise.
2370 * final.c (struct label_alignment): Remove not used structure.
2371 (LABEL_ALIGN): Change type to align_flags.
2372 (LOOP_ALIGN): Likewise.
2373 (JUMP_ALIGN): Likewise.
2374 (default_loop_align_max_skip): Remove.
2375 (default_label_align_max_skip): Likewise.
2376 (default_jump_align_max_skip): Likewise.
2377 (default_label_align_after_barrier_max_skip):
2378 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
2379 (LABEL_TO_MAX_SKIP): Remove.
2380 (label_to_alignment): Return align_flags type instead of integer.
2381 (label_to_max_skip): Remove.
2382 (align_fuzz): Use align_flags type.
2383 (compute_alignments): Use align_flags type and use align_flags::max
2384 to combine multiple alignments.
2385 (grow_label_align): Grow vec instead of C array.
2386 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
2387 (shorten_branches): Use align_flags type and use align_flags::max
2388 to combine multiple alignments.
2389 (final_scan_insn_1): Remove usage of secondary alignment that comes
2390 from label alignment, but instead use proper secondary alignment
2391 which is computed in grow_label_align.
2392 * flags.h (struct align_flags_tuple): Move to align.h.
2393 (struct align_flags): Likewise.
2394 (state_align_loops): Rename to align_loops.
2395 (state_align_jumps): Rename to align_jumps.
2396 (state_align_labels): Rename to align_labels.
2397 (state_align_functions): Rename to align_functions.
2398 (align_loops_log): Remove.
2399 (align_jumps_log): Remove.
2400 (align_labels_log): Remove.
2401 (align_functions_log): Remove.
2402 (align_loops_max_skip): Remove.
2403 (align_jumps_max_skip): Remove.
2404 (align_labels_max_skip): Remove.
2405 (align_functions_max_skip): Remove.
2406 (align_loops_value): Remove.
2407 (align_jumps_value): Remove.
2408 (align_labels_value): Remove.
2409 (align_functions_value): Remove.
2410 * output.h (label_to_alignment): Change return type to align_flags.
2411 (label_to_max_skip): Remove.
2412 * target.def: Remove loop_align_max_skip, label_align_max_skip,
2413 jump_align_max_skip macros.
2414 * targhooks.h (default_loop_align_max_skip): Remove.
2415 (default_label_align_max_skip): Likewise.
2416 (default_jump_align_max_skip): Likewise.
2417 (default_label_align_after_barrier_max_skip): Remove.
2418 * toplev.c (read_log_maxskip): Use ::normalize function.
2419 (parse_N_M): Remove not used argument and also call ::normalize.
2420 (parse_alignment_opts): Do not pass unused arguments.
2421 * varasm.c (assemble_start_function): Use directly align_functions
2422 instead of removed macros.
2423 * system.h: Do not poison removed macros.
2425 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2428 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
2429 and cond temporaries don't have reference type if iterator has
2430 pointer type. For init use &for_pre_body instead of pre_p if
2431 for_pre_body is non-empty.
2433 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2435 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
2436 double-double modes to SFmode directly directly.
2437 (trunc<mode>sf2_fprs): Delete.
2439 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2441 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
2442 for conversions between IFmode and the decimal floating point modes.
2443 (init_float128_ieee): Use the correct names for conversions between
2444 KFmode and the decimal floating point modes.
2446 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2448 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
2449 for the conversions between TDmode and IFmode.
2450 (init_float128_ieee): Use more correct names for the conversions
2451 between TDmode and KFmode.
2453 2018-07-16 Jakub Jelinek <jakub@redhat.com>
2455 PR tree-optimization/86526
2456 * builtins.c (expand_builtin_memcmp): Formatting fixes.
2457 (inline_expand_builtin_string_cmp): Likewise.
2458 (inline_string_cmp): Likewise. Use c_readstr instead of
2459 builtin_memcpy_read_str. Add unit_mode temporary.
2461 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2464 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
2465 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
2467 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2469 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
2470 Alphabetize prototypes of built-in functions, separating out
2471 built-in functions that are listed in this section but should be
2472 described elsewhere.
2474 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
2477 * expmed.c (emit_store_flag): Do not emit setcc followed by a
2478 conditional move when trapping comparison was split to a
2479 non-trapping one (and vice versa).
2481 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2483 * config/s390/s390.c (s390_function_profiler): Generate nops
2484 instead of profiler call sequences.
2485 * config/s390/s390.opt: Add the new option.
2487 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2489 * config/s390/s390.c (s390_function_profiler): Generate
2490 __mcount_loc section.
2491 * config/s390/s390.opt: Add the new option.
2493 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2495 * common.opt: Add the new warning.
2496 * config/s390/s390.c (s390_function_profiler): Emit "brasl
2497 %r0,__fentry__" when -mfentry is specified.
2498 (s390_option_override_internal): Disallow -mfentry for 31-bit
2500 * config/s390/s390.opt: Add the new option.
2502 2018-07-16 Richard Biener <rguenther@suse.de>
2505 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
2506 for function-local FUNCTION_DECL and RESULT_DECL immediately.
2508 2018-07-16 Martin Liska <mliska@suse.cz>
2511 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
2514 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2516 * config/arc/arcHS.md: Update ARCHS scheduling rules.
2518 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2520 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
2522 * config/arc/arc-cpus.def (hs4x): New cpu.
2524 * config/arc/arc-tables.opt: Regenerate.
2525 * config/arc/arc.c (arc_sched_issue_rate): New function.
2526 (TARGET_SCHED_ISSUE_RATE): Define.
2527 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
2528 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
2530 (attr tune): Add ARCHS4x tune values.
2531 (attr tune_dspmpy): Define.
2532 (*tst): Correct instruction type.
2533 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
2534 * config/arc/arcHS4x.md: New file.
2535 * config/arc/fpu.md: Update instruction type attributes.
2536 * config/arc/t-multilib: Regenerate.
2538 2018-07-16 Tom de Vries <tdevries@suse.de>
2541 * var-tracking.c (vt_initialize): Fix pre_dec handling.
2543 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2545 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
2548 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
2550 PR tree-optimization/86514
2551 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
2552 conversion to a boolean type from a type with greater precision.
2554 2018-07-16 Tom de Vries <tdevries@suse.de>
2556 * var-tracking.c (vt_initialize): Print adjusted insn slim if
2557 dump_flags request TDF_SLIM.
2559 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
2561 * fold-const.c (int_const_binop_1): Abstract...
2562 (wide_int_binop): ...wide int code here.
2563 (poly_int_binop): ...poly int code here.
2564 Abstract the rest of int_const_binop_1 into int_const_binop.
2565 * fold-const.h (wide_int_binop): New.
2566 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
2567 Remove useless PLUS/MINUS_EXPR case.
2568 (zero_nonzero_bits_from_vr): Move wide int code...
2569 (zero_nonzero_bits_from_bounds): ...here.
2570 (extract_range_from_binary_expr_1): Move mask optimization code...
2571 (range_easy_mask_min_max): ...here.
2572 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
2573 (range_easy_mask_min_max): New.
2575 2018-07-15 Jeff Law <law@redhat.com>
2578 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
2581 2018-07-14 Jim Wilson <jimw@sifive.com>
2583 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
2585 2018-07-14 Paul Koning <ni1d@arrl.net>
2587 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
2589 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
2591 * lto-streamer-out.c (copy_function_or_variable): Dump info about
2594 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
2595 Steve Munroe <munroesj52@gmail.com>
2597 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
2598 (_mm_andnot_si128): Likewise.
2599 (_mm_or_si128): Likewise.
2600 (_mm_xor_si128): Likewise.
2602 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
2605 * builtins.c (expand_builtin_memcmp): Inline the calls first
2606 when result_eq is false.
2607 (expand_builtin_strcmp): Inline the calls first.
2608 (expand_builtin_strncmp): Likewise.
2609 (inline_string_cmp): New routine. Expand a string compare
2610 call by using a sequence of char comparison.
2611 (inline_expand_builtin_string_cmp): New routine. Inline expansion
2612 a call to str(n)cmp/memcmp.
2613 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
2615 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
2617 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
2619 * config/arm/driver-arm.c: Include arm-native.h.
2620 (host_detect_local_cpu): Use auto-generated data tables.
2621 (vendors, arm_cpu_table): Delete. Move part information to ...
2622 * config/arm/arm-cpus.in: ... here.
2623 * config/arm/parsecpu.awk (gen_native): New function.
2624 (vendor, part): New CPU fields.
2625 (END): Add support for building the native CPU detection tables.
2626 * config/arm/t-arm (arm-native.h): Add build rule.
2627 (driver-arm.o): Add dependency on arm-native.h.
2629 2018-07-13 Richard Biener <rguenther@suse.de>
2632 * match.pd (addr1 - addr2): Allow either of the operand to
2635 2018-07-13 Tom de Vries <tdevries@suse.de>
2637 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
2640 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
2642 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
2643 arrays instead of numbered variables.
2645 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
2647 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
2648 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
2650 2018-07-13 Richard Biener <rguenther@suse.de>
2653 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
2654 instead of get_context_die.
2656 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
2657 Richard Biener <rguenther@suse.de>
2660 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
2661 that the loop latch destination where phi is defined.
2663 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
2665 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
2666 (riscv_expand_epilogue): Add assertion to check interrupt mode.
2667 (riscv_set_current_function): Extract getting interrupt type to new
2669 (riscv_get_interrupt_type): New function.
2670 (riscv_merge_decl_attributes): New function, checking interrupt type is
2672 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
2674 2018-07-12 Paul Koning <ni1d@arrl.net>
2676 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
2679 2018-07-12 Paul Koning <ni1d@arrl.net>
2681 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
2682 zero reference, add doloop_end instead.
2683 * doc/md.texi (decrement_and_branch_until_zero): Remove.
2684 (Looping patterns): Remove decrement_and_branch_until_zero. Add
2685 detail for doloop_end.
2687 2018-07-12 Martin Sebor <msebor@redhat.com>
2690 * attribs.c (decl_attributes): Reject conflicting attributes before
2691 calling attribute handlers.
2693 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2695 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
2697 (gcc::dump_manager::get_dump_file_name): likewise.
2698 (dump_begin): Likewise.
2699 * dumpfile.h (dump_begin): Update prototype.
2700 (gcc::dump_manager::get_dump_file_name,
2701 gcc::dump_manager::get_dump_file_name): Update prototype.
2703 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2705 * internal-fn.h (vectorizable_internal_fn_p): New function.
2706 * tree-vect-slp.c (compatible_calls_p): Likewise.
2707 (vect_build_slp_tree_1): Remove nops argument. Handle calls
2708 to internal functions.
2709 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
2711 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2713 * fold-const.h (inverse_conditions_p): Declare.
2714 * fold-const.c (inverse_conditions_p): New function.
2715 * match.pd: Use inverse_conditions_p. Add folds of view_converts
2716 that test the inverse condition of a conditional internal function.
2717 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
2718 * internal-fn.c (internal_fn_mask_index): Handle conditional
2720 (vectorized_internal_fn_supported_p): New function.
2721 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
2722 (any_pred_load_store): Replace with...
2723 (need_to_predicate): ...this new variable.
2724 (redundant_ssa_names): New variable.
2725 (ifcvt_can_use_mask_load_store): Move initial checks to...
2726 (ifcvt_can_predicate): ...this new function. Handle tree codes
2727 for which a conditional internal function exists.
2728 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
2729 instead of ifcvt_can_use_mask_load_store. Update after variable
2731 (predicate_load_or_store): New function, split out from
2732 predicate_mem_writes.
2733 (check_redundant_cond_expr): New function.
2734 (value_available_p): Likewise.
2735 (predicate_rhs_code): Likewise.
2736 (predicate_mem_writes): Rename to...
2737 (predicate_statements): ...this. Use predicate_load_or_store
2738 and predicate_rhs_code.
2739 (combine_blocks, tree_if_conversion): Update after above name changes.
2740 (ifcvt_local_dce): Handle redundant_ssa_names.
2741 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
2742 general conditional functions.
2743 * tree-vect-stmts.c (vectorizable_call): Likewise.
2745 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2746 Alan Hayward <alan.hayward@arm.com>
2747 David Sherwood <david.sherwood@arm.com>
2749 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
2750 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
2751 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
2752 plus and minus and convert them into IFN_COND_FMA-based sequences.
2753 (convert_mult_to_fma): Handle conditional plus and minus.
2755 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2757 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
2758 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
2759 (cond_fnms_optab): New optabs.
2760 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
2762 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
2763 * internal-fn.h (get_conditional_internal_fn): Declare.
2764 (get_unconditional_internal_fn): Likewise.
2765 * internal-fn.c (cond_ternary_direct): New macro.
2766 (expand_cond_ternary_optab_fn): Likewise.
2767 (direct_cond_ternary_optab_supported_p): Likewise.
2768 (FOR_EACH_COND_FN_PAIR): Likewise.
2769 (get_conditional_internal_fn): New function.
2770 (get_unconditional_internal_fn): Likewise.
2771 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
2772 (gimple_match_op::gimple_match_op): Add a new overload for 5
2774 (gimple_match_op::set_op): Likewise.
2775 (gimple_resimplify5): Declare.
2776 * genmatch.c (decision_tree::gen): Generate simplifications for
2778 * gimple-match-head.c (gimple_simplify): Define an overload for
2779 5 operands. Handle calls with 5 arguments in the top-level overload.
2780 (convert_conditional_op): Handle conversions from unconditional
2781 internal functions to conditional ones.
2782 (gimple_resimplify5): New function.
2783 (build_call_internal): Pass a fifth operand.
2784 (maybe_push_res_to_seq): Likewise.
2785 (try_conditional_simplification): Try converting conditional
2786 internal functions to unconditional internal functions.
2787 Handle 3-operand unconditional forms.
2788 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
2789 Define ternary equivalents of the current rules for binary conditional
2791 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
2793 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
2794 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
2795 (optab): Handle them.
2796 (SVE_COND_FP_TERNARY): New int iterator.
2797 (sve_fmla_op, sve_fmad_op): New int attributes.
2798 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
2799 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
2800 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
2802 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2804 * target.def (preferred_else_value): New target hook.
2805 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
2806 * doc/tm.texi: Regenerate.
2807 * targhooks.h (default_preferred_else_value): Declare.
2808 * targhooks.c (default_preferred_else_value): New function.
2809 * internal-fn.h (conditional_internal_fn_code): Declare.
2810 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
2811 (get_conditional_internal_fn): Use it.
2812 (conditional_internal_fn_code): New function.
2813 * gimple-match.h (gimple_match_cond): New struct.
2814 (gimple_match_op): Add a cond member function.
2815 (gimple_match_op::gimple_match_op): Update all forms to take a
2817 * genmatch.c (expr::gen_transform): Use the same condition as res_op
2818 for the suboperation, but don't specify a particular else_value.
2819 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
2820 (visit_nary_op, visit_reference_op_load): Pass
2821 gimple_match_cond::UNCOND to the gimple_match_op constructor.
2822 * gimple-match-head.c: Include tree-eh.h
2823 (convert_conditional_op): New function.
2824 (maybe_resimplify_conditional_op): Likewise.
2825 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
2826 (gimple_resimplify2): Likewise.
2827 (gimple_resimplify3): Likewise.
2828 (gimple_resimplify4): Likewise.
2829 (maybe_push_res_to_seq): Return null for conditional operations.
2830 (try_conditional_simplification): New function.
2831 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
2833 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
2835 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
2837 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
2839 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2841 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2843 (hash_tree): Do not hash DECL_FCONTEXT
2844 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
2845 Do not stream DECL_FCONTEXT.
2846 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
2847 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
2849 2018-07-12 Richard Biener <rguenther@suse.de>
2852 * dwarf2out.c (gen_block_die): Only output blocks when they have
2853 at least one !DECL_IGNORED_P variable.
2855 2018-07-12 Richard Biener <rguenther@suse.de>
2858 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2859 Remove -mieee-fp handling.
2861 2018-07-12 Richard Biener <rguenther@suse.de>
2863 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
2864 left-over from last patch.
2866 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2868 PR tree-optimization/86492
2869 * gimple-ssa-store-merging.c
2870 (imm_store_chain_info::coalesce_immediate_stores): Call
2871 check_no_overlap even for the merge_overlapping case. Formatting fix.
2873 2018-07-12 Richard Biener <rguenther@suse.de>
2876 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
2877 move possibly trapping operations into the conditional.
2879 2018-07-12 Richard Biener <rguenther@suse.de>
2881 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2882 (vn_lookup_simplify_result): Remove recursion limit applied
2884 (vn_nary_build_or_lookup_1): Adjust.
2885 (try_to_simplify): Likewise.
2886 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2888 (gimple_resimplify2): Likewise.
2889 (gimple_resimplify3): Likewise.
2890 (gimple_resimplify4): Likewise.
2892 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2894 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2895 Use __mmask64 type instead of __mmask8 for __M argument.
2896 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2897 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2899 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2900 __mmask16 for __M argument.
2901 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2902 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2903 to __mmask16 instead of __mmask8.
2904 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2905 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2906 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2907 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2908 instead of __mmask16 for __U argument.
2909 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2910 __mmask16 instead of __mmask8 for __U argument.
2911 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2913 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2915 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2917 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2919 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2921 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2923 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2925 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2927 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2929 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2931 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2932 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
2933 return type as well as __M argument type and all casts from __mmask8
2935 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
2936 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
2937 return type as well as __M argument type and all casts from __mmask8
2939 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
2940 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
2941 return type as well as __M argument type and all casts from __mmask8
2943 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
2944 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
2945 return type as well as __M argument type and all casts from __mmask8
2947 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
2948 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
2951 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
2953 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
2954 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
2957 2018-07-11 Paul Koning <ni1d@arrl.net>
2959 * doc/md.texi (define_subst): Document how multiple occurrences of
2960 the same argument in the replacement pattern are handled.
2962 2018-07-11 Paul Koning <ni1d@arrl.net>
2964 * doc/extend.texi (Common Variable Attributes): Move "mode" into
2966 (Common Type Attributes): Add "mode" attribute.
2968 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
2970 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
2971 stream DECL_ORIGINAL_TYPE.
2972 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
2973 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
2974 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
2975 Do not walk original type.
2976 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
2978 (write_ts_decl_non_common_tree_pointers): Do not stream
2980 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
2981 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
2983 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2985 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
2986 thread twice from the same starting edge.
2988 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2990 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
2991 * gimple.c (gimple_call_nonnull_result_p): ...here...
2992 (gimple_call_nonnull_arg): ...and here.
2993 * gimple.h (gimple_call_nonnull_result_p): New.
2994 (gimple_call_nonnull_arg): New.
2996 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
2998 * config/arm/arm-cpus.in: Move information from fpu field of each
2999 cpu definition to the isa field.
3000 * config/arm/parsecpu.awk (fpu): Delete match rule.
3001 (gen_comm_data): Don't add bits from the CPU's FPU entry.
3003 2018-07-11 Richard Biener <rguenther@suse.de>
3006 * dwarf2out.c (init_sections_and_labels): Use
3007 output_asm_line_debug_info consistently.
3008 (dwarf2out_early_finish): Likewise.
3009 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
3012 2018-07-11 Richard Biener <rguenther@suse.de>
3014 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
3015 Rework father_bb setting in a way to avoid propagating constants
3016 multiple times on a loop body.
3018 2018-07-10 Mark Wielaard <mark@klomp.org>
3021 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
3023 2018-07-10 Richard Biener <rguenther@suse.de>
3025 * hash-map.h (hash_map::iterator::operator*): Return
3026 references to key and value.
3028 2018-07-10 Jakub Jelinek <jakub@redhat.com>
3031 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
3032 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
3033 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
3034 (gimplify_omp_for): For composite loops, move outer
3035 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
3036 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
3037 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
3038 TREE_LIST for both the original class iterator and the "last" helper
3039 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
3040 loop, remember has_decl_expr from outer composite loops for the
3041 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
3043 2018-07-09 Martin Sebor <msebor@redhat.com>
3047 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
3048 accessing implicitly initialized array elements.
3049 * expr.c (string_constant): Handle string initializers of
3050 character arrays within aggregates.
3051 * gimple-fold.c (fold_array_ctor_reference): Add argument.
3052 Store element offset. As a special case, handle zero size.
3053 (fold_nonarray_ctor_reference): Same.
3054 (fold_ctor_reference): Add argument. Store subobject offset.
3055 * gimple-fold.h (fold_ctor_reference): Add argument.
3057 2018-07-09 Paul Koning <ni1d@arrl.net>
3059 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
3060 (pdp11_insn_cost): New function.
3061 (pdp11_md_asm_adjust): New function.
3062 (TARGET_INVALID_WITHIN_DOLOOP): Define.
3063 (pdp11_rtx_costs): Update to match machine better.
3064 (output_addr_const_pdp11): Correct format mismatch warnings.
3065 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
3066 * config/pdp11/pdp11.md: General change to add base_cost and/or
3067 length attributes for use by new pdp11_insn_cost function.
3068 (MIN_BRANCH): Correct definition.
3070 (doloop_end): Use standard pattern name for looping pattern.
3071 (doloop_end_nocc): New.
3072 (movsf): Add another constraint alternative.
3073 (zero_extendqihi2): Add constraint alternatives for not in place
3075 (zero_extendhisi2): Remove.
3076 (shift patterns): Add CC handling variants.
3080 (define_peephole2): New peephole to recognize mov that sets CC for
3083 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3085 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
3086 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
3089 2018-07-09 Jakub Jelinek <jakub@redhat.com>
3092 * real.c (real_nextafter): Return true if result is denormal.
3094 2018-07-09 Martin Liska <mliska@suse.cz>
3096 * common.opt: Add back wrongly removed attribute.
3098 2018-07-09 Richard Biener <rguenther@suse.de>
3101 * dwarf2out.c (gen_block_die): For an early generated DIE
3102 always output high/low PC attributes.
3104 2018-07-09 Tom de Vries <tdevries@suse.de>
3106 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
3107 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
3108 onto VAR_DECL with abstract origin.
3110 2018-07-07 Jim Wilson <jimw@sifive.com>
3112 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
3114 2018-07-07 Tom de Vries <tdevries@suse.de>
3116 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
3118 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
3120 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
3122 (combine_bound): Use wide-int overflow calculation instead of
3124 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
3126 * fold-const.c (int_const_binop_2): Same.
3127 (extract_muldiv_1): Same.
3128 (fold_div_compare): Same.
3129 (fold_abs_const): Same.
3131 * poly-int.h (add): Same.
3135 * predict.c (predict_iv_comparison): Same.
3136 * profile-count.c (slow_safe_scale_64bit): Same.
3137 * simplify-rtx.c (simplify_const_binary_operation): Same.
3138 * tree-chrec.c (tree_fold_binomial): Same.
3139 * tree-data-ref.c (split_constant_offset_1): Same.
3140 * tree-if-conv.c (idx_within_array_bound): Same.
3141 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
3142 * tree-ssa-phiopt.c (minmax_replacement): Same.
3143 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
3144 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
3145 * vr-values.c (vr_values::adjust_range_with_scev): Same.
3146 * wide-int.cc (wi::add_large): Same.
3147 (wi::mul_internal): Same.
3148 (wi::sub_large): Same.
3149 (wi::divmod_internal): Same.
3150 * wide-int.h: Change overflow type to overflow_type for neg, add,
3151 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
3152 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
3153 mul_internal, divmod_internal.
3154 (overflow_type): New enum.
3155 (accumulate_overflow): New.
3157 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3159 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
3160 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
3162 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3164 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
3165 argument is checked for zero before entering loop, avoid checking again.
3167 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
3169 * gimplify.h (generic_expr_could_trap_p): Set as global function.
3170 * gimplify.h (generic_expr_could_trap_p): Likwise.
3171 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
3173 2018-07-06 Jakub Jelinek <jakub@redhat.com>
3175 PR tree-optimization/86401
3176 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
3177 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
3178 (fold_bit_and_mask): ... here. New helper function for match.pd.
3179 * fold-const.h (fold_bit_and_mask): Declare.
3180 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
3182 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
3185 * target.def (translate_mode_attribute): New hook.
3186 * targhooks.h (default_translate_mode_attribute): Declare.
3187 * targhooks.c (default_translate_mode_attribute): New function.
3188 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
3189 * doc/tm.texi: Regenerate.
3190 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
3191 (rs6000_translate_mode_attribute): New function.
3193 2018-07-06 Paul Koning <ni1d@arrl.net>
3195 * doc/md.texi (define_split): Document DONE and FAIL.
3196 (define_peephole2): Ditto.
3198 2018-07-05 Jeff Law <law@redhat.com>
3200 PR tree-optimization/86010
3201 * tree-ssa-dse.c (compute_trims): More aggressively trim at
3202 both the head and tail of mem* and str* calls.
3204 2018-07-05 Jim Wilson <jimw@sifive.com>
3206 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
3208 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
3210 * config/aarch64/aarch64-simd.md: correct flags text for
3211 MIN_EXPR replacement.
3213 2018-07-05 James Clarke <jrtc27@jrtc27.com>
3215 * configure: Regenerated.
3217 2018-07-05 Carl Love <cel@us.ibm.com>
3219 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3220 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3221 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3222 VSX_BUILTIN_DOUBLEL_V4SF.
3224 2018-07-05 Martin Sebor <msebor@redhat.com>
3227 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3228 than its domain to compute its the upper bound of a char array.
3230 2018-07-05 Nathan Sidwell <nathan@acm.org>
3232 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
3233 * doc/cpp.texi: Update comment.
3234 * doc/tm.texi: Rebuilt.
3235 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
3236 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
3237 * doc/extend.texi (Backwards Compatibility): Clarify it is system
3238 headers affected by extern "C".
3239 * system.h: Poison NO_IMPLICIT_EXTERN_C.
3240 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
3241 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
3242 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
3243 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
3244 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
3245 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
3246 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
3247 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
3248 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
3249 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
3250 config/sparc/sp64-elf.h, config/spu/spu.h,
3251 config/stormy16/stormy16.h, config/v850/v850.h,
3252 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
3253 define NO_IMPLICIT_EXTERN_C.
3254 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
3256 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3259 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3260 instead of GET_MODE_SIZE when comparing Units.
3262 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3265 * rtlanal.c (set_noop_p): Constrain on mode change,
3266 include hard-reg-set.h
3268 2018-07-05 Tamar Christina <tamar.christina@arm.com>
3270 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
3272 2018-07-05 Jakub Jelinek <jakub@redhat.com>
3275 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3278 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
3280 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
3282 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
3285 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
3287 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
3289 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
3292 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
3294 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
3296 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
3297 neg_*_op* variables.
3299 2018-07-04 Martin Liska <mliska@suse.cz>
3301 * tree-switch-conversion.c: Define
3302 max_ratio_for_speed and max_ratio_for_size constants.
3304 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
3305 Martin Liska <mliska@suse.cz>
3310 * common.opt: Rename align options with 'str_' prefix.
3311 * common/config/i386/i386-common.c (set_malign_value): New
3313 (ix86_handle_option): Use it to set -falign-* options/
3314 * config/aarch64/aarch64-protos.h (struct tune_params): Change
3315 type from int to string.
3316 * config/aarch64/aarch64.c: Update default values from int
3318 * config/alpha/alpha.c (alpha_override_options_after_change):
3320 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
3321 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3322 max skip conditionally.
3323 * config/i386/freebsd.h (SUBALIGN_LOG): New.
3324 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3325 max skip conditionally.
3326 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3327 max skip conditionally.
3328 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
3329 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3330 max skip conditionally.
3331 * config/i386/i386.c (struct ptt): Change type from int to
3333 (ix86_default_align): Set default values.
3334 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
3335 max skip conditionally.
3336 * config/i386/iamcu.h (SUBALIGN_LOG): New.
3337 (ASM_OUTPUT_MAX_SKIP_ALIGN):
3338 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
3339 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3340 max skip conditionally.
3341 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
3342 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
3343 * config/i386/x86-64.h (SUBALIGN_LOG): New.
3344 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
3345 max skip conditionally.
3346 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3347 * config/ia64/ia64.c (ia64_option_override): Set default values
3348 for alignment options.
3349 * config/m68k/m68k.c: Handle new str_align_* options.
3350 * config/mips/mips.c (mips_set_compression_mode): Change
3352 (mips_option_override): Set default values for options.
3353 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3355 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3357 * config/rx/rx.c (rx_option_override): Likewise.
3358 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
3359 (LABEL_ALIGN): Use align_labels_log.
3360 (LOOP_ALIGN): Use align_loops_align.
3361 * config/s390/s390.c (s390_asm_output_function_label): Use new
3363 * config/sh/sh.c (sh_override_options_after_change):
3364 Change type of constants.
3365 * config/spu/spu.c (spu_sched_init): Likewise.
3366 * config/sparc/sparc.c (sparc_option_override): Set default
3368 * config/visium/visium.c (visium_option_override): Likewise.
3369 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
3370 emit p2align format with last argument if it's not needed.
3371 * doc/invoke.texi: Document extended format of -falign-*.
3372 * final.c: Use align_labels alignment.
3373 * flags.h (struct target_flag_state): Change type to use
3375 (struct align_flags_tuple): New.
3376 (struct align_flags): Likewise.
3377 (align_loops_log): Redefine macro to use new types.
3378 (align_loops_max_skip): Redefine macro to use new types.
3379 (align_jumps_log): Redefine macro to use new types.
3380 (align_jumps_max_skip): Redefine macro to use new types.
3381 (align_labels_log): Redefine macro to use new types.
3382 (align_labels_max_skip): Redefine macro to use new types.
3383 (align_functions_log): Redefine macro to use new types.
3384 (align_loops): Redefine macro to use new types.
3385 (align_jumps): Redefine macro to use new types.
3386 (align_labels): Redefine macro to use new types.
3387 (align_functions): Redefine macro to use new types.
3388 (align_functions_max_skip): Redefine macro to use new types.
3389 (align_loops_value): New macro.
3390 (align_jumps_value): New macro.
3391 (align_labels_value): New macro.
3392 (align_functions_value): New macro.
3393 * function.c (invoke_set_current_function_hook): Propagate
3394 alignment values from flags to global variables default in
3396 * ipa-icf.c (sem_function::equals_wpa): Use
3397 cl_optimization_option_eq instead of memcmp.
3398 * lto-streamer.h (cl_optimization_stream_out): Support streaming
3400 (cl_optimization_stream_in): Likewise.
3401 * optc-save-gen.awk: Support strings in cl_optimization.
3402 * opth-gen.awk: Likewise.
3403 * opts.c (finish_options): Remove error checking of invalid
3405 (MAX_CODE_ALIGN): Remove.
3406 (MAX_CODE_ALIGN_VALUE): Likewise.
3407 (parse_and_check_align_values): New function.
3408 (check_alignment_argument): Likewise.
3409 (common_handle_option): Use check_alignment_argument.
3410 * opts.h (parse_and_check_align_values): Declare.
3411 * toplev.c (init_alignments): Remove.
3412 (read_log_maxskip): New.
3413 (parse_N_M): Likewise.
3414 (parse_alignment_opts): Likewise.
3415 (backend_init_target): Remove usage of init_alignments.
3416 * toplev.h (parse_alignment_opts): Declare.
3417 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
3419 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
3420 * tree.c (cl_option_hasher::equal): New.
3421 * varasm.c: Use new global macros.
3423 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
3425 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
3426 Use a simpler align directive also if MAXSKIP = ALIGN-1.
3427 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3428 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3429 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3430 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3431 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
3432 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
3433 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
3434 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3435 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3436 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3437 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3439 2018-07-04 Martin Liska <mliska@suse.cz>
3440 Jonathan Wakely <jwakely@redhat.com>
3442 * coverage.c: Use correct type.
3443 * doc/invoke.texi: Language correction.
3445 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
3448 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
3449 ENDBRANCH for non-tail call which may return via indirect branch.
3450 * doc/extend.texi: Document indirect_return attribute.
3452 2018-07-03 Martin Sebor <msebor@redhat.com>
3454 PR tree-optimization/86274
3455 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
3457 (format_floating): Correct handling of infinities and NaNs.
3459 2018-07-03 Martin Sebor <msebor@redhat.com>
3461 * print-tree.c (print_real_cst): New function.
3462 (print_node_brief): Call it.
3463 (print_node): Ditto.
3465 2018-07-03 Jeff Law <law@redhat.com>
3467 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
3468 into a single pattern.
3470 * config/h8300/h8300.md (ors code_iterator): New.
3471 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
3472 a single pattern and single splitter.
3473 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
3474 (iorqi3_1, xorqi3_1): Likewise.
3475 (iorqi3, xorqi3 expanders): Similarly.
3477 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
3478 (movmd_internal) into a single pattern using the P mode iterator.
3479 (movmd splitters): Similarly.
3480 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
3481 (movsd splitters): Similarly.
3483 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
3484 ADDB, ADDW and ADDL into a single ADD attribute which selects the
3485 right table based on the size of the operand.
3486 * config/h8300/h8300.md (length_table): Corresponding changes. All
3487 references to "addb", "addw" and "addl" changed to "add".
3488 (btst patterns): Merge two variants into a single pattern.
3489 (tstqi, tsthi): Likewise.
3490 (addhi3_incdec, addsi3_incdec): Likewise.
3491 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
3492 (mulhi3, mulsi3): Likewise.
3493 (udivhi3, udivsi3): Likewise.
3494 (divhi3, divsi3): Likewise.
3495 (andorqi3, andorhi3, andorsi3): Likewise.
3497 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
3500 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
3501 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
3503 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3505 PR tree-optimization/85694
3506 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
3507 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
3509 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
3510 (<u>avg<mode>3_ceil): New patterns.
3512 2018-07-03 David Malcolm <dmalcolm@redhat.com>
3514 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
3515 scan-tree-dump directive.
3516 * gcc.dg/vect/slp-perm-2.c: Likewise.
3517 * gcc.dg/vect/slp-perm-3.c: Likewise.
3518 * gcc.dg/vect/slp-perm-5.c: Likewise.
3519 * gcc.dg/vect/slp-perm-6.c: Likewise.
3520 * gcc.dg/vect/slp-perm-7.c: Likewise.
3521 * gcc.dg/vect/slp-perm-8.c: Likewise.
3523 2018-07-03 Marek Polacek <polacek@redhat.com>
3526 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
3528 2018-07-03 Richard Biener <rguenther@suse.de>
3531 * tree-ssa-structalias.c (find_func_clobbers): Properly
3532 handle indirect calls.
3534 2018-07-03 Jeff Law <law@redhat.com>
3536 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
3537 (shifts): New code iterator.
3538 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
3539 expander. Fix HImode handling on H8/SX.
3540 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
3541 (subqi3, subhi3, subsi3 expanders): Likewise.
3542 (andqi3, andhi3, andsi3 expanders): Likewise.
3543 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
3544 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
3545 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
3546 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
3547 (zero_extendqihi2, zero_extendqisi2): Likewise.
3548 (extendqihi2, extendqisi2): Likewise.
3549 (rotlqi3, rotlhi3, rotlsi3): Likewise.
3550 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
3551 (rotlqi3_1, rotlhi3_1): Likewise.
3552 (logicalhi3_sn, logicalsi3_sn): Likewise.
3553 (logicalhi3, logicalsi3): Likewise.
3555 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3557 * tree-vect-patterns.c (vect_recog_rotate_pattern)
3558 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3559 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
3560 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
3561 type to append_pattern_def_seq instead of creating a stmt_vec_info
3563 (build_mask_conversion): Likewise. Remove vinfo argument.
3564 (vect_add_conversion_to_patterm): Likewise, renaming to...
3565 (vect_add_conversion_to_pattern): ...this.
3566 (vect_recog_mask_conversion_pattern): Update call to
3567 build_mask_conversion. Pass the vector type to
3568 append_pattern_def_seq here too.
3569 (vect_recog_gather_scatter_pattern): Update call to
3570 vect_add_conversion_to_pattern.
3572 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3574 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
3575 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3576 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
3577 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
3578 STMT_VINFO_PATTERN_DEF_SEQ to null here.
3579 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
3580 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
3581 append_pattern_def_seq instead of new_pattern_def_seq.
3582 (vect_recog_divmod_pattern): Do both of the above.
3583 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
3586 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3588 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
3589 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
3590 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
3591 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
3592 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
3593 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3594 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
3595 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3596 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3597 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
3598 parameter with a single stmt_vec_info.
3599 (vect_recog_func_ptr): Likewise.
3600 (vect_recog_gather_scatter_pattern): Likewise, folding in...
3601 (vect_try_gather_scatter_pattern): ...this.
3602 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
3603 the stmt_vec_info of the statement to be matched. Don't clear
3604 STMT_VINFO_RELATED_STMT.
3605 (vect_pattern_recog): Update call accordingly.
3607 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3609 PR tree-optimization/85694
3610 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
3611 (uavgM3_ceil): Document new optabs.
3612 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
3613 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
3615 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
3616 (savg_ceil_optab): New optabs.
3617 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
3618 (vect_vect_recog_func_ptrs): Add it.
3619 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
3620 constant directly from the associated lhs.
3622 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3624 * tree-vect-patterns.c (vect_split_statement): New function.
3625 (vect_convert_input): Use it to try to split an existing cast.
3627 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3629 * poly-int.h (print_hex): New function.
3630 * dumpfile.h (dump_dec, dump_hex): Declare.
3631 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
3632 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
3633 min_input_precision, operation_precision and operation_sign.
3634 * tree-vect-patterns.c (vect_get_range_info): New function.
3635 (vect_same_loop_or_bb_p, vect_single_imm_use)
3636 (vect_operation_fits_smaller_type): Delete.
3637 (vect_look_through_possible_promotion): Add an optional
3638 single_use_p parameter.
3639 (vect_recog_over_widening_pattern): Rewrite to use new
3640 stmt_vec_info infomration. Handle one operation at a time.
3641 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
3642 (vect_truncatable_operation_p, vect_set_operation_type)
3643 (vect_set_min_input_precision): New functions.
3644 (vect_determine_min_output_precision_1): Likewise.
3645 (vect_determine_min_output_precision): Likewise.
3646 (vect_determine_precisions_from_range): Likewise.
3647 (vect_determine_precisions_from_users): Likewise.
3648 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
3649 (vect_vect_recog_func_ptrs): Put over_widening first.
3651 (vect_pattern_recog): Call vect_determine_precisions.
3653 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3655 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
3656 statements that have been replaced by further pattern statements.
3657 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
3659 2018-07-03 Richard Biener <rguenther@suse.de>
3661 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
3662 always set *dt. Dump vectype in vectype overload.
3663 * dumpfile.h (dump_gimple_expr): New function.
3664 (dump_gimple_expr_loc): Likewise.
3665 * dumpfile.c (dump_gimple_expr): New function.
3666 (dump_gimple_expr_loc): Likewise.
3668 2018-07-02 Jeff Law <law@redhat.com>
3670 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
3671 the H8/300, H8/300H and H8/S variants into a single pattern.
3672 (movhi_h8300, movqi_h8300hs): Similarly.
3673 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
3674 (QHI mode iterator): New.
3676 * config/h8300/h8300.md: Remove trailing whitespace.
3678 2018-07-02 Jim Wilson <jimw@sifive.com>
3680 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
3681 instead of emit_insn for interrupt returns.
3682 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
3683 (riscv_sret, riscv_uret): Likewise.
3685 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3687 * pretty-print.c (selftest::test_pp_format): Move save and restore
3688 of quotes to class auto_fix_quotes, and add an instance.
3689 * selftest.c: Include "intl.h".
3690 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
3691 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
3692 * selftest.h (selftest::auto_fix_quotes): New class.
3694 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
3696 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
3697 (aarch64_sve_prepare_conditional_op): Remove.
3698 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
3699 Allow aarch64_simd_reg_or_zero as select operand; remove
3700 the aarch64_sve_prepare_conditional_op call.
3701 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
3702 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3703 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
3704 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
3705 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
3706 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
3707 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
3708 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
3709 and a splitters to match all of the *_any patterns.
3710 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
3712 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
3713 (SVE_COND_FP_BINARY_REV): Remove.
3714 (sve_int_op_rev, sve_fp_op_rev): New.
3715 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
3716 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
3717 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
3718 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
3719 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
3720 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
3721 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
3722 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
3723 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
3725 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
3726 Remove match_dup 1 from the inner unspec.
3727 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3729 * config/aarch64/aarch64.md (movprfx): New attr.
3730 (length): Default movprfx to 8.
3731 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
3732 (*madd<SVE_I>, *msub<SVE_I): Likewise.
3733 (*<su>mul<SVE_I>3_highpart): Likewise.
3734 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
3735 (*v<ASHIFT><SVE_I>3): Likewise.
3736 (*<su><MAXMIN><SVE_I>3): Likewise.
3737 (*<su><MAXMIN><SVE_F>3): Likewise.
3738 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
3739 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
3740 (*div<SVE_F>4): Likewise.
3742 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
3744 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
3747 2018-07-02 Richard Biener <rguenther@suse.de>
3749 PR tree-optimization/86363
3750 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
3751 memset argument refers to a non-variable address.
3753 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3755 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
3756 {PLUS,MINUS}_EXPR code to...
3757 (adjust_symbolic_bound): ...here,
3758 (combine_bound): ...here,
3759 (set_value_range_with_overflow): ...and here.
3761 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3763 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
3765 (extract_range_from_abs_expr): ...here.
3767 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3769 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3770 -fno-omit-frame-pointer when not optimizing.
3772 2018-07-02 Martin Liska <mliska@suse.cz>
3775 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
3776 (propagate_nothrow): Likewise.
3778 2018-07-02 Martin Liska <mliska@suse.cz>
3781 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
3783 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3785 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
3786 function in r262149, changing "loc" param from source_location to
3787 const dump_location_t &.
3788 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
3789 declaration, as above.
3791 2018-07-01 Paul Koning <ni1d@arrl.net>
3793 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3794 -munit-asm, -mgnu-asm, -mdec-asm.
3795 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
3796 (pdp11_output_labelref): New.
3797 (pdp11_output_def): New.
3798 (pdp11_output_addr_vec_elt): New.
3799 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
3800 %# and %@ format codes.
3801 (pdp11_option_override): New.
3802 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
3803 (pdp11_output_ident): New.
3804 (pdp11_asm_named_section): New.
3805 (pdp11_asm_init_sections): New.
3806 (pdp11_file_start): New.
3807 (pdp11_file_end): New.
3808 (output_ascii): Use .ascii/.asciz for -mdec-asm.
3809 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
3810 %o, like %c but octal.
3811 (pdp11_option_override): New.
3812 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
3814 (DATA_SECTION_ASM_OP): Ditto.
3815 (READONLY_DATA_SECTION_ASM_OP): New.
3816 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
3817 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
3818 (ASM_OUTPUT_LABELREF): Ditto.
3819 (ASM_OUTPUT_DEF): Ditto.
3820 (ASM_OUTPUT_EXTERNAL): New.
3821 (ASM_OUTPUT_SOURCE_FILENAME): New.
3822 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
3823 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
3824 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
3825 %# and %@ format codes.
3826 * config/pdp11/pdp11.opt (mgnu-asm): New.
3827 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
3828 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
3829 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
3831 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
3833 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
3834 dereferencing path[] beyond its length.
3836 (debug_all_paths): New.
3837 (rewire_first_differing_edge): New.
3838 (adjust_paths_after_duplication): New.
3839 (duplicate_thread_path): Call adjust_paths_after_duplication.
3841 (thread_through_all_blocks): Add new argument to
3842 duplicate_thread_path.
3844 2018-06-30 Jim Wilson <jimw@sifive.com>
3846 * config/riscv/predicates.md (p2m1_shift_operand): New.
3847 (high_mask_shift_operand): New.
3848 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
3849 pattern using p2m1_shift_operand.
3850 (lshsi3_zero_extend_3+2): New combiner pattern using
3851 high_mask_shift_operand.
3853 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3855 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
3857 (vect_recog_rotate_pattern): ...here.
3858 (vect_convert_input): Try to insert casts of invariants in the
3860 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
3861 preheader to be empty.
3863 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3865 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
3866 vector type. If given, install it in the new statement's
3868 (vect_element_precision): New function.
3869 (vect_unpromoted_value): New struct.
3870 (vect_unpromoted_value::vect_unpromoted_value): New function.
3871 (vect_unpromoted_value::set_op): Likewise.
3872 (vect_look_through_possible_promotion): Likewise.
3873 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
3874 (vect_widened_op_tree, vect_convert_input): Likewise.
3875 (vect_convert_inputs, vect_convert_output): Likewise.
3876 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
3877 to handle the optional cast of the multiplication result and
3878 vect_widened_op_tree to detect the widened multiplication itself.
3879 Do not require the input and output of promotion casts to have
3880 the same sign, but base the signedness of the operation on the
3881 input rather than the result. If the pattern includes two
3882 promotions, check that those promotions have the same sign.
3883 Do not restrict the MULT_EXPR handling to a double-width result;
3884 handle quadruple-width results and wider. Use vect_convert_inputs
3885 to convert the inputs to the common type.
3886 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
3887 to handle the optional cast of the ABS result. Also allow a sign
3888 change or a sign extension between the ABS and MINUS.
3889 Use vect_widened_op_tree to detect the widened subtraction and use
3890 vect_convert_inputs to convert the inputs to the common type.
3891 (vect_handle_widen_op_by_const): Delete.
3892 (vect_recog_widen_op_pattern): New function.
3893 (vect_recog_widen_mult_pattern): Use it.
3894 (vect_recog_widen_shift_pattern): Likewise.
3895 (vect_recog_widen_sum_pattern): Use
3896 vect_look_through_possible_promotion to handle the promoted
3899 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3901 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
3902 the containing gimple_seq *.
3903 * gimple-iterator.h (gsi_for_stmt): Declare it.
3904 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
3905 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
3906 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3907 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3908 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
3910 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
3912 (vect_mark_pattern_stmts): ...here. Handle cases in which the
3913 statement being replaced is part of an existing pattern
3914 definition sequence, inserting the new pattern statements before
3916 (vect_pattern_recog_1): Don't return a bool. If the statement
3917 is already part of a pattern, instead apply pattern matching
3918 to the pattern definition statements. Don't clear the
3919 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
3920 (vect_pattern_recog): Don't break after the first match;
3921 continue processing the pattern definition statements instead.
3922 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
3924 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3926 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3927 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3928 (vect_recog_widen_sum_pattern): Use it.
3930 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3932 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3933 phi is not a pattern statement and has not been replaced by
3934 a pattern statement.
3935 * tree-vect-patterns.c (type_conversion_p): Don't check
3936 STMT_VINFO_IN_PATTERN_P.
3937 (vect_recog_vector_vector_shift_pattern): Likewise.
3938 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
3939 the pattern statement rather than the original statement; check
3940 directly for a WIDEN_MULT_EXPR here.
3941 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
3942 vect_is_simple_use to return the pattern statement rather
3943 than the original statement; use is_pattern_stmt_p to check
3944 for such a pattern statement.
3945 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
3946 to return the pattern statement rather than the original statement;
3947 don't do the same transformation here.
3948 (vect_is_simple_use): If the defining statement has been replaced
3949 by a pattern statement, return the pattern statement instead.
3950 Remove the corresponding (local) transformation from the vectype
3953 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3955 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
3956 end and default to null.
3957 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3958 (vectorizable_reduction): Update calls accordingly, dropping the
3959 gimple ** argument if the passed-back statement isn't needed.
3960 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
3961 (vect_recog_rotate_pattern): Likewise.
3962 (vect_recog_mask_conversion_pattern): Likewise.
3963 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3964 (vect_mask_constant_operand_p): Likewise.
3965 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
3966 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
3967 (get_group_load_store_type, get_load_store_type): Likewise.
3968 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
3969 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
3970 (vectorizable_conversion, vectorizable_assignment): Likewise.
3971 (vectorizable_shift, vectorizable_operation): Likewise.
3972 (vectorizable_store, vect_is_simple_cond): Likewise.
3973 (vectorizable_condition, vectorizable_comparison): Likewise.
3974 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
3975 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
3976 and move it to the end. Cope with null def_stmt_outs.
3978 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3980 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
3982 2018-06-29 Jeff Law <law@redhat.com>
3984 * config/v850/v850.c (v850_legitimate_address_p): Handle large
3985 displacements for TARGET_V850E2V3 and newer.
3986 (TARGET_LRA_P): Remove. Defaults to LRA now.
3987 * config/v850/v850.md (sign23byte_load): Remove.
3988 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
3989 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
3991 2018-06-29 Martin Liska <mliska@suse.cz>
3994 * coverage.c (coverage_init): Mangle full path name.
3995 * doc/invoke.texi: Document the change.
3996 * gcov-io.c (mangle_path): New.
3997 * gcov-io.h (mangle_path): Likewise.
3998 * gcov.c (mangle_name): Use mangle_path for path mangling.
4000 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4002 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
4003 if starting source register is not even.
4005 2018-06-29 Martin Liska <mliska@suse.cz>
4007 PR tree-optimization/86263
4008 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
4009 Make edge redirection.
4011 2018-06-29 David Malcolm <dmalcolm@redhat.com>
4013 * dumpfile.c (dump_loc): Add indentation based on scope depth.
4014 (dump_scope_depth): New variable.
4015 (get_dump_scope_depth): New function.
4016 (dump_begin_scope): New function.
4017 (dump_end_scope): New function.
4018 * dumpfile.h (get_dump_scope_depth): New declaration.
4019 (dump_begin_scope): New declaration.
4020 (dump_end_scope): New declaration.
4021 (class auto_dump_scope): New class.
4022 (AUTO_DUMP_SCOPE): New macro.
4023 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
4026 2018-06-29 Richard Biener <rguenther@suse.de>
4028 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
4029 compute_all_dependences succeeds.
4030 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
4031 exceed --param loop-max-datarefs-for-datadeps.
4033 2018-06-29 Jakub Jelinek <jakub@redhat.com>
4035 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
4037 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
4040 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
4041 alternative 0 in preferred_for_speed attribute.
4043 2018-06-28 Paul Koning <ni1d@arrl.net>
4045 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
4046 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
4047 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
4048 * config/pdp11/pdp11.md: Correct "length" attribute calculation
4049 for shift insn patterns.
4051 2018-06-28 David Malcolm <dmalcolm@redhat.com>
4053 * cgraph.c (cgraph_node::get_body): Replace assignments to
4054 "dump_file" with calls to set_dump_file.
4055 * dumpfile.c (alt_dump_file): Make static, and group with...
4056 (alt_flags): ...this definition.
4057 (dumps_are_enabled): New variable.
4058 (refresh_dumps_are_enabled): New function.
4059 (set_dump_file): New function.
4060 (set_alt_dump_file): New function.
4061 (gcc::dump_manager::dump_start): Replace assignments to
4062 "dump_file" and "alt_dump_file" with calls to set_dump_file and
4064 (gcc::dump_manager::dump_finish): Likewise.
4065 * dumpfile.h (alt_dump_file): Delete decl.
4066 (dumps_are_enabled): New variable decl.
4067 (set_dump_file): New function decl.
4068 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
4070 * tree-nested.c (lower_nested_functions): Replace assignments to
4071 "dump_file" with calls to set_dump_file.
4073 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4075 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4076 goto_locus of each outgoing edge of each basic block.
4078 2018-06-28 Richard Biener <rguenther@suse.de>
4080 * dwarf2out.c (decl_scope_table): Remove.
4081 (push_decl_scope): Likewise.
4082 (pop_decl_scope): Likewise.
4083 (gen_type_die_for_member): Do not call push/pop_decl_scope.
4084 (gen_struct_or_union_type_die): Likewise.
4085 (gen_tagged_type_die): Likewise.
4086 (dwarf2out_init): Do not initialize decl_scope_table.
4087 (dwarf2out_c_finalize): Do not free it.
4089 2018-06-28 Richard Biener <rguenther@suse.de>
4091 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
4092 deciding whether to not re-use a DIE.
4094 2018-06-28 Richard Biener <rguenther@suse.de>
4096 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
4097 DW_AT_abstract_origin attribute.
4099 2018-06-28 Martin Liska <mliska@suse.cz>
4101 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4102 Use newly introduced constants.
4103 * tree-switch-conversion.h (struct jump_table_cluster):
4104 Define max_ratio_for_size and max_ratio_for_speed.
4106 2018-06-28 Martin Liska <mliska@suse.cz>
4108 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4109 Add new checking assert to catch invalid state.
4110 (jump_table_cluster::can_be_handled): Handle single case
4112 (jump_table_cluster::is_beneficial): Bail out for such case.
4113 (bit_test_cluster::find_bit_tests):
4114 Add new checking assert to catch invalid state.
4115 (bit_test_cluster::can_be_handled): Handle single case
4117 (bit_test_cluster::is_beneficial): Bail out for such case.
4118 (switch_decision_tree::analyze_switch_statement):
4121 2018-06-28 Martin Liska <mliska@suse.cz>
4123 * common.opt: Introduce -completion option.
4124 * gcc.c (driver_handle_option): Handle it.
4125 (driver::main): Print completions if completion
4127 * opt-suggestions.c (option_proposer::get_completions):
4129 (option_proposer::suggest_completion): Likewise.
4130 (option_proposer::find_param_completions): Likewise.
4131 (verify_autocompletions): Likewise.
4132 (test_completion_valid_options): Likewise.
4133 (test_completion_valid_params): Likewise.
4134 (in_completion_p): Likewise.
4135 (empty_completion_p): Likewise.
4136 (test_completion_partial_match): Likewise.
4137 (test_completion_garbage): Likewise.
4138 (opt_proposer_c_tests): Likewise.
4139 * opt-suggestions.h: Declare new functions.
4140 * opts.c (common_handle_option): Handle OPT__completion_.
4141 * selftest-run-tests.c (selftest::run_tests): Add
4142 opt_proposer_c_tests.
4143 * selftest.c (assert_str_startswith): New.
4144 * selftest.h (assert_str_startswith): Likewise.
4145 (opt_proposer_c_tests): New.
4146 (ASSERT_STR_STARTSWITH): Likewise.
4148 2018-06-28 Martin Liska <mliska@suse.cz>
4150 * Makefile.in: Add opt-suggestions.o.
4151 * gcc-main.c: Include opt-suggestions.h.
4152 * gcc.c (driver::driver): Likewise.
4153 (driver::~driver): Remove m_option_suggestions.
4154 (driver::build_option_suggestions): Moved to option_proposer.
4155 (driver::suggest_option): Likewise.
4156 (driver::handle_unrecognized_options): Use option_proposer.
4157 * gcc.h (class driver): Add new memver m_option_proposer.
4158 * opt-suggestions.c: New file.
4159 * opt-suggestions.h: New file.
4161 2018-06-28 Martin Liska <mliska@suse.cz>
4163 * vec.h (class auto_string_vec): New (moved from auto_argvec).
4164 (auto_string_vec::~auto_string_vec): Likewise.
4166 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
4168 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
4169 prevent_decl_creation_for_types fields up and add reset_location field.
4170 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
4171 statement if id->reset_location is true.
4172 (copy_edges_for_bb): Do not set goto_locus on the new edges if
4173 id->reset_location is true.
4174 (copy_phis_for_bb): Force input_location on the arguments if
4175 id->reset_location is true.
4176 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
4177 is set on the function to be inlined.
4179 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
4181 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
4183 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
4185 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
4186 registers for Pmode.
4187 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
4188 hard registers for the clobbered pseudo.
4190 2018-06-27 Paul Koning <ni1d@arrl.net>
4192 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
4193 mutually exclusive options.
4194 * config/pdp11/constraints.md (h): New constraint.
4195 (O): Update definition to match shift code generation.
4196 (D): New constraint.
4197 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
4199 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
4201 (output_jump): Change arguments.
4202 (pdp11_fixed_cc_regs): New function.
4203 (pdp11_cc_mode): Ditto.
4204 (pdp11_expand_shift): Ditto.
4205 (pdp11_assemble_shift): Ditto.
4206 (pdp11_small_shift): Ditto.
4207 (pdp11_branch_cost): Remove.
4208 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
4210 (pdp11_register_move_cost): Update for CC registers.
4211 (pdp11_rtx_costs): Add case for LSHIFTRT.
4212 (pdp11_output_jump): Add CCNZ mode conditional branches.
4213 (notice_update_cc_on_set): Remove.
4214 (pdp11_cc_mode): New function.
4215 (simple_memory_operand): Correct pre/post decrement case.
4216 (no_side_effect_operand): New function.
4217 (pdp11_regno_reg_class): Add CC_REGS class.
4218 (pdp11_fixed_cc_regs): New function.
4219 (pdp11_small_shift): New function.
4220 (pdp11_expand_shift): New function to expand shift insns.
4221 (pdp11_assemble_shift): New function to output shifts.
4222 (pdp11_branch_cost): Remove.
4223 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
4224 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
4225 (WCHAR_TYPE): Ditto.
4226 (PTRDIFF_TYPE): Ditto.
4227 (ADJUST_INSN_LENGTH): New macro.
4228 (FIXED_REGISTERS): Add CC registers.
4229 (CALL_USED_REGISTERS): Ditto.
4231 (REG_CLASS_NAMES): Ditto.
4232 (REG_CLASS_CONTENTS): Ditto.
4233 (SELECT_CC_MODE): Use new function.
4234 (TARGET_FLAGS_REGNUM): New macro.
4235 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
4236 (cc0_reg_rtx): Remove.
4237 (CC_STATUS_MDEP): Remove.
4238 (CC_STATUS_MDEFP_INIT): Remove.
4239 (CC_IN_FPU): Remove.
4240 (NOTICE_UPDATE_CC): Remove.
4241 (REGISTER_NAMES): Add CC registers.
4242 (BRANCH_COST): Change to constant 1.
4243 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
4245 * config/pdp11/pdp11.opt (mbcopy): Remove.
4246 (mbcopy-builtin): Remove.
4247 (mbranch-cheap): Remove.
4248 (mbranch-expensive): Remove.
4249 * config/pdp11/predicates.md (expand_shift_operand): Update to
4250 match shift code generation.
4251 (ccnz_operator): New predicate.
4252 * doc/invoke.texi (PDP-11 Options): Remove deleted options
4253 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
4254 Remove non-existent option -mabshi, -mno-abshi. Document mutually
4256 * doc/md.texi (PDP-11): Document new D and h constraints. Update
4257 description of O constraint.
4259 2018-06-27 Jeff Law <law@redhat.com>
4260 Austin Law <austinklaw@gmail.com>
4262 * config/v850/v850.md (addsi3_set_flags): New pattern.
4263 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
4264 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
4265 (zero_extendhisi2_v850_set_flags): Likewise.
4266 (zero_extendqisi2_v850_set_flags): Likewise.
4267 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
4268 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
4269 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
4271 * config/v850/v850-protos.h (notice_update_cc): Remove.
4272 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
4273 (v850_print_operand): Handle 'D' and "d".
4274 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
4275 Add handling of arithmetic/logical operations compared against zero.
4276 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
4277 Do not look at v850_compare_op, instead get mode from last argument.
4278 (v850_gen_compare): Remove
4279 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
4280 after reload for prologue insns.
4281 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
4283 (construct_save_jarl): Likewise.
4284 (TARGET_FLAGS_REGNUM): Define.
4285 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
4286 (NOTICE_UPDATE_CC): Remove.
4287 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
4288 than cc0. Conditionalize on reload_completed.
4289 (cmpsi_insn, setfcc_insn): Likewise.
4290 (tst1 splitter): Turn into define_and_split which sets the flags
4292 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
4293 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
4294 (cstoresf4, cstoredf4): Clobber the flags.
4295 (cmpsi, cmpsf, cmpdf): Remove expanders.
4296 (setf_insn): Remove pattern.
4297 (addsi3): Turn into define_and_split which clobbers the flags after
4298 reload and a suitable pattern (addsi3_clobber_flags) for use after
4300 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
4301 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
4302 (ashrsi3, ashrsi3_v850e2): Likewise.
4303 (bins): Clobber the flags.
4304 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
4305 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
4306 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
4307 (call_value_internal_short, call_value_internal_long): Likewise.
4308 (callt_save_interrupt, callt_return_interrupt): Likewise.
4309 (save_interrupt, return_interrupt): Likewise.
4310 (callt_save_all_interrupt, save_all_interrupt): Likewise.
4311 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
4312 (restore_all_interrupt, _restore_all_interrupt): Likewise.
4313 (All FP comparisons): Only allow after reload has completed.
4315 (divh, divhu): Tweak output template.
4316 (branch_z_normal, branch_z_invert): Remove
4317 (branch_nz_normal, branch_nz_invert): Likewise.
4318 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
4320 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
4321 * config/v850/v850.c (notice_update_cc): Remove.
4322 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
4323 (CC_NO_CARRY): Likewise.
4324 (NOTICE_UPDATE_CC): Define to nothing.
4325 * config/v850/v850.md: Remove block comment on cc0 handling
4326 Remove "cc" attribute from all patterns. Remove cc_status handling
4327 from all patterns. Minor formatting fixes.
4329 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4331 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
4332 (cortex-a76.cortex-a55): Likewise.
4333 * config/aarch64/aarch64-tune.md: Regenerate.
4334 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
4335 cortex-a76.cortex-a55.
4337 2018-06-27 Jeff Law <law@redhat.com>
4339 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
4340 (MULTILIB_DIRNAMES): Similarly.
4342 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
4344 * gimple.h (gimple_return_retbnd): Delete.
4345 (gimple_return_set_retbnd): Likewise.
4346 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
4347 gimple_return_set_retbnd.
4348 * gimple-pretty-print.c (dump_gimple_return): Remove call to
4349 gimple_return_retbnd and adjust.
4350 * tree-inline.h (struct copy_body_data): Remove retbnd field.
4351 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
4352 Explicitly return NULL in a couple more cases. Move assertion
4353 on debug statements and remove unreachable code.
4354 (reset_debug_binding): Do not test id->retbnd.
4355 (expand_call_inline): Do not set it.
4357 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4359 * configure.ac: Add --disable-gcov option.
4360 * configure: Regenerate.
4361 * Makefile.in: Honour @enable_gcov@.
4362 * doc/install.texi: Document --disable-gcov.
4364 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4366 * config/arm/arm-cpus.in (cortex-a76): New entry.
4367 (cortex-a76.cortex-a55): Likewise.
4368 * config/arm/arm-tables.opt: Regenerate.
4369 * config/arm/arm-tune.md: Likewise.
4370 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
4371 * doc/invoke.texi (ARM Options): Document cortex-a76 and
4372 cortex-a76.cortex-a55.
4374 2018-06-27 Tamar Christina <tamar.christina@arm.com>
4377 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
4379 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
4381 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
4383 (EPILOGUE_USES): Likewise.
4385 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
4387 * tree-inline.c (remap_location): New function extracted from...
4388 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
4389 (copy_phis_for_bb): ...here. Call remap_location.
4390 (copy_cfg_body): Adjust call to copy_edges_for_bb.
4392 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
4394 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
4395 unaligned vsx for 16B memset.
4397 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
4400 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
4401 ieee128_float_type_node to long_double_type_node unless
4402 TARGET_LONG_DOUBLE_128 is set.
4404 2018-06-26 David Malcolm <dmalcolm@redhat.com>
4406 * cfgloop.c (get_loop_location): Convert return type from
4407 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
4408 by implicit construction from rtx_insn *, and using
4409 dump_user_location_t::from_function_decl for the fallback case.
4410 * cfgloop.h (get_loop_location): Convert return type from
4411 location_t to dump_user_location_t.
4412 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
4413 dump_printf_loc to pass in a dump_location_t rather than a
4414 location_t, via the gimple stmt.
4415 * coverage.c (get_coverage_counts): Update calls to
4416 dump_printf_loc to pass in dump_location_t rather than a
4418 * doc/optinfo.texi (Dump types): Convert example of
4419 dump_printf_loc from taking "locus" to taking "insn". Update
4420 description of the "_loc" calls to cover dump_location_t.
4421 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
4423 (dump_user_location_t::dump_user_location_t): New constructors,
4424 from gimple *stmt and rtx_insn *.
4425 (dump_user_location_t::from_function_decl): New function.
4426 (dump_loc): Make static.
4427 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
4428 const dump_location_t &.
4429 (dump_generic_expr_loc): Delete.
4430 (dump_printf_loc): Convert param "loc" from location_t to
4431 const dump_location_t &.
4432 (selftest::test_impl_location): New function.
4433 (selftest::dumpfile_c_tests): New function.
4434 * dumpfile.h: Include "profile-count.h".
4435 (class dump_user_location_t): New class.
4436 (struct dump_impl_location_t): New struct.
4437 (class dump_location_t): New class.
4438 (dump_printf_loc): Convert 2nd param from source_location to
4439 const dump_location_t &.
4440 (dump_generic_expr_loc): Delete.
4441 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
4442 const dump_location_t &.
4443 * gimple-fold.c (fold_gimple_assign): Update call to
4444 dump_printf_loc to pass in a dump_location_t rather than a
4445 location_t, via the gimple stmt.
4446 (gimple_fold_call): Likewise.
4447 * gimple-loop-interchange.cc
4448 (loop_cand::analyze_iloop_reduction_var): Update for change to
4449 check_reduction_path.
4450 (tree_loop_interchange::interchange): Update for change to
4452 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
4453 change in return-type of find_loop_location.
4454 (graphite_regenerate_ast_isl): Likewise.
4455 * graphite-optimize-isl.c (optimize_isl): Likewise.
4456 * graphite.c (graphite_transform_loops): Likewise.
4457 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
4458 pass in a dump_location_t rather than a location_t, via the
4460 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4461 * ipa.c (walk_polymorphic_call_targets): Likewise.
4462 * loop-unroll.c (report_unroll): Convert "locus" param from
4463 location_t to dump_location_t.
4464 (decide_unrolling): Update for change to get_loop_location's
4466 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
4467 location_t to dump_user_location_t.
4468 (grid_find_single_omp_among_assignments_1): Updates calls to
4469 dump_printf_loc to pass in a dump_location_t rather than a
4470 location_t, via the gimple stmt.
4471 (grid_parallel_clauses_gridifiable): Convert "tloc" from
4472 location_t to dump_location_t. Updates calls to dump_printf_loc
4473 to pass in a dump_location_t rather than a location_t, via the
4475 (grid_inner_loop_gridifiable_p): Likewise.
4476 (grid_dist_follows_simple_pattern): Likewise.
4477 (grid_gfor_follows_tiling_pattern): Likewise.
4478 (grid_target_follows_gridifiable_pattern): Likewise.
4479 (grid_attempt_target_gridification): Convert initialization
4480 of local "grid" from memset to zero-initialization; FIXME: does
4481 this require C++11? Update call to dump_printf_loc to pass in a
4482 optinfo_location rather than a location_t, via the gimple stmt.
4483 * profile.c (read_profile_edge_counts): Updates call to
4484 dump_printf_loc to pass in a dump_location_t rather than a
4486 (compute_branch_probabilities): Likewise.
4487 * selftest-run-tests.c (selftest::run_tests): Call
4489 * selftest.h (dumpfile_c_tests): New decl.
4490 * tree-loop-distribution.c (pass_loop_distribution::execute):
4491 Update for change in return type of find_loop_location.
4492 * tree-parloops.c (parallelize_loops): Likewise.
4493 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
4494 "locus" from location_t to dump_user_location_t.
4495 (canonicalize_loop_induction_variables): Likewise.
4496 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
4497 for change in return type of find_loop_location.
4498 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
4499 to dump_printf_loc to pass in a dump_location_t rather than a
4500 location_t, via the stmt.
4501 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
4503 * tree-vect-loop-manip.c (find_loop_location): Convert return
4504 type from source_location to dump_user_location_t.
4505 (vect_do_peeling): Update for above change.
4506 (vect_loop_versioning): Update for change in type of
4508 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
4509 from location_t to dump_user_location_t.
4510 (vect_estimate_min_profitable_iters): Update for change in type
4512 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
4513 location_t to dump_location_t.
4514 (vect_slp_bb): Update for change in type of vect_location.
4515 * tree-vectorizer.c (vect_location): Convert from source_location
4516 to dump_user_location_t.
4517 (try_vectorize_loop_1): Update for change in vect_location's type.
4518 (vectorize_loops): Likewise.
4519 (increase_alignment): Likewise.
4520 * tree-vectorizer.h (vect_location): Convert from source_location
4521 to dump_user_location_t.
4522 (find_loop_location): Convert return type from source_location to
4523 dump_user_location_t.
4524 (check_reduction_path): Convert 1st param from location_t to
4525 dump_user_location_t.
4526 * value-prof.c (check_counter): Update call to dump_printf_loc to
4527 pass in a dump_user_location_t rather than a location_t; update
4528 call to error_at for change in type of "locus".
4529 (check_ic_target): Update call to dump_printf_loc to
4530 pass in a dump_user_location_t rather than a location_t, via the
4533 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
4535 * config/s390/s390.h (enum processor_flags): Do not use
4536 default tune parameter when -march was specified.
4538 2018-06-26 Jakub Jelinek <jakub@redhat.com>
4541 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
4542 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
4545 2018-06-26 Richard Biener <rguenther@suse.de>
4547 PR tree-optimization/86287
4549 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
4550 (vect_analyze_loop): Initialize n_stmts.
4552 2018-06-26 Richard Biener <rguenther@suse.de>
4555 * fold-const.c (fold_convertible_p): Pointer extension
4558 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
4561 * dwarf2out.c (loc_list_has_views): Adjust comments.
4562 (dw_loc_list): Split single cross-partition range with
4565 2018-06-25 Jeff Law <law@redhat.com>
4567 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
4568 on -mbig-switch by default.
4570 * config/v850/predicates.md (const_float_1_operand): Fix match_code
4572 (const_float_0_operand): Remove unused predicate.
4573 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
4574 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
4575 (recipsf2): New expander. Original pattern now called
4577 (recipdf2, recipdf2_insn): Similarly.
4578 (rsqrtsf2, rsqrtsf2_insn): Similarly
4579 (rsqrtdf2, rsqrtdf2_insn): Similarly
4581 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
4583 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
4584 Simplify logic for FreeBSD (twice).
4586 2018-06-25 Martin Sebor <msebor@redhat.com>
4588 PR tree-optimization/86204
4589 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
4590 a strnlen result if it's less than the length of the string.
4592 2018-06-25 Martin Sebor <msebor@redhat.com>
4594 PR tree-optimization/85700
4595 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
4596 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
4597 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
4599 2018-06-25 Martin Sebor <msebor@redhat.com>
4601 * doc/extend.texi (Zero-length arrays): Update and clarify.
4603 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
4605 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
4606 added IEEE/IBM long double multilib support on PowerPC little
4607 endian Linux systems.
4608 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
4609 (MULTILIB_DEFAULTS): Likewise.
4610 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4612 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
4613 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
4614 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
4616 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
4619 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
4620 (REORDER_45): Likewise.
4622 2018-06-25 Jeff Law <law@redhat.com>
4624 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
4625 dividend to 32 bits. Adjust length.
4626 (udivmodhi4): Cleanup output template. Fix length.
4628 2018-06-25 Carl Love <cel@us.ibm.com>
4630 * config/rs6000/vsx.md: Change word selector to prefered location.
4632 2018-06-25 Richard Biener <rguenther@suse.de>
4634 PR tree-optimization/86304
4635 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
4636 epilogue-if-converted loops as well.
4638 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
4640 * lto-section-out.c (lto_begin_section): Do not print section
4641 name for noaddr and unnumbered dumps.
4643 2018-06-25 Richard Biener <rguenther@suse.de>
4645 * tree-vectorizer.h (struct vec_info_shared): New structure
4646 with parts split out from struct vec_info and loop_nest from
4647 struct _loop_vec_info.
4648 (struct vec_info): Adjust accordingly.
4649 (struct _loop_vec_info): Likewise.
4650 (LOOP_VINFO_LOOP_NEST): Adjust.
4651 (LOOP_VINFO_DATAREFS): Likewise.
4652 (LOOP_VINFO_DDRS): Likewise.
4653 (struct _bb_vec_info): Likewise.
4654 (BB_VINFO_DATAREFS): Likewise.
4655 (BB_VINFO_DDRS): Likewise.
4656 (struct _stmt_vec_info): Add dr_aux member.
4657 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
4658 (DR_MISALIGNMENT_UNINITIALIZED): New.
4659 (set_dr_misalignment): Adjust.
4660 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
4661 (vect_analyze_loop): Adjust prototype.
4662 (vect_analyze_loop_form): Likewise.
4663 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4664 Compute dependences lazily.
4665 (vect_record_base_alignments): Use shared datarefs/ddrs.
4666 (vect_verify_datarefs_alignment): Likewise.
4667 (vect_analyze_data_refs_alignment): Likewise.
4668 (vect_analyze_data_ref_accesses): Likewise.
4669 (vect_analyze_data_refs): Likewise.
4670 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
4671 constructor parameter for shared part.
4672 (vect_analyze_loop_form): Pass in shared part and adjust.
4673 (vect_analyze_loop_2): Pass in storage for the number of
4674 stmts. Move loop nest finding to the caller. Compute
4676 (vect_analyze_loop): Pass in shared part.
4677 (vect_transform_loop): Verify shared datarefs are unchanged.
4678 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
4679 constructor parameter for shared part.
4680 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
4681 (vect_slp_bb): Verify shared datarefs are unchanged before
4683 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
4685 (new_stmt_vec_info): Initialize DR_AUX misalignment to
4686 DR_MISALIGNMENT_UNINITIALIZED.
4687 * tree-vectorizer.c (vec_info::vec_info): Add constructor
4688 parameter for shared part.
4689 (vec_info::~vec_info): Adjust.
4690 (vec_info_shared::vec_info_shared): New.
4691 (vec_info_shared::~vec_info_shared): Likewise.
4692 (vec_info_shared::save_datarefs): Likewise.
4693 (vec_info_shared::check_datarefs): Likewise.
4694 (try_vectorize_loop_1): Construct shared part live for analyses
4695 of a single loop for multiple vector sizes.
4696 * tree-parloops.c (gather_scalar_reductions): Adjust.
4698 2018-06-25 Richard Biener <rguenther@suse.de>
4700 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
4701 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
4702 (vect_analyze_data_refs): Remove similar code from here and
4703 simplify accordingly.
4705 2018-06-25 Richard Biener <rguenther@suse.de>
4707 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
4708 for reverse storage order accesses rather than asserting
4709 they cannot happen here.
4711 2018-06-25 Tom de Vries <tdevries@suse.de>
4714 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
4715 Use data16 instead of .byte for insn prefix.
4717 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
4720 * parser.c (make_char_string_pack): Pass this literal chars
4721 through cpp_interpret_string.
4722 (cp_parser_userdef_numeric_literal): Check the result of
4723 make_char_string_pack.
4725 2018-06-24 Maya Rashish <coypu@sdf.org>
4727 * ginclude/stddef.h: Simplify conditions around avoiding
4728 re-definition of __size_t.
4730 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4732 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
4733 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
4735 2018-06-22 Maya Rashish <coypu@sdf.org>
4737 * doc/invoke.texi (mno-fancy-math-387): Update for changes
4738 made to OpenBSD and NetBSD through the years.
4740 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4742 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4743 behavior of vec_pack (vector double, vector double) to match
4744 behavior of vec_float2 (vector double, vector double).
4746 2018-06-22 Olivier Hainque <hainque@adacore.com>
4748 * gimplify.c (gimplify_function_tree): Prevent creation
4749 of a trampoline for the address of the current function
4750 passed to entry/exit instrumentation hooks.
4752 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
4755 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
4758 2018-06-22 Martin Liska <mliska@suse.cz>
4760 PR tree-optimization/86263
4761 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4762 Bail out if is_enabled is false.
4763 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
4765 (jump_table_cluster::is_enabled): New function.
4767 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4769 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4770 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
4771 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
4772 (lto_input_ts_binfo_tree_pointers): Likewise.
4773 * tree-streamer-out.c (streamer_write_tree_bitfields,
4774 write_ts_binfo_tree_pointers): Likewise.
4775 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
4777 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4779 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
4781 2018-06-22 Martin Liska <mliska@suse.cz>
4783 * symbol-summary.h (get): Make it pure and inline move
4784 functionality from ::get function.
4785 (get): Remove and inline into ::get and ::get_create.
4786 (get_create): Move code from ::get function.
4788 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4791 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
4792 -x assembler-with-cpp.
4794 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4796 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
4797 _FILE_OFFSET_BITS=64 for C++.
4799 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
4801 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
4802 conversion insn that shows up when pr85657-3.c is compiled using
4803 IEEE 128-bit long double.
4804 (neg<mode>2_internal): Use the correct mode to check whether the
4805 mode is IBM extended.
4806 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
4807 multiply and divide external functions from being created more
4810 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4812 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
4814 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
4815 the edge can be forwarded.
4816 (cfg_layout_merge_blocks): Likewise.
4818 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4820 * except.c (finish_eh_generation): Commit edge insertions only after
4821 the EH edges have been redirected from post-landing to landing pads.
4823 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4825 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
4826 create_tmp_var_for to create the FRAME decl.
4827 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
4829 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4831 * tree-inline.c (copy_edges_for_bb): Minor tweak.
4832 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
4833 debug statement when resetting its value.
4834 (expand_call_inline): Copy the locus of the call onto the assignment
4835 of the return value, if any. Use local variable in more cases.
4837 2018-06-21 Martin Liska <mliska@suse.cz>
4839 * ipa-pure-const.c (propagate_nothrow): Use
4840 funct_state_summaries->get.
4841 (dump_malloc_lattice): Likewise.
4842 (propagate_malloc): Likewise.
4844 2018-06-21 Richard Biener <rguenther@suse.de>
4846 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
4847 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
4848 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
4850 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
4851 BLOCK_ABSTRACT_ORIGIN unconditionally.
4853 2018-06-21 David Malcolm <dmalcolm@redhat.com>
4855 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
4857 * ipa-reference.c (ipa_reference_c_finalize): Delete
4858 ipa_ref_opt_sum_summaries and set it to NULL.
4860 2018-06-21 Tom de Vries <tdevries@suse.de>
4862 PR tree-optimization/85859
4863 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
4864 test with comment from bb_no_side_effects_p.
4866 2018-06-21 Richard Biener <rguenther@suse.de>
4868 PR tree-optimization/86232
4869 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
4870 max for constant niter.
4872 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4874 * config/aarch64/aarch64-simd.md
4875 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
4877 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4879 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
4880 Make opernads of the unspec commutative.
4882 2018-06-21 Richard Biener <rguenther@suse.de>
4884 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4885 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
4886 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
4887 (vect_analyze_data_ref_dependence): Re-order checks to deal with
4889 (vect_record_base_alignments): Do not record base alignment
4890 for gathers or scatters.
4891 (vect_compute_data_ref_alignment): Drop return value that is always
4892 true. Bail out early for gathers or scatters.
4893 (vect_enhance_data_refs_alignment): Bail out early for gathers
4895 (vect_find_same_alignment_drs): Likewise.
4896 (vect_analyze_data_refs_alignment): Remove dead code.
4897 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4898 (vect_analyze_data_refs): For possible gathers or scatters do
4899 not create an alternate DR, just check their possible validity
4900 and mark them. Adjust DECL_NONALIASED handling to not rely
4902 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4903 update inits of gathers or scatters.
4904 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4905 Also copy gather/scatter flag to pattern vinfo.
4907 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
4909 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4910 behavior of vec_packsu (vector unsigned long long, vector unsigned
4911 long long) to match behavior of vec_packs with same signature.
4913 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
4914 Thomas Schwinge <thomas@codesourcery.com>
4915 Cesar Philippidis <cesar@codesourcery.com>
4917 * gimplify.c (gimplify_scan_omp_clauses): Add support for
4918 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4919 (gimplify_adjust_omp_clauses): Likewise.
4920 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
4921 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
4922 (gimplify_omp_target_update): Update handling of acc update and
4924 * omp-low.c (install_var_field): Remove unused parameter
4925 base_pointers_restrict.
4926 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
4927 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
4929 (omp_target_base_pointers_restrict_p): Delete.
4930 (scan_omp_target): Update call to scan_sharing_clauses.
4931 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
4933 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4934 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4935 (convert_local_omp_clauses): Likewise.
4936 * tree-pretty-print.c (dump_omp_clause): Likewise.
4937 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
4939 (omp_clause_code_name): Likewise.
4941 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4944 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4947 PR tree-optimization/86231
4948 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4949 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4951 2018-06-20 Tom de Vries <tdevries@suse.de>
4953 PR tree-optimization/86097
4954 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
4955 iv type if signedness of iv type is not the same as that of *nit.
4957 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4959 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
4960 EDGE_EH edges, verify they are all EDGE_EH.
4962 2018-06-20 Maya Rashish <coypu@sdf.org>
4964 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
4966 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4968 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
4969 * config/aarch64/aarch64.c (xgene1_tunings): Add
4970 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
4971 (aarch64_mode_valid_for_sched_fusion_p):
4972 Allow 16-byte modes.
4973 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
4974 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
4976 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
4978 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
4979 * config/aarch64/iterators.md (VQ2): New mode iterator.
4981 2018-06-20 Martin Liska <mliska@suse.cz>
4983 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4984 Change default ratio from 10 to 8.
4986 2018-06-20 Martin Liska <mliska@suse.cz>
4988 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4990 (bit_test_cluster::find_bit_tests): Likewise.
4991 (switch_decision_tree::analyze_switch_statement): Find clusters.
4992 * tree-switch-conversion.h (struct jump_table_cluster): Document
4995 2018-06-20 Martin Liska <mliska@suse.cz>
4997 * tree-switch-conversion.c (switch_conversion::collect):
4998 Record m_uniq property.
4999 (switch_conversion::expand): Bail out for special conditions.
5000 (group_cluster::~group_cluster): New.
5001 (group_cluster::group_cluster): Likewise.
5002 (group_cluster::dump): Likewise.
5003 (jump_table_cluster::emit): New.
5004 (switch_decision_tree::fix_phi_operands_for_edges): New.
5005 (struct case_node): Remove struct.
5006 (jump_table_cluster::can_be_handled): New.
5007 (case_values_threshold): Moved to header.
5008 (reset_out_edges_aux): Likewise.
5009 (jump_table_cluster::is_beneficial): New.
5010 (bit_test_cluster::can_be_handled): Likewise.
5011 (add_case_node): Remove.
5012 (bit_test_cluster::is_beneficial): New.
5013 (case_bit_test::cmp): New.
5014 (bit_test_cluster::emit): New.
5015 (expand_switch_as_decision_tree_p): Remove.
5016 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
5017 (fix_phi_operands_for_edge): Likewise.
5018 (switch_decision_tree::analyze_switch_statement): New.
5019 (compute_cases_per_edge): Move ...
5020 (switch_decision_tree::compute_cases_per_edge): ... here.
5021 (try_switch_expansion): Likewise.
5022 (switch_decision_tree::try_switch_expansion): Likewise.
5023 (record_phi_operand_mapping): Likewise.
5024 (switch_decision_tree::record_phi_operand_mapping): Likewise.
5025 (emit_case_decision_tree): Likewise.
5026 (switch_decision_tree::emit): Likewise.
5027 (balance_case_nodes): Likewise.
5028 (switch_decision_tree::balance_case_nodes): Likewise.
5029 (dump_case_nodes): Likewise.
5030 (switch_decision_tree::dump_case_nodes): Likewise.
5031 (emit_jump): Likewise.
5032 (switch_decision_tree::emit_jump): Likewise.
5033 (emit_cmp_and_jump_insns): Likewise.
5034 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
5035 (emit_case_nodes): Likewise.
5036 (switch_decision_tree::emit_case_nodes): Likewise.
5037 (conditional_probability): Remove.
5038 * tree-switch-conversion.h (enum cluster_type): New.
5040 (struct cluster): Likewise.
5041 (cluster::cluster): Likewise.
5042 (struct simple_cluster): Likewise.
5043 (simple_cluster::simple_cluster): Likewise.
5044 (struct group_cluster): Likewise.
5045 (struct jump_table_cluster): Likewise.
5046 (struct bit_test_cluster): Likewise.
5047 (struct min_cluster_item): Likewise.
5048 (struct case_tree_node): Likewise.
5049 (case_tree_node::case_tree_node): Likewise.
5050 (jump_table_cluster::case_values_threshold): Likewise.
5051 (struct case_bit_test): Likewise.
5052 (struct switch_decision_tree): Likewise.
5053 (struct switch_conversion): Likewise.
5054 (switch_decision_tree::reset_out_edges_aux): Likewise.
5056 2018-06-20 Martin Liska <mliska@suse.cz>
5058 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
5059 (hoist_edge_and_branch_if_true): Likewise.
5060 (expand_switch_using_bit_tests_p): Likewise.
5061 (struct case_bit_test): Likewise.
5062 (case_bit_test_cmp): Likewise.
5063 (emit_case_bit_tests): Likewise.
5064 (switch_conversion::switch_conversion): New class.
5065 (struct switch_conv_info): Remove old struct.
5066 (collect_switch_conv_info): More to ...
5067 (switch_conversion::collect): ... this.
5068 (check_range): Likewise.
5069 (switch_conversion::check_range): Likewise.
5070 (check_all_empty_except_final): Likewise.
5071 (switch_conversion::check_all_empty_except_final): Likewise.
5072 (check_final_bb): Likewise.
5073 (switch_conversion::check_final_bb): Likewise.
5074 (create_temp_arrays): Likewise.
5075 (switch_conversion::create_temp_arrays): Likewise.
5076 (free_temp_arrays): Likewise.
5077 (gather_default_values): Likewise.
5078 (switch_conversion::gather_default_values): Likewise.
5079 (build_constructors): Likewise.
5080 (switch_conversion::build_constructors): Likewise.
5081 (constructor_contains_same_values_p): Likewise.
5082 (switch_conversion::contains_same_values_p): Likewise.
5083 (array_value_type): Likewise.
5084 (switch_conversion::array_value_type): Likewise.
5085 (build_one_array): Likewise.
5086 (switch_conversion::build_one_array): Likewise.
5087 (build_arrays): Likewise.
5088 (switch_conversion::build_arrays): Likewise.
5089 (gen_def_assigns): Likewise.
5090 (switch_conversion::gen_def_assigns): Likewise.
5091 (prune_bbs): Likewise.
5092 (switch_conversion::prune_bbs): Likewise.
5093 (fix_phi_nodes): Likewise.
5094 (switch_conversion::fix_phi_nodes): Likewise.
5095 (gen_inbound_check): Likewise.
5096 (switch_conversion::gen_inbound_check): Likewise.
5097 (process_switch): Use the newly created class.
5098 (switch_conversion::expand): New.
5099 (switch_conversion::~switch_conversion): New.
5100 * tree-switch-conversion.h: New file.
5102 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5104 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
5105 tree-vect-patterns.c.
5106 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
5107 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
5108 (vect_recog_sad_pattern): Likewise.
5109 (vect_recog_widen_sum_pattern): Likewise.
5110 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
5111 (vect_recog_widen_shift_pattern): Remove the type_in argument.
5112 (vect_recog_rotate_pattern): Likewise.
5113 (vect_recog_mult_pattern): Likewise.
5114 (vect_recog_vector_vector_shift_pattern): Likewise.
5115 (vect_recog_divmod_pattern): Likewise.
5116 (vect_recog_mixed_size_cond_pattern): Likewise.
5117 (vect_recog_bool_pattern): Likewise.
5118 (vect_recog_mask_conversion_pattern): Likewise.
5119 (vect_try_gather_scatter_pattern): Likewise.
5120 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
5121 (vect_recog_over_widening_pattern): Likewise.
5122 (vect_recog_gather_scatter_pattern): Likewise.
5123 (vect_recog_func_ptr): Move from tree-vectorizer.h
5124 (vect_vect_recog_func_ptrs): Move further down the file.
5125 (vect_recog_func): Likewise. Remove the third argument.
5126 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
5127 (vect_pattern_recog_1): Expect the pattern function to do any
5128 necessary target tests. Also expect it to provide a vector type.
5129 Remove the type_in handling.
5131 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5133 * tree-vect-patterns.c (vect_pattern_detected): New function.
5134 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
5135 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
5136 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
5137 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
5138 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5139 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5140 (vect_recog_mask_conversion_pattern)
5141 (vect_try_gather_scatter_pattern): Likewise.
5143 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5145 * tree-vect-patterns.c (vect_get_internal_def): New function.
5146 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5147 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
5148 (search_type_for_mask_1): Use it.
5150 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5152 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5153 redundant WIDEN_SUM_EXPR handling.
5154 (vect_recog_sad_pattern): Likewise.
5156 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5158 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
5159 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
5160 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
5161 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
5162 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
5164 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5166 * tree-vect-stmts.c (vectorizable_call): Make sure that we
5167 use the stmt_vec_info of the original bb statement for the
5168 new zero assignment, even if the call is part of a pattern.
5170 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5172 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
5173 that the sequence is attached to the original statement rather
5174 than the pattern statement.
5175 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
5176 PATTERN_DEF_SEQ from the original statement rather than
5177 the main pattern statement.
5178 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
5179 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
5180 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
5182 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
5184 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
5185 definition statements before the early exit for statements that aren't
5187 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
5189 (vect_transform_loop): ...here. Process pattern definition
5190 statements without first checking whether the main pattern
5191 statement is live or relevant.
5193 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
5195 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
5196 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
5198 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
5200 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
5201 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
5202 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
5203 (expand_block_compare): Change select_block_compare_mode call.
5204 (expand_strncmp_align_check): Use new functions, fix comment.
5205 (emit_final_str_compare_gpr): New function.
5206 (expand_strn_compare): Refactor and clean up code.
5207 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
5209 2018-06-19 Tony Reix <tony.reix@atos.com>
5210 Damien Bergamini <damien.bergamini@atos.com>
5211 David Edelsohn <dje.gcc@gmail.com>
5213 * collect2.c (static_obj): New variable.
5214 (static_libs): New variable.
5215 (is_in_list): Uncomment declaration.
5216 (main): Track AIX libraries linked statically.
5217 (is_in_list): Uncomment definition.
5218 (scan_prog_file): Don't add AIX shared libraries initializer
5219 to constructor list if linking statically.
5221 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
5223 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
5225 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
5227 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
5229 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
5232 2018-06-19 Martin Liska <mliska@suse.cz>
5234 * config/i386/i386.c (ix86_can_inline_p): Do not use
5235 ipa_fn_summaries::get_create.
5236 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
5238 (devirtualization_time_bonus): Likewise.
5239 (ipcp_propagate_stage): Likewise.
5240 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5241 (edge_set_predicate): Likewise.
5242 (evaluate_conditions_for_known_args): Likewise.
5243 (evaluate_properties_for_edge): Likewise.
5244 (ipa_call_summary::reset): Tranform to ...
5245 (ipa_call_summary::~ipa_call_summary): ... this.
5246 (ipa_fn_summary::reset): Transform to ...
5247 (ipa_fn_summary::~ipa_fn_summary): ... this.
5248 (ipa_fn_summary_t::remove): Rename to ...
5249 (ipa_fn_summary_t::remove_callees): ... this.
5250 (ipa_fn_summary_t::duplicate): Use placement new
5251 instead of memory copy.
5252 (ipa_call_summary_t::duplicate): Likewise.
5253 (ipa_call_summary_t::remove): Remove.
5254 (dump_ipa_call_summary): Change get_create to get.
5255 (ipa_dump_fn_summary): Dump only when summary exists.
5256 (analyze_function_body): Use symbol_summary::get instead
5258 (compute_fn_summary): Likewise.
5259 (estimate_edge_devirt_benefit): Likewise.
5260 (estimate_edge_size_and_time): Likewise.
5261 (inline_update_callee_summaries): Likewise.
5262 (remap_edge_change_prob): Likewise.
5263 (remap_edge_summaries): Likewise.
5264 (ipa_merge_fn_summary_after_inlining): Likewise.
5265 (write_ipa_call_summary): Likewise.
5266 (ipa_fn_summary_write): Likewise.
5267 (ipa_free_fn_summary): Likewise.
5268 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
5269 (struct ipa_call_summary): Likewise.
5270 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
5272 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
5273 (estimate_size_after_inlining): Likewise.
5274 (estimate_growth): Likewise.
5275 (growth_likely_positive): Likewise.
5276 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5277 (inline_call): Likewise.
5278 * ipa-inline.c (caller_growth_limits): Likewise.
5279 (can_inline_edge_p): Likewise.
5280 (can_inline_edge_by_limits_p): Likewise.
5281 (compute_uninlined_call_time): Likewise.
5282 (compute_inlined_call_time): Likewise.
5283 (want_inline_small_function_p): Likewise.
5284 (edge_badness): Likewise.
5285 (update_caller_keys): Likewise.
5286 (update_callee_keys): Likewise.
5287 (inline_small_functions): Likewise.
5288 (inline_to_all_callers_1): Likewise.
5289 (dump_overall_stats): Likewise.
5290 (early_inline_small_functions): Likewise.
5291 (early_inliner): Likewise.
5292 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5293 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5294 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5295 * ipa-split.c (execute_split_functions): Likewise.
5296 * symbol-summary.h: Likewise.
5297 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5299 2018-06-19 Richard Biener <rguenther@suse.de>
5301 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
5302 (vectorize_loops): ... here. Fix dbgcnt handling.
5303 (try_vectorize_loop): Wrap try_vectorize_loop_1.
5305 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
5308 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
5309 ieee128 argument takes up only one (vector) register, not two (floating
5312 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
5314 * gimplify.c (gimplify_init_constructor): Really never clear for an
5315 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
5317 2018-06-19 Richard Biener <rguenther@suse.de>
5319 PR tree-optimization/86179
5320 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
5321 after failed recognition.
5323 2018-06-18 Martin Sebor <msebor@redhat.com>
5326 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
5327 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
5328 Handle integer subtraction.
5329 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
5330 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
5332 2018-06-18 David Malcolm <dmalcolm@redhat.com>
5334 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
5335 param from rtx to rtx_insn *.
5336 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
5338 (frv_ifcvt_modify_insn): Likwise.
5339 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
5340 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
5341 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
5342 as_a <rtx_insn *> cast to local "unprotected_region" once
5343 it's been established that it's not NULL or pc_rtx.
5344 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
5345 param "sethi" from rtx to rtx_insn *.
5346 (nds32_group_float_insns): Likewise for param "insn".
5347 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
5349 (vax_output_int_subtract): Likewise.
5350 * config/vax/vax.c (vax_output_int_add): Likewise for param
5352 (vax_output_int_subtract): Likewise.
5353 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
5354 (emit_pattern_after): Likewise for param "after".
5355 (emit_insn_after): Likewise.
5356 (emit_jump_insn_after): Likewise.
5357 (emit_call_insn_after): Likewise.
5358 (emit_debug_insn_after): Likewise.
5359 (emit_pattern_before): Likewise for param "before".
5360 (emit_insn_before): Likewise.
5361 (emit_jump_insn_before): Likewise.
5362 * final.c (get_insn_template): Likewise for param "insn", removing
5364 * output.h (get_insn_template): Likewise for 2nd param.
5365 * rtl.h (emit_insn_before): Likewise.
5366 (emit_jump_insn_before): Likewise.
5367 (emit_debug_insn_before_noloc): Likewise.
5368 (emit_insn_after): Likewise.
5369 (emit_jump_insn_after): Likewise.
5370 (emit_call_insn_after): Likewise.
5371 (emit_debug_insn_after): Likewise.
5372 (set_insn_deleted): Likewise for param.
5374 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
5377 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
5378 floating point modes, so that IFmode is numerically greater than
5379 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
5380 to declare the ordering. This prevents IFmode from being
5381 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
5382 machine. Include rs6000-modes.h to share the fractional values
5383 between genmodes* and the rest of the compiler.
5387 * config/rs6000/rs6000-modes.h: New file.
5388 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
5389 meaning of rs6000_long_double_size so that 126..128 selects an
5390 appropriate 128-bit floating point type.
5391 (rs6000_option_override_internal): Likewise.
5392 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
5393 (TARGET_LONG_DOUBLE_128): Change the meaning of
5394 rs6000_long_double_size so that 126..128 selects an appropriate
5395 128-bit floating point type.
5396 (LONG_DOUBLE_TYPE_SIZE): Update comment.
5397 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
5398 source and destination to match the standard usage.
5399 (truncifkf2): Likewise.
5400 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
5401 ISA 2.07 to use an explicit clobber, instead of passing in a
5403 (copysign<mode>3_soft): Likewise.
5405 2018-06-18 David Malcolm <dmalcolm@redhat.com>
5407 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5408 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
5409 (vect_slp_analyze_instance_dependence): Likewise.
5410 (vect_enhance_data_refs_alignment): Likewise.
5411 (vect_analyze_data_refs_alignment): Likewise.
5412 (vect_slp_analyze_and_verify_instance_alignment
5413 (vect_analyze_data_ref_accesses): Likewise.
5414 (vect_prune_runtime_alias_test_list): Likewise.
5415 (vect_analyze_data_refs): Likewise.
5416 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
5417 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5418 (vect_analyze_scalar_cycles_1): Likewise.
5419 (vect_get_loop_niters): Likewise.
5420 (vect_analyze_loop_form_1): Likewise.
5421 (vect_update_vf_for_slp): Likewise.
5422 (vect_analyze_loop_operations): Likewise.
5423 (vect_analyze_loop): Likewise.
5424 (vectorizable_induction): Likewise.
5425 (vect_transform_loop): Likewise.
5426 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
5427 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5428 (vect_make_slp_decision): Likewise.
5429 (vect_detect_hybrid_slp): Likewise.
5430 (vect_slp_analyze_operations): Likewise.
5431 (vect_slp_bb): Likewise.
5432 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
5433 (vectorizable_bswap): Likewise.
5434 (vectorizable_call): Likewise.
5435 (vectorizable_simd_clone_call): Likewise.
5436 (vectorizable_conversion): Likewise.
5437 (vectorizable_assignment): Likewise.
5438 (vectorizable_shift): Likewise.
5439 (vectorizable_operation): Likewise.
5440 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
5442 2018-06-18 Martin Sebor <msebor@redhat.com>
5444 PR tree-optimization/81384
5445 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
5446 * builtins.c (expand_builtin_strnlen): New function.
5447 (expand_builtin): Call it.
5448 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
5449 * builtins.def (BUILT_IN_STRNLEN): New.
5450 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
5451 Warn for bounds in excess of maximum object size.
5452 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
5453 single-value ranges. Handle strnlen.
5454 (handle_builtin_strlen): Handle strnlen.
5455 (strlen_check_and_optimize_stmt): Same.
5456 * doc/extend.texi (Other Builtins): Document strnlen.
5458 2018-06-18 Maya Rashish <coypu@sdf.org>
5460 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
5461 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
5462 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
5464 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
5466 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
5468 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5470 * tree.c (escaped_string::escape): Replace cast to char * by
5471 const_cast<char *> (unescaped).
5473 2018-06-18 Nick Clifton <nickc@redhat.com>
5476 * tree.c (escaped_string): New class. Converts an unescaped
5477 string into its escaped equivalent.
5478 (warn_deprecated_use): Use the new class to convert the
5479 deprecation message, if present.
5480 (test_escaped_strings): New self test.
5481 (test_c_tests): Add test_escaped_strings.
5482 * doc/extend.texi (deprecated): Add a note that the
5483 deprecation message is affected by the -fmessage-length
5484 option, and that control characters will be escaped.
5485 (#pragma GCC error): Document this pragma.
5486 (#pragma GCC warning): Likewise.
5487 * doc/invoke.texi (-fmessage-length): Document this option's
5488 effect on the #warning and #error preprocessor directives and
5489 the deprecated attribute.
5491 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
5493 * tree.c (decl_value_expr_lookup): Revert latest change.
5494 (decl_value_expr_insert): Likewise.
5496 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
5498 * gimplify.c (nonlocal_vlas): Delete.
5499 (nonlocal_vla_vars): Likewise.
5500 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
5502 (gimplify_body): Do not create and destroy nonlocal_vlas.
5503 * tree-nested.c: Include diagnostic.h.
5504 (use_pointer_in_frame): Tweak.
5505 (lookup_field_for_decl): Add assertion and declare the transformation.
5506 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
5507 internal error when the reference is in a wrong context. Do not
5508 create a debug decl by default.
5509 (note_nonlocal_block_vlas): Delete.
5510 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
5511 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
5512 create a debug decl by default.
5513 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
5514 call is in a wrong context.
5515 (fixup_vla_decls): New function.
5516 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
5517 debug variables were created.
5518 * tree.c (decl_value_expr_lookup): Add checking assertion.
5519 (decl_value_expr_insert): Likewise.
5521 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5524 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
5525 * tree-scalar-evolution.c (interpret_expr): Likewise.
5526 (expression_expensive_p): Likewise.
5527 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
5528 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
5529 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
5530 (ssa_defined_by_minus_one_stmt_p): New.
5532 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5535 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
5536 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
5537 * dojump.c (do_jump): Likewise.
5538 * expr.c (expand_expr_real_2): Check operand type's sign.
5539 * fold-const.c (const_unop): Handle ABSU_EXPR.
5540 (fold_abs_const): Likewise.
5541 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
5542 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
5543 (strip_sign_op_1): Likesise.
5544 * match.pd: Add new pattern to generate ABSU_EXPR.
5545 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
5546 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5547 * tree-eh.c (operation_could_trap_helper_p): Likewise.
5548 * tree-inline.c (estimate_operator_cost): Likewise.
5549 * tree-pretty-print.c (dump_generic_node): Likewise.
5550 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
5551 * tree.def (ABSU_EXPR): New.
5553 2018-06-16 Jakub Jelinek <jakub@redhat.com>
5556 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
5557 documented as preserved for backward compatibility only.
5558 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
5560 PR rtl-optimization/86108
5561 * bb-reorder.c (create_forwarder_block): Renamed to ...
5562 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
5563 jump from new landing pad to the second part.
5564 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
5567 2018-06-15 Jakub Jelinek <jakub@redhat.com>
5570 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
5571 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
5572 Only call store_expr for halves if the mode is the same.
5575 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
5576 Fix up comment formatting.
5578 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
5580 * typed-splay-tree.h (typed_splay_tree::remove): New function.
5581 (typed_splay_tree::closure,
5582 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
5583 (typed_splay_tree::typed_splay_tree,
5584 typed_splay_tree::operator =): Declared private.
5585 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
5586 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
5587 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
5588 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
5589 typed_splay_tree::splay_tree_splay,
5590 typed_splay_tree::splay_tree_foreach_helper,
5591 typed_splay_tree::splay_tree_insert,
5592 typed_splay_tree::splay_tree_remove,
5593 typed_splay_tree::splay_tree_lookup,
5594 typed_splay_tree::splay_tree_predecessor,
5595 typed_splay_tree::splay_tree_successor,
5596 typed_splay_tree::splay_tree_min,
5597 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
5598 (typed_splay_tree::root, typed_splay_tree::comp,
5599 typed_splay_tree::delete_key,
5600 typed_splay_tree::delete_value): New data members.
5601 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
5602 typed_splay_tree::remove.
5604 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
5606 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
5607 -mginv and -mno-ginv to the assembler.
5608 * config/mips/mips.opt (-mcrc): New option.
5610 * doc/invoke.text (-mcrc): Document.
5613 2018-06-15 Nick Clifton <nickc@redhat.com>
5616 * tree.c (escaped_string): New class. Converts an unescaped
5617 string into its escaped equivalent.
5618 (warn_deprecated_use): Use the new class to convert the
5619 deprecation message, if present.
5620 (test_escaped_strings): New self test.
5621 (test_c_tests): Add test_escaped_strings.
5622 * doc/extend.texi (deprecated): Add a note that the
5623 deprecation message is affected by the -fmessage-length
5624 option, and that control characters will be escaped.
5625 (#pragma GCC error): Document this pragma.
5626 (#pragma GCC warning): Likewise.
5627 * doc/invoke.texi (-fmessage-length): Document this option's
5628 effect on the #warning and #error preprocessor directives and
5629 the deprecated attribute.
5631 2018-06-15 Richard Biener <rguenther@suse.de>
5633 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
5634 here, also noting vector size used.
5635 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
5636 size used in MSG_OPTIMIZED_LOCATIONS dump.
5637 (pass_slp_vectorize::execute): Adjust.
5639 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
5642 * config/arc/arc.c (arc_return_address_register): Fix
5645 2018-06-15 Richard Biener <rguenther@suse.de>
5648 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
5649 leave useless conversion stripping to force_gimple_operand_gsi.
5650 (gimplify_build2): Likewise.
5651 (gimplify_build1): Likewise.
5653 2018-06-15 Richard Biener <rguenther@suse.de>
5656 * tree-cfg.c (move_stmt_op): unshare invariant addresses
5657 before adjusting their block.
5659 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
5661 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
5662 multilibs for *-*-rtems*.
5663 * config/riscv/t-rtems: New file.
5665 2018-06-14 Jakub Jelinek <jakub@redhat.com>
5668 * match.pd ((A +- CST1) +- CST2): Punt if last resort
5669 unsigned_type_for returns NULL.
5672 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
5673 subregs of multi-word pseudos unless the float mode has word size.
5675 2018-06-14 Richard Biener <rguenther@suse.de>
5678 * tree-vect-generic.c (build_word_mode_vector_type): Remove
5679 duplicate and harmful type_hash_canon.
5680 * tree.c (type_hash_canon): Assert we didn't find ourselves.
5682 2018-06-14 Richard Biener <rguenther@suse.de>
5685 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
5688 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
5690 * config/rtems.h (STDINT_LONG32): Define.
5692 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
5693 Prachi Godbole <prachi.godbole@imgtec.com>
5695 * config/mips/mips-cpus.def: Define P6600.
5696 * config/mips/mips-tables.opt: Regenerate.
5697 * config/mips/mips.c (mips_ucbranch_type): New enum.
5698 (mips_rtx_cost_data): Add support for P6600.
5699 (mips_issue_rate): Likewise.
5700 (mips_multipass_dfa_lookahead): Likewise.
5701 (mips_avoid_hazard): Likewise.
5702 (mips_reorg_process_insns): Likewise.
5703 (mips_classify_branch_p6600): New function.
5704 * config/mips/mips.h (TUNE_P6600): New define.
5705 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
5706 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
5707 * config/mips/mips.md: Include p6600.md.
5708 (processor): Add p6600.
5709 * config/mips/p6600.md: New file.
5710 * doc/invoke.texi: Add p6600 to supported architectures.
5712 2018-06-13 Martin Sebor <msebor@redhat.com>
5714 PR tree-optimization/86114
5715 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
5717 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
5719 2018-06-13 Richard Biener <rguenther@suse.de>
5721 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
5722 Properly set vector type of the intermediate stmt.
5723 * tree-vect-stmts.c (vectorizable_operation): The destination
5724 var always has vectype_out type.
5726 2018-06-13 Jeff Law <law@redhat.com>
5728 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
5729 integer 0 for argument to print_rtl_with_bb.
5730 (rl78_reorg): Likewise.
5732 2018-06-13 David Malcolm <dmalcolm@redhat.com>
5734 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
5735 from rtx to rtx_insn *.
5736 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
5738 (add_sched_insns_for_speculation): Likewise for local "target",
5739 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5740 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
5741 from rtx_insn ** to rtx_code_label **.
5742 (reorg_emit_nops): Likewise.
5743 (c6x_reorg): Likewise for local "call_labels".
5744 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
5746 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
5747 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
5748 the loops over LABEL_REFS.
5749 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
5751 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5752 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
5753 (split_branches): Strengthen local "olabel" from rtx to
5754 rtx_insn *, adding a safe_as_a cast.
5755 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
5757 (add_insn_after): Likewise for first two params.
5758 (add_insn_before): Likewise.
5759 (remove_insn): Likewise for param.
5760 (emit_pattern_before_noloc): Likewise for second and third params.
5761 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
5762 (emit_call_insn_before_noloc): Likewise.
5763 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
5765 (emit_barrier_before): Likewise.
5766 (emit_label_before): Strengthen "label" param from "rtx" to
5767 "rtx_code_label *". Strengthen "before" param from "rtx" to
5769 (emit_insn_after_1): Strengthen "after" param from "rtx" to
5771 (emit_pattern_after_noloc): Likewise.
5772 (emit_insn_after_noloc): Likewise.
5773 (emit_jump_insn_after_noloc): Likewise.
5774 (emit_call_insn_after_noloc): Likewise.
5775 (emit_debug_insn_after_noloc): Likewise.
5776 (emit_barrier_after): Likewise.
5777 (emit_label_after): Likewise for both params.
5778 (emit_pattern_after_setloc): Likewise for "after" param. Convert
5779 "loc" param from "int" to "location_t".
5780 (emit_insn_after_setloc): Likewise.
5781 (emit_jump_insn_after_setloc): Likewise.
5782 (emit_call_insn_after_setloc): Likewise.
5783 (emit_debug_insn_after_setloc): Likewise.
5784 (emit_pattern_before_setloc): Likewise for "before" param. Convert
5785 "loc" param from "int" to "location_t".
5786 (emit_pattern_before): Convert NULL_RTX to NULL.
5787 (emit_insn_before_setloc): Convert "loc" param from "int" to
5789 (emit_jump_insn_before_setloc): Likewise.
5790 (emit_call_insn_before_setloc): Likewise.
5791 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
5792 rtx_insn *. Convert "loc" param from "int" to "location_t".
5793 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
5794 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
5795 Convert 3rd param from "int" to "location_t".
5796 (emit_barrier_before, emit_barrier_after, next_real_insn):
5797 Strengthen param from rtx to rtx_insn *.
5798 (emit_label_before): Strengthen 1st param from "rtx" to
5799 "rtx_code_label *". Strengthen 2nd param from "rtx" to
5801 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
5802 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
5803 Strengthen 2nd param from "rtx" to "rtx_insn *".
5804 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
5805 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
5806 Likewise. Convert 3rd param from "int" to "location_t".
5807 (emit_label_after): Strengthen 1st param from "rtx" to
5809 (next_real_insn, remove_insn): Strengthen param from "rtx" to
5811 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
5812 from "rtx" to "rtx_insn *".
5814 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
5816 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
5817 bodies streamed in with -Q.
5818 * dumpfile.c (dump_files): Add lto-stream-out dump file.
5819 * dumpfile.h (tree_dump_index): Add lto_stream_out.
5820 * gimple-streamer-out.c: Include gimple-pretty-print.h
5821 (output_bb): Dump stmts streamed.
5822 * lto-section-out.c: Include print-tree.h
5823 (lto_begin_section): Dump sections created.
5824 (lto_output_decl_index): Dump decl encoded.
5825 * lto-streamer-out.c: Include print-tree.h
5826 (create_output_block): Dump output block created.
5827 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
5828 (output_function): Dump function output.
5829 (output_constructor): Dump constructor streamed.
5830 (write_global_stream): Output indexes encoded.
5831 (produce_asm_for_decls): Dump streams encoded.
5832 * lto-streamer.c (streamer_dump_file): New global var.
5833 * lto-streamer.h (streamer_dump_file): Declare.
5834 * passes.c (ipa_write_summaries): Initialize streamer dump.
5835 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
5838 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
5841 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
5842 offsets for register save directives. Emit a second batch of save
5843 directives, if need be, when the function accesses prior frames.
5845 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5847 * config/arc/fpu.md (fmasf4): Force operand to register.
5848 (fnmasf4): Likewise.
5850 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5852 * config/arc/arc-protos.h (arc_pad_return): Remove.
5853 * config/arc/arc.c (machine_function): Remove force_short_suffix
5855 (arc_print_operand): Adjust printing of '&'.
5856 (arc_verify_short): Remove conditional printing of short suffix.
5857 (arc_final_prescan_insn): Remove reference to size_reason.
5858 (pad_return): New function.
5859 (arc_reorg): Call pad_return.
5860 (arc_pad_return): Remove.
5861 (arc_init_machine_status): Remove reference to force_short_suffix.
5862 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
5863 (attr length): When attribute iscompact is true force to 2
5864 regardless; in the case of maybe check if we want to force the
5865 instruction to have 4 bytes length.
5866 (nopv): Change it to generate 4 byte long nop as well.
5867 (blockage): New pattern.
5868 (simple_return): Remove call to arc_pad_return.
5869 (p_return_i): Likewise.
5871 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5873 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
5875 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5877 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
5880 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5882 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
5883 for ARC700 and ARCv2.
5885 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
5888 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
5889 operands[2] instead of operands[1].
5892 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5894 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
5895 case, check whether the outer register overlaps an unallocatable
5896 register, not just whether it fits the required class.
5898 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5900 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5902 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
5903 (memrefs_conflict_p): Likewise.
5904 (init_alias_analysis): Likewise.
5905 * cfgexpand.c (expand_debug_expr): Likewise.
5906 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
5907 * cse.c (fold_rtx): Likewise.
5908 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
5909 * expr.c (emit_block_move_hints): Likewise.
5910 (clear_storage_hints, push_block, emit_push_insn): Likewise.
5911 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
5912 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
5913 (emit_group_store): Likewise.
5914 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
5915 to read the PRE/POST_MODIFY increment.
5916 * calls.c (store_one_arg): Use strip_offset.
5917 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
5919 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
5921 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
5922 (simplify_binary_operation_1): Extend CONST_INT handling to
5924 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
5925 than a HOST_WIDE_INT.
5926 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
5928 (adjust_mems, add_stores): Update accodingly.
5929 (vt_canonicalize_addr): Track polynomial offsets.
5930 (emit_note_insn_var_location): Likewise.
5931 (vt_add_function_parameter): Likewise.
5932 (vt_initialize): Likewise.
5934 2018-06-12 Jeff Law <law@redhat.com>
5936 * config.gcc (alpha*-*-freebsd*): Remove.
5937 * config/alpha/freebsd.h: Remove.
5939 2018-06-12 David Malcolm <dmalcolm@redhat.com>
5942 * spellcheck-tree.c (levenshtein_distance): Rename to...
5943 (get_edit_distance): ...this, and update for underlying renaming.
5944 * spellcheck-tree.h (levenshtein_distance): Rename to...
5945 (get_edit_distance): ...this.
5946 * spellcheck.c (levenshtein_distance): Rename to...
5947 (get_edit_distance): ...this. Convert from Levenshtein distance
5948 to Damerau-Levenshtein distance by supporting transpositions of
5949 adjacent characters. Rename "v1" to "v_next" and "v0" to
5951 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
5952 (selftest::test_edit_distance_unit_test_oneway): ...this, and
5953 update for underlying renaming.
5954 (selftest::levenshtein_distance_unit_test): Rename to...
5955 (selftest::test_get_edit_distance_unit): ...this, and update for
5956 underlying renaming.
5957 (selftest::test_find_closest_string): Add example from PR 69968
5958 where transposition helps
5959 (selftest::test_metric_conditions): Update for renaming.
5960 (selftest::test_metric_conditions): Likewise.
5961 (selftest::spellcheck_c_tests): Likewise.
5962 * spellcheck.h (levenshtein_distance): Rename both overloads to...
5963 (get_edit_distance): ...this.
5964 (best_match::consider): Update for renaming.
5966 2018-06-12 Martin Sebor <msebor@redhat.com>
5968 PR tree-optimization/85259
5969 * builtins.c (compute_objsize): Handle constant offsets.
5970 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
5971 true iff a warning has been issued.
5972 * gimple.h (gimple_nonartificial_location): New function.
5973 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
5974 gimple_nonartificial_location and handle -Wno-system-headers.
5975 (handle_builtin_stxncpy): Same.
5977 2018-06-12 Martin Sebor <msebor@redhat.com>
5980 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
5982 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
5984 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5985 BUILTIN_VEC_XST entries for pointer to double and long long.
5987 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
5990 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5992 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5995 2018-06-12 Martin Liska <mliska@suse.cz>
5997 * doc/options.texi: Document IntegerRange.
5999 2018-06-12 Martin Liska <mliska@suse.cz>
6001 * config/i386/i386.opt: Make MPX-related options as Deprecated.
6002 * opt-functions.awk: Handle Deprecated flag.
6003 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
6005 (read_cmdline_option): Report warning for a deprecated option.
6006 * opts.h (struct cl_option): Add new field cl_deprecated.
6007 (CL_ERR_DEPRECATED): New.
6009 2018-06-12 Martin Liska <mliska@suse.cz>
6011 * doc/options.texi: Document Deprecated option flag.
6013 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6015 * config/arc/arc-arch.h (arc_extras): New enum.
6016 (arc_cpu_t):Add field extra.
6017 (arc_cpu_types): Consider the extras.
6018 * config/arc/arc-cpus.def: Add extras info.
6019 * config/arc/arc-opts.h (processor_type): Consider extra field.
6020 * config/arc/arc.c (arc_override_options): Handle extra field.
6022 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
6024 * config/arc/arc-arch.h: Update ARC_OPTX macro.
6025 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
6027 * config/arc/arc.c (arc_init): Update pic warning.
6028 (irq_range): Update irq range parsing warnings.
6029 (arc_override_options): Update various warning messages.
6030 (arc_handle_aux_attribute): Likewise.
6032 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
6034 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
6036 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6038 * doc/sourcebuild.texi: Document usage of line number 0 in verify
6039 compiler messages directives.
6041 2018-06-12 Matthew Fortune <mfortune@gmail.com>
6043 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
6044 * config/mips/mips-tables.opt: Regenerate.
6045 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
6047 * doc/invoke.texi: Document -march=i6500.
6049 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
6051 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
6052 (i6400_gpmul): Add cpu_unit.
6053 (i6400_gpdiv): Likewise.
6054 (i6400_msa_add_d): Update reservations.
6055 (i6400_msa_int_add) Likewise.
6056 (i6400_msa_short_logic3) Likewise.
6057 (i6400_msa_short_logic2) Likewise.
6058 (i6400_msa_short_logic) Likewise.
6059 (i6400_msa_move) Likewise.
6060 (i6400_msa_cmp) Likewise.
6061 (i6400_msa_short_float2) Likewise.
6062 (i6400_msa_div_d) Likewise.
6063 (i6400_msa_long_logic1) Likewise.
6064 (i6400_msa_long_logic2) Likewise.
6065 (i6400_msa_mult) Likewise.
6066 (i6400_msa_long_float2) Likewise.
6067 (i6400_msa_long_float4) Likewise.
6068 (i6400_msa_long_float5) Likewise.
6069 (i6400_msa_long_float8) Likewise.
6070 (i6400_fpu_fadd): Include frint type.
6071 (i6400_fpu_store): New define_insn_reservation.
6072 (i6400_fpu_load): Likewise.
6073 (i6400_fpu_move): Likewise.
6074 (i6400_fpu_fcmp): Likewise.
6075 (i6400_fpu_fmadd): Likewise.
6076 (i6400_int_mult): Include imul3nc type and update reservation.
6077 (i6400_int_div): Include idiv3 type and update reservation.
6078 (i6400_int_load): Update to check type not move_type.
6079 (i6400_int_store): Likewise.
6080 (i6400_int_prefetch): Set zero latency.
6082 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
6084 * gcc.c: Document new %@{...} sequence.
6085 (LINK_COMMAND_SPEC): Use it for the -L switches.
6086 (cpp_unique_options): Use it for the -I switches.
6087 (at_file_argbuf): New global variable.
6088 (in_at_file): Likewise.
6089 (alloc_args): Create at_file_argbuf.
6090 (clear_args): Truncate at_file_argbuf.
6091 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
6092 (open_at_file): New function.
6093 (close_at_file): Likewise.
6094 (create_at_file): Delete.
6095 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
6098 (validate_switches_from_spec): Deal with %@{...} sequence.
6099 (validate_switches): Likewise.
6100 (driver::finalize): Call clear_args.
6102 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6104 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
6106 2018-06-11 Martin Sebor <msebor@redhat.com>
6108 * doc/invoke.texi (-Wall): List -Wc++17-compat.
6109 (Wno-class-memaccess): Add @opindex.
6110 (Wno-templates, Wno-multiple-inheritance): Same.
6111 (Wno-virtual-inheritance, Wno-namespaces): Same.
6112 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
6113 (Wno-format-overflow, Wno-format-truncation): Same.
6114 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
6115 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
6116 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
6117 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
6118 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
6119 (Wno-misspelled-isr): Same.
6121 2018-06-11 Martin Sebor <msebor@redhat.com>
6123 * PR tree-optimization/86083
6124 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
6126 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
6128 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
6130 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
6133 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
6134 on the correct operand.
6135 (*movdi_internal64): Ditto.
6137 2018-06-11 Martin Liska <mliska@suse.cz>
6139 PR tree-optimization/86089
6140 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
6142 2018-06-11 Julia Koval <julia.koval@intel.com>
6144 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
6145 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
6146 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
6148 2018-06-11 Olivier Hainque <hainque@adacore.com>
6150 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
6151 for Ada with strict dwarf2.
6153 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
6156 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
6159 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
6161 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
6163 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
6165 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
6168 2018-06-08 Martin Liska <mliska@suse.cz>
6170 * tree-cfg.h (debug_function): Fix argument type to match
6173 2018-06-08 Martin Liska <mliska@suse.cz>
6175 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
6176 Remove usage of MPX-related (and removed) fields.
6177 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
6179 2018-06-08 David Malcolm <dmalcolm@redhat.com>
6181 * cfg.c (debug): Use TDF_NONE rather than 0.
6182 * cfghooks.c (debug): Likewise.
6183 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
6184 (struct dump_option_value_info): Convert to...
6185 (struct kv_pair): ...this template type.
6186 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
6188 (optinfo_verbosity_options): Likewise.
6189 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
6191 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
6192 than int for "optgroup_flags" param.
6193 (dump_generic_expr_loc): Use dump_flags_t rather than int for
6195 (dump_dec): Likewise.
6196 (dump_finish): Use TDF_NONE rather than 0.
6197 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
6198 rather than int for "optgroup_flags" param. Use TDF_NONE rather
6199 than 0. Update for change to option_ptr.
6200 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
6201 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
6202 0. Update for changes to optinfo_verbosity_options and
6204 (opt_info_switch_p): Convert optgroup_flags from int to
6206 (dump_basic_block): Use dump_flags_t rather than int
6207 for "dump_kind" param.
6208 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
6209 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
6210 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
6211 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
6212 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
6213 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
6214 TDF_NONE): Convert from macros to...
6215 (enum dump_flag): ...this new enum.
6216 (dump_flags_t): Update to use enum.
6217 (operator|, operator&, operator~, operator|=, operator&=):
6218 Implement for dump_flags_t.
6219 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
6220 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
6221 Convert from macros to...
6222 (enum optgroup_flag): ...this new enum.
6223 (optgroup_flags_t): New typedef.
6224 (operator|, operator|=): Implement for optgroup_flags_t.
6225 (struct dump_file_info): Convert field "alt_flags" to
6226 dump_flags_t. Convert field "optgroup_flags" to
6228 (dump_basic_block): Use dump_flags_t rather than int for param.
6229 (dump_generic_expr_loc): Likewise.
6230 (dump_dec): Likewise.
6231 (dump_register): Convert param "optgroup_flags" to
6233 (opt_info_enable_passes): Likewise.
6234 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
6236 * gimple-pretty-print.c (debug): Likewise.
6237 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
6238 (merged_store_group::apply_stores): Likewise.
6239 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
6240 * gimple.c (verify_gimple_pp): Likewise.
6241 * graphite-poly.c (print_pbb_body): Likewise.
6242 * passes.c (pass_manager::register_one_dump_file): Convert
6243 local "optgroup_flags" to optgroup_flags_t.
6244 * print-tree.c (print_node): Use TDF_NONE rather than 0.
6246 (debug_body): Likewise.
6247 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
6248 to optgroup_flags_t.
6249 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
6251 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
6252 (convert_mult_to_fma): Likewise.
6253 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
6254 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
6255 * tree-vect-data-refs.c (dump_lower_bound): Convert param
6256 "dump_kind" to dump_flags_t.
6258 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6260 * config/rs6000/rs6000.c (min, max): Delete.
6262 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
6264 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
6265 -mabi=spe and -mabi=no-spe.
6267 2018-06-08 Martin Liska <mliska@suse.cz>
6269 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
6270 where we expect an existing summary.
6272 2018-06-08 Martin Liska <mliska@suse.cz>
6274 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
6275 * ipa-inline.h (estimate_edge_growth): Likewise.
6277 2018-06-08 Martin Liska <mliska@suse.cz>
6279 * cgraph.c (function_version_hasher::hash): Use
6280 cgraph_node::get_uid ().
6281 (function_version_hasher::equal):
6282 * cgraph.h (cgraph_node::get_uid): New method.
6283 * ipa-inline.c (update_caller_keys): Use
6284 cgraph_node::get_uid ().
6285 (update_callee_keys): Likewise.
6286 * ipa-utils.c (searchc): Likewise.
6287 (ipa_reduced_postorder): Likewise.
6288 * lto-cgraph.c (input_node): Likewise.
6289 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
6290 * symbol-summary.h (symtab_insertion): Likewise.
6291 (symtab_removal): Likewise.
6292 (symtab_duplication): Likewise.
6293 * tree-pretty-print.c (dump_function_header): Likewise.
6294 * tree-sra.c (convert_callers_for_node): Likewise.
6296 2018-06-08 Martin Liska <mliska@suse.cz>
6298 * cgraph.c (symbol_table::create_edge): Always assign a new
6300 (symbol_table::free_edge): Do not recycle numbers.
6301 * cgraph.h (cgraph_edge::get): New method.
6302 * symbol-summary.h (symtab_removal): Use it.
6303 (symtab_duplication): Likewise.
6304 (call_summary::hashable_uid): Remove.
6306 2018-06-08 Martin Liska <mliska@suse.cz>
6308 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
6309 (initialize_growth_caches): Remove.
6310 (free_growth_caches): Likewise.
6311 (do_estimate_edge_time): Use edge_growth_cache.
6312 (do_estimate_edge_size): Likewise.
6313 (do_estimate_edge_hints): Likewise.
6314 * ipa-inline.c (reset_edge_caches): Likewise.
6315 (recursive_inlining): Likewise.
6316 (inline_small_functions): Likewise.
6317 * ipa-inline.h (initialize_growth_caches): Remove.
6318 (estimate_edge_size): Likewise.
6319 (estimate_edge_time): Likewise.
6320 (estimate_edge_hints): Likewise.
6321 (reset_edge_growth_cache): Likewise.
6322 * symbol-summary.h (call_summary::remove): New method.
6324 2018-06-08 Martin Liska <mliska@suse.cz>
6326 * ipa-cp.c (class edge_clone_summary): New summary.
6327 (grow_edge_clone_vectors): Remove.
6328 (ipcp_edge_duplication_hook): Remove.
6329 (class edge_clone_summary_t): New call_summary class.
6330 (ipcp_edge_removal_hook): Remove.
6331 (edge_clone_summary_t::duplicate): New function.
6332 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
6333 (create_specialized_node): Likewise.
6334 (ipcp_driver): Initialize edge_clone_summaries and do not
6337 2018-06-08 Martin Liska <mliska@suse.cz>
6339 * symbol-summary.h (get): New function.
6340 (call_summary::m_initialize_when_cloning): New class member.
6342 2018-06-08 Martin Liska <mliska@suse.cz>
6344 * cgraph.c (cgraph_node::remove): Do not recycle uid.
6345 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
6346 (symbol_table::allocate_cgraph_symbol): Do not set uid.
6347 * passes.c (uid_hash_t): Record removed_nodes by their uids.
6348 (remove_cgraph_node_from_order): Use the removed_nodes set.
6349 (do_per_function_toporder): Likwise.
6350 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
6351 instead of summary_uid.
6352 (symtab_removal): Likewise.
6353 (symtab_duplication): Likewise.
6355 2018-06-08 Martin Liska <mliska@suse.cz>
6357 * ipa-cp.c (ipcp_store_bits_results): Use
6358 ipcp_transformation_sum.
6359 (ipcp_store_vr_results): Likewise.
6360 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
6362 (ipcp_transformation_initialize): ... this.
6363 (ipa_set_node_agg_value_chain):
6364 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
6365 (write_ipcp_transformation_info): Likewise.
6366 (read_ipcp_transformation_info): Likewise.
6367 (ipcp_update_bits): Likewise.
6368 (ipcp_update_vr): Likewise.
6369 (ipcp_transform_function): Likewise.
6370 * ipa-prop.h: Rename ipcp_transformation_summary to
6371 ipcp_transformation.
6372 (class ipcp_transformation_t): New function summary.
6373 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
6374 (ipa_get_agg_replacements_for_node): Likewise.
6376 2018-06-08 Martin Liska <mliska@suse.cz>
6378 * ipa-pure-const.c (struct funct_state_d): Do it class instead
6380 (class funct_state_summary_t): New function_summary class.
6381 (has_function_state): Remove.
6382 (get_function_state): Likewise.
6383 (set_function_state): Likewise.
6384 (add_new_function): Likewise.
6385 (funct_state_summary_t::insert): New function.
6386 (duplicate_node_data): Remove.
6387 (remove_node_data): Remove.
6388 (funct_state_summary_t::duplicate): New function.
6389 (register_hooks): Create new funct_state_summaries.
6390 (pure_const_generate_summary): Use it.
6391 (pure_const_write_summary): Likewise.
6392 (pure_const_read_summary): Likewise.
6393 (propagate_pure_const): Likewise.
6394 (propagate_nothrow): Likewise.
6395 (dump_malloc_lattice): Likewise.
6396 (propagate_malloc): Likewise.
6397 (execute): Do not register hooks, just remove summary
6399 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
6402 2018-06-08 Martin Liska <mliska@suse.cz>
6404 * ipa-reference.c (remove_node_data): Remove.
6405 (duplicate_node_data): Likewise.
6406 (class ipa_ref_var_info_summary_t): New class.
6407 (class ipa_ref_opt_summary_t): Likewise.
6408 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
6409 (get_reference_optimization_summary): Use
6410 ipa_ref_opt_sum_summaries.
6411 (set_reference_vars_info): Remove.
6412 (set_reference_optimization_summary): Likewise.
6413 (ipa_init): Create summaries.
6414 (init_function_info): Use function summary.
6415 (ipa_ref_opt_summary_t::duplicate): New function.
6416 (ipa_ref_opt_summary_t::remove): New function.
6417 (get_read_write_all_from_node): Fix GNU coding style.
6418 (propagate): Use function summary.
6419 (write_node_summary_p): Fix GNU coding style.
6420 (stream_out_bitmap): Likewise.
6421 (ipa_reference_read_optimization_summary): Use function summary.
6422 (ipa_reference_c_finalize): Do not release hooks.
6424 2018-06-08 Martin Liska <mliska@suse.cz>
6426 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
6427 (analyze_function_body): Extract multiple calls of get_create.
6428 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6429 * ipa-inline.c (recursive_inlining): Use ::get method.
6430 * ipa-inline.h (estimate_edge_growth): Likewise.
6432 2018-06-08 Martin Liska <mliska@suse.cz>
6434 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
6436 (hsa_function_summary::hsa_function_summary): Use the new enum
6438 (hsa_gpu_implementation_p): Use hsa_summaries::get.
6439 * hsa-gen.c (hsa_get_host_function): Likewise.
6440 (get_brig_function_name): Likewise.
6441 * ipa-hsa.c (process_hsa_functions): Likewise.
6442 (ipa_hsa_write_summary): Likewise.
6443 * symbol-summary.h (symtab_duplication): Use ::get function/
6444 (get): New function.
6446 2018-06-08 Martin Liska <mliska@suse.cz>
6448 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
6450 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
6451 (hsa_register_kernel): Likewise.
6452 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
6453 * hsa-gen.c (hsa_get_host_function): Likewise.
6454 (get_brig_function_name): Likewise.
6455 (generate_hsa): Likewise.
6456 (pass_gen_hsail::execute): Likewise.
6457 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
6458 (devirtualization_time_bonus): Likewise.
6459 (ipcp_propagate_stage): Likewise.
6460 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6461 (edge_set_predicate): Likewise.
6462 (evaluate_conditions_for_known_args): Likewise.
6463 (evaluate_properties_for_edge): Likewise.
6464 (ipa_fn_summary::reset): Likewise.
6465 (ipa_fn_summary_t::duplicate): Likewise.
6466 (dump_ipa_call_summary): Likewise.
6467 (ipa_dump_fn_summary): Likewise.
6468 (analyze_function_body): Likewise.
6469 (compute_fn_summary): Likewise.
6470 (estimate_edge_devirt_benefit): Likewise.
6471 (estimate_edge_size_and_time): Likewise.
6472 (estimate_calls_size_and_time): Likewise.
6473 (estimate_node_size_and_time): Likewise.
6474 (inline_update_callee_summaries): Likewise.
6475 (remap_edge_change_prob): Likewise.
6476 (remap_edge_summaries): Likewise.
6477 (ipa_merge_fn_summary_after_inlining): Likewise.
6478 (ipa_update_overall_fn_summary): Likewise.
6479 (read_ipa_call_summary): Likewise.
6480 (inline_read_section): Likewise.
6481 (write_ipa_call_summary): Likewise.
6482 (ipa_fn_summary_write): Likewise.
6483 (ipa_free_fn_summary): Likewise.
6484 * ipa-hsa.c (process_hsa_functions): Likewise.
6485 (ipa_hsa_write_summary): Likewise.
6486 (ipa_hsa_read_section): Likewise.
6487 * ipa-icf.c (sem_function::merge): Likewise.
6488 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6489 (do_estimate_edge_time): Likewise.
6490 (estimate_size_after_inlining): Likewise.
6491 (estimate_growth): Likewise.
6492 (growth_likely_positive): Likewise.
6493 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6494 (inline_call): Likewise.
6495 * ipa-inline.c (caller_growth_limits): Likewise.
6496 (can_inline_edge_p): Likewise.
6497 (can_inline_edge_by_limits_p): Likewise.
6498 (compute_uninlined_call_time): Likewise.
6499 (compute_inlined_call_time): Likewise.
6500 (want_inline_small_function_p): Likewise.
6501 (edge_badness): Likewise.
6502 (update_caller_keys): Likewise.
6503 (update_callee_keys): Likewise.
6504 (recursive_inlining): Likewise.
6505 (inline_small_functions): Likewise.
6506 (inline_to_all_callers_1): Likewise.
6507 (dump_overall_stats): Likewise.
6508 (early_inline_small_functions): Likewise.
6509 (early_inliner): Likewise.
6510 * ipa-inline.h (estimate_edge_growth): Likewise.
6511 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6512 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6513 * ipa-prop.h (IPA_NODE_REF): Likewise.
6514 (IPA_EDGE_REF): Likewise.
6515 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6516 (propagate_malloc): Likewise.
6517 * ipa-split.c (execute_split_functions): Likewise.
6518 * symbol-summary.h: Rename get to get_create.
6520 (get_create): Likewise.
6521 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6523 2018-06-08 Martin Liska <mliska@suse.cz>
6525 * symbol-summary.h (release): Move definition out of class
6527 (symtab_removal): Likewise.
6528 (symtab_duplication): Likewise.
6530 2018-06-08 Martin Liska <mliska@suse.cz>
6532 * symbol-summary.h (function_summary): Move constructor
6533 implementation out of class declaration.
6534 (release): Likewise.
6535 (symtab_insertion): Likewise.
6536 (symtab_removal): Likewise.
6537 (symtab_duplication): Likewise.
6540 2018-06-08 Martin Liska <mliska@suse.cz>
6542 * Makefile.in: Remove support for MPX (macros, related functions,
6543 fields in cgraph_node, ...).
6544 * builtin-types.def (BT_BND): Likewise.
6545 (BT_FN_BND_CONST_PTR): Likewise.
6546 (BT_FN_CONST_PTR_BND): Likewise.
6547 (BT_FN_VOID_PTR_BND): Likewise.
6548 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
6549 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
6550 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
6551 (expand_builtin_mempcpy_with_bounds): Likewise.
6552 (expand_builtin_memset_with_bounds): Likewise.
6553 (expand_builtin_memset_args): Likewise.
6554 (std_expand_builtin_va_start): Likewise.
6555 (expand_builtin): Likewise.
6556 (expand_builtin_with_bounds): Likewise.
6557 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
6558 (DEF_LIB_BUILTIN_CHKP): Likewise.
6559 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
6560 (DEF_CHKP_BUILTIN): Likewise.
6561 (BUILT_IN_MEMCPY): Likewise.
6562 (BUILT_IN_MEMMOVE): Likewise.
6563 (BUILT_IN_MEMPCPY): Likewise.
6564 (BUILT_IN_MEMSET): Likewise.
6565 (BUILT_IN_STPCPY): Likewise.
6566 (BUILT_IN_STRCAT): Likewise.
6567 (BUILT_IN_STRCHR): Likewise.
6568 (BUILT_IN_STRCPY): Likewise.
6569 (BUILT_IN_STRLEN): Likewise.
6570 (BUILT_IN_MEMCPY_CHK): Likewise.
6571 (BUILT_IN_MEMMOVE_CHK): Likewise.
6572 (BUILT_IN_MEMPCPY_CHK): Likewise.
6573 (BUILT_IN_MEMSET_CHK): Likewise.
6574 (BUILT_IN_STPCPY_CHK): Likewise.
6575 (BUILT_IN_STRCAT_CHK): Likewise.
6576 (BUILT_IN_STRCPY_CHK): Likewise.
6577 * calls.c (store_bounds): Likewise.
6578 (emit_call_1): Likewise.
6579 (special_function_p): Likewise.
6580 (maybe_warn_nonstring_arg): Likewise.
6581 (initialize_argument_information): Likewise.
6582 (finalize_must_preallocate): Likewise.
6583 (compute_argument_addresses): Likewise.
6584 (expand_call): Likewise.
6585 * cfgexpand.c (expand_call_stmt): Likewise.
6586 (expand_return): Likewise.
6587 (expand_gimple_stmt_1): Likewise.
6588 (pass_expand::execute): Likewise.
6589 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
6590 (cgraph_node::remove): Likewise.
6591 (cgraph_node::dump): Likewise.
6592 (cgraph_node::verify_node): Likewise.
6593 * cgraph.h (chkp_function_instrumented_p): Likewise.
6594 (symtab_node::get_alias_target): Likewise.
6595 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
6596 (cgraph_local_p): Likewise.
6597 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
6598 (cgraph_edge::rebuild_references): Likewise.
6599 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
6600 (walk_polymorphic_call_targets): Likewise.
6601 (cgraph_node::expand_thunk): Likewise.
6602 (symbol_table::output_weakrefs): Likewise.
6603 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
6604 (ix86_handle_option): Likewise.
6605 * config/i386/constraints.md: Likewise.
6606 * config/i386/i386-builtin-types.def (BND): Likewise.
6610 * config/i386/i386-builtin.def (BDESC_END): Likewise.
6611 (BDESC_FIRST): Likewise.
6613 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
6614 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
6615 * config/i386/i386.c (enum reg_class): Likewise.
6616 (ix86_target_string): Likewise.
6617 (ix86_option_override_internal): Likewise.
6618 (ix86_conditional_register_usage): Likewise.
6619 (ix86_valid_target_attribute_inner_p): Likewise.
6620 (ix86_set_indirect_branch_type): Likewise.
6621 (ix86_set_current_function): Likewise.
6622 (ix86_function_arg_regno_p): Likewise.
6623 (init_cumulative_args): Likewise.
6624 (ix86_function_arg_advance): Likewise.
6625 (ix86_function_arg): Likewise.
6626 (ix86_pass_by_reference): Likewise.
6627 (ix86_function_value_regno_p): Likewise.
6628 (ix86_function_value_1): Likewise.
6629 (ix86_function_value_bounds): Likewise.
6630 (ix86_return_in_memory): Likewise.
6631 (ix86_setup_incoming_vararg_bounds): Likewise.
6632 (ix86_va_start): Likewise.
6633 (indirect_thunk_need_prefix): Likewise.
6634 (print_reg): Likewise.
6635 (ix86_print_operand): Likewise.
6636 (ix86_expand_call): Likewise.
6637 (ix86_output_function_return): Likewise.
6638 (reg_encoded_number): Likewise.
6639 (BDESC_VERIFYS): Likewise.
6640 (ix86_init_mpx_builtins): Likewise.
6641 (ix86_init_builtins): Likewise.
6642 (ix86_emit_cmove): Likewise.
6643 (ix86_emit_move_max): Likewise.
6644 (ix86_expand_builtin): Likewise.
6645 (ix86_builtin_mpx_function): Likewise.
6646 (ix86_get_arg_address_for_bt): Likewise.
6647 (ix86_load_bounds): Likewise.
6648 (ix86_store_bounds): Likewise.
6649 (ix86_load_returned_bounds): Likewise.
6650 (ix86_store_returned_bounds): Likewise.
6651 (ix86_class_likely_spilled_p): Likewise.
6652 (ix86_hard_regno_mode_ok): Likewise.
6653 (x86_order_regs_for_local_alloc): Likewise.
6654 (ix86_mitigate_rop): Likewise.
6655 (ix86_bnd_prefixed_insn_p): Likewise.
6656 (ix86_mpx_bound_mode): Likewise.
6657 (ix86_make_bounds_constant): Likewise.
6658 (ix86_initialize_bounds): Likewise.
6659 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
6660 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
6661 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
6662 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
6663 (TARGET_CHKP_BOUND_MODE): Likewise.
6664 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
6665 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
6666 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
6667 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
6668 * config/i386/i386.h (TARGET_MPX): Likewise.
6669 (TARGET_MPX_P): Likewise.
6670 (VALID_BND_REG_MODE): Likewise.
6671 (FIRST_BND_REG): Likewise.
6672 (LAST_BND_REG): Likewise.
6673 (enum reg_class): Likewise.
6674 (BND_REG_P): Likewise.
6675 (BND_REGNO_P): Likewise.
6676 (BNDmode): Likewise.
6677 (ADJUST_INSN_LENGTH): Likewise.
6678 * config/i386/i386.md: Likewise.
6679 * config/i386/i386.opt: Likewise.
6680 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
6681 (defined): Likewise.
6682 (LINK_MPX): Likewise.
6683 (MPX_SPEC): Likewise.
6684 (LIBMPX_SPEC): Likewise.
6685 (LIBMPXWRAPPERS_SPEC): Likewise.
6686 (CHKP_SPEC): Likewise.
6687 * config/i386/predicates.md: Likewise.
6688 * dbxout.c (dbxout_type): Likewise.
6689 * doc/extend.texi: Likewise.
6690 * doc/invoke.texi: Likewise.
6691 * doc/md.texi: Likewise.
6692 * doc/tm.texi: Likewise.
6693 * doc/tm.texi.in: Likewise.
6694 * dwarf2out.c (is_base_type): Likewise.
6695 (gen_formal_types_die): Likewise.
6696 (gen_subprogram_die): Likewise.
6697 (gen_type_die_with_usage): Likewise.
6698 (gen_decl_die): Likewise.
6699 (dwarf2out_late_global_decl): Likewise.
6700 * expr.c (expand_assignment): Likewise.
6701 (emit_storent_insn): Likewise.
6702 (store_expr_with_bounds): Likewise.
6703 (store_expr): Likewise.
6704 (expand_expr_real_1): Likewise.
6705 * expr.h (store_expr_with_bounds): Likewise.
6706 * function.c (use_register_for_decl): Likewise.
6707 (struct bounds_parm_data): Likewise.
6708 (assign_parms_augmented_arg_list): Likewise.
6709 (assign_parm_find_entry_rtl): Likewise.
6710 (assign_parm_is_stack_parm): Likewise.
6711 (assign_parm_load_bounds): Likewise.
6712 (assign_bounds): Likewise.
6713 (assign_parms): Likewise.
6714 (expand_function_start): Likewise.
6715 * gcc.c (CHKP_SPEC): Likewise.
6716 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
6717 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
6718 (wrestrict_dom_walker::check_call): Likewise.
6719 * gimple.c (gimple_build_call_from_tree): Likewise.
6720 * gimple.h (enum gf_mask): Likewise.
6721 (gimple_call_with_bounds_p): Likewise.
6722 (gimple_call_set_with_bounds): Likewise.
6723 * gimplify.c (gimplify_init_constructor): Likewise.
6724 * ipa-cp.c (initialize_node_lattices): Likewise.
6725 (propagate_constants_across_call): Likewise.
6726 (find_more_scalar_values_for_callers_subset): Likewise.
6727 * ipa-hsa.c (process_hsa_functions): Likewise.
6728 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
6729 * ipa-icf.c (sem_function::merge): Likewise.
6730 * ipa-inline.c (early_inliner): Likewise.
6731 * ipa-pure-const.c (warn_function_noreturn): Likewise.
6732 (warn_function_cold): Likewise.
6733 (propagate_pure_const): Likewise.
6734 * ipa-ref.h (enum GTY): Likewise.
6735 * ipa-split.c (find_retbnd): Likewise.
6736 (consider_split): Likewise.
6737 (split_function): Likewise.
6738 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6739 * ipa.c (walk_polymorphic_call_targets): Likewise.
6740 (symbol_table::remove_unreachable_nodes): Likewise.
6741 (process_references): Likewise.
6742 (cgraph_build_static_cdtor_1): Likewise.
6743 * lto-cgraph.c (lto_output_node): Likewise.
6744 (output_refs): Likewise.
6745 (compute_ltrans_boundary): Likewise.
6746 (input_overwrite_node): Likewise.
6747 (input_node): Likewise.
6748 (input_cgraph_1): Likewise.
6749 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
6750 * passes.c (pass_manager::execute_early_local_passes): Likewise.
6751 (class pass_chkp_instrumentation_passes): Likewise.
6752 (make_pass_chkp_instrumentation_passes): Likewise.
6753 * passes.def: Likewise.
6754 * rtl.h (struct GTY): Likewise.
6755 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
6756 * stor-layout.c (layout_type): Likewise.
6757 * symtab.c: Likewise.
6758 * target.def: Likewise.
6759 * targhooks.c (default_chkp_bound_type): Likewise.
6760 (default_chkp_bound_mode): Likewise.
6761 (default_builtin_chkp_function): Likewise.
6762 (default_chkp_function_value_bounds): Likewise.
6763 (default_chkp_make_bounds_constant): Likewise.
6764 (default_chkp_initialize_bounds): Likewise.
6765 * targhooks.h (default_chkp_bound_type): Likewise.
6766 (default_chkp_bound_mode): Likewise.
6767 (default_builtin_chkp_function): Likewise.
6768 (default_chkp_function_value_bounds): Likewise.
6769 (default_chkp_make_bounds_constant): Likewise.
6770 (default_chkp_initialize_bounds): Likewise.
6771 * toplev.c (compile_file): Likewise.
6772 (process_options): Likewise.
6773 * tree-core.h (DEF_BUILTIN): Likewise.
6774 (DEF_BUILTIN_CHKP): Likewise.
6775 * tree-inline.c (declare_return_variable): Likewise.
6776 (remap_gimple_stmt): Likewise.
6777 (copy_bb): Likewise.
6778 (initialize_inlined_parameters): Likewise.
6779 (expand_call_inline): Likewise.
6780 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
6781 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
6782 (make_pass_ipa_chkp_produce_thunks): Likewise.
6783 (make_pass_chkp): Likewise.
6784 (make_pass_chkp_opt): Likewise.
6785 (make_pass_chkp_instrumentation_passes): Likewise.
6786 * tree-pretty-print.c (dump_generic_node): Likewise.
6787 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
6788 * tree-ssa-dce.c (propagate_necessity): Likewise.
6789 (eliminate_unnecessary_stmts): Likewise.
6790 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
6791 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6792 * tree-ssa-sccvn.h: Likewise.
6793 * tree-ssa-strlen.c (get_string_length): Likewise.
6794 (valid_builtin_call): Likewise.
6795 (adjust_last_stmt): Likewise.
6796 (handle_builtin_strchr): Likewise.
6797 (handle_builtin_strcpy): Likewise.
6798 (handle_builtin_stxncpy): Likewise.
6799 (handle_builtin_memcpy): Likewise.
6800 (handle_builtin_strcat): Likewise.
6801 (strlen_check_and_optimize_stmt): Likewise.
6802 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
6803 * tree-streamer-in.c: Likewise.
6804 * tree-streamer.c (record_common_node): Likewise.
6805 * tree.c (tree_code_size): Likewise.
6806 (wide_int_to_tree_1): Likewise.
6807 (type_contains_placeholder_1): Likewise.
6808 (build_common_tree_nodes): Likewise.
6809 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
6810 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
6811 (POINTER_BOUNDS_P): Likewise.
6812 (BOUNDED_TYPE_P): Likewise.
6813 (BOUNDED_P): Likewise.
6814 (CALL_WITH_BOUNDS_P): Likewise.
6815 (pointer_bounds_type_node): Likewise.
6816 * value-prof.c (gimple_ic): Likewise.
6817 * var-tracking.c (vt_add_function_parameters): Likewise.
6818 * varasm.c (make_decl_rtl): Likewise.
6819 (assemble_start_function): Likewise.
6820 (output_constant): Likewise.
6821 (maybe_assemble_visibility): Likewise.
6822 * varpool.c (ctor_for_folding): Likewise.
6823 * chkp-builtins.def: Remove.
6824 * ipa-chkp.c: Remove.
6825 * ipa-chkp.h: Remove.
6826 * rtl-chkp.c: Remove.
6827 * rtl-chkp.h: Remove.
6828 * tree-chkp-opt.c: Remove.
6829 * tree-chkp.c: Remove.
6830 * tree-chkp.h: Remove.
6832 2018-06-07 Carl Love <cel@us.ibm.com>
6834 * config/rs6000/vsx.md (vextract_fp_from_shorth,
6835 vextract_fp_from_shortl): Add BE support.
6837 2018-06-07 Paul Koning <ni1d@arrl.net>
6839 * compare-elim.c (try_merge_compare): Don't merge compare if
6840 address contains a side effect.
6841 (try_eliminate_compare): Likewise.
6843 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
6845 * config.gcc: Support "tremont".
6846 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
6847 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6849 * config/i386/i386.c (m_TREMONT): Define.
6850 (processor_target_table): Add "tremont".
6851 (PTA_TREMONT): Define.
6852 (ix86_lea_outperforms): Add TARGET_TREMONT.
6853 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
6854 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
6855 and M_INTEL_GOLDMONT_PLUS.
6856 (fold_builtin_cpu): Add "tremont".
6857 (ix86_add_stmt_cost): Add TARGET_TREMONT.
6858 (ix86_option_override_internal): Add "tremont".
6859 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
6860 (processor_type): Add PROCESSOR_TREMONT.
6861 * config/i386/x86-tune.def: Add m_TREMONT.
6862 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
6864 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6866 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
6867 symbol defined for msp430i* devices to be lower case.
6869 2018-06-07 Richard Biener <rguenther@suse.de>
6871 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
6872 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
6873 Properly wrap signed arithmetic if overflow wraps.
6875 2018-06-07 Jakub Jelinek <jakub@redhat.com>
6877 PR tree-optimization/69615
6878 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
6879 of a cast from a same precision integral SSA_NAME in a bb dominated
6880 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
6881 cast to utype if rhs2 has already a compatible type.
6883 2018-06-07 Richard Biener <rguenther@suse.de>
6885 PR tree-optimization/85935
6886 * graphite-scop-detection.c (find_params_in_bb): Analyze
6887 condition operands with respect to the correct loop. Assert
6888 the analysis doesn't fail.
6890 2018-06-04 Carl Love <cel@us.ibm.com>
6892 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
6893 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
6894 as it is slightly cheaper.
6895 (first_match_or_eos_index_<mode>):
6896 Calculate index using natural element order.
6897 (first_match_index_<mode>):
6898 Calculate index using natural element order.
6899 (first_match_or_eos_index_<mode>):
6900 Calculate index using natural order.
6901 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
6902 for BE and LE modes.
6903 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
6904 P9V_BUILTIN_VCLZLSBB_V16QI.
6905 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
6908 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6910 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
6911 indentation and line wrap for many prototypes. Add missing
6912 @smallexample directives around block of prototypes for vec_xl and
6915 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
6917 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
6918 track if we pass or return IEEE 128-bit floating point.
6919 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6920 C++ mangling that is compatible with GCC 8.1.
6921 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
6922 (init_cumulative_args): Note if we pass or return IEEE 128-bit
6923 floating point types.
6924 (rs6000_function_arg_advance_1): Likewise.
6925 (rs6000_mangle_type): Optionally generate mangled names that match
6926 what GCC 8.1 generated for IEEE 128-bit floating point types.
6927 (rs6000_globalize_decl_name): If we have an external function that
6928 passes or returns IEEE 128-bit types, generate a weak reference
6929 from the mangled name used in GCC 8.1 to the current mangled
6931 (rs6000_init_builtins): Make __ibm128 use the long double type if
6932 long double is IBM extended double. Make __float128 use the long
6933 double type if long double is IEEE 128-bit.
6936 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
6937 macro for __ibm128 built-in functions.
6938 (PACK_IF): Add __ibm128 pack/unpack functions.
6939 (UNPACK_IF): Likewise.
6940 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
6941 enable long double built-in functions if long double is IEEE
6942 128-bit floating point.
6943 (rs6000_invalid_builtin): Update long double built-in function
6945 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
6946 functions, adjust the built-in function to use the long double
6947 built-in function if __ibm128 and long double are the same type.
6948 * doc/extend.texi (PowerPC builtins): Update documention for
6949 __builtin_{,un}pack_longdouble. Add documentation for
6950 __builtin_{,un}pack_ibm128.
6952 2018-06-06 Jim Wilson <jimw@sifive.com>
6954 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
6955 (struct machine_function): New field interrupt_mode.
6956 (riscv_handle_type_attribute): New function. Add forward declaration.
6957 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
6958 (riscv_expand_epilogue): Check interrupt_mode field.
6959 (riscv_set_current_function): Check interrupt attribute args and
6960 set interrupt_mode field.
6961 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
6962 (riscv_sret, riscv_uret): New.
6963 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
6964 new arguments to interrupt attribute.
6966 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
6969 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
6970 Don't handle -mcpu=power8 if -mpower9-vector is also used.
6972 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6974 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
6975 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
6976 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6977 several redundant entries.
6979 2018-06-06 David Malcolm <dmalcolm@redhat.com>
6981 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
6982 type from "rtx" to "rtx_insn *".
6983 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
6984 for local "call_insn", removing cast.
6985 (ix86_expand_call): Likewise, introducing a "call_insn" local.
6987 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
6989 PR tree-optimization/86066
6990 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
6991 for BIT_INSERT_EXPR stores.
6993 2018-06-06 Richard Biener <rguenther@suse.de>
6995 PR tree-optimization/86062
6996 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6997 component refs ontop
6998 of to be offsetted base.
7000 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7002 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
7003 to be static and remove check on interrupt attribute name.
7005 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
7007 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7008 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
7010 2018-06-05 Steve Ellcey <sellcey@cavium.com>
7013 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
7015 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
7016 Remove second argument, change how error is called.
7017 (aarch64_layout_arg): Remove second argument from
7018 aarch64_err_no_fpadvsimd call.
7019 (aarch64_init_cumulative_args): Ditto.
7020 (aarch64_gimplify_va_arg_expr): Ditto.
7021 * config/aarch64/aarch64.md (mov<mode>): Ditto.
7023 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
7025 * config/i386/i386.md (simple_return_indirect_internal): New expander.
7026 (*simple_return_indirect_internal<mode>): Rename from
7027 simple_return_indirect_internal. Use W mode iterator.
7028 (rstorssp): New expander.
7029 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
7030 (clrssbsy): New expander.
7031 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
7033 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
7035 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
7037 (cmse_check_pointed_object): Likewise.
7039 2018-06-05 Martin Liska <mliska@suse.cz>
7041 PR gcov-profile/47618
7042 * doc/invoke.texi: Document how -fprofile-dir format
7045 2018-06-05 Richard Biener <rguenther@suse.de>
7047 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
7048 removal pretend DOM info isn't available so we do not update
7049 it and only remove edges, not dominated blocks. Actually free
7050 DOM info in case we removed something. Remove unreachable blocks.
7051 (mfb_keep_latches): Work with either DOM info or marked backedges.
7052 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
7053 first. Mark backedges if DOM info isn't available.
7054 (Re-)compute DOM info after cleanup_control_flow_pre.
7056 2018-06-05 Richard Biener <rguenther@suse.de>
7058 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
7059 (locus_discrim_hasher::hash): Adjust.
7060 (locus_discrim_hasher::equal): Likewise.
7061 (next_discriminator_for_locus): Work on line directly.
7062 (same_line_p): Pass in expanded locus1 as well.
7063 (assign_discriminators): Avoid redundant location expansions.
7065 2018-06-05 Richard Biener <rguenther@suse.de>
7067 PR tree-optimization/86046
7068 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
7069 if required after clearing TREE_ADDRESSABLE.
7071 2018-06-05 Richard Biener <rguenther@suse.de>
7073 PR tree-optimization/86047
7074 * tree-ssa-loop.c (for_each_index): Glob handling of all
7075 decls and constants and really handle all of them.
7077 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7080 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
7081 qualifier_void_pointer and qualifier_const_void_pointer.
7082 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7083 (arm_init_builtins): Handle the above.
7084 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
7085 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
7086 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
7089 2018-06-05 Martin Liska <mliska@suse.cz>
7091 * auto-profile.c (read_autofdo_file): Do not use
7092 gcov_ctr_summary struct.
7093 (afdo_callsite_hot_enough_for_early_inline): Likewise.
7094 * coverage.c (struct counts_entry): Likewise.
7095 (read_counts_file): Read just single summary entry.
7096 (get_coverage_counts): Use gcov_summary struct.
7097 * coverage.h (get_coverage_counts): Likewise.
7098 * gcov-dump.c (dump_working_sets): Likewise.
7099 (tag_summary): Dump just single summary.
7100 * gcov-io.c (gcov_write_summary): Write just histogram
7102 (gcov_read_summary): Read just single summary.
7103 (compute_working_sets): Use gcov_summary struct.
7104 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
7105 of GCOV_COUNTERS_SUMMABLE.
7106 (GCOV_COUNTERS_SUMMABLE): Remove.
7107 (GCOV_FIRST_VALUE_COUNTER): Replace with
7108 GCOV_COUNTER_V_INTERVAL.
7109 (struct gcov_ctr_summary): Remove.
7110 (struct gcov_summary): Directly use fields of former
7112 (compute_working_sets): Use gcov_summary struct.
7113 * gcov.c (read_count_file): Do not use ctrs fields.
7114 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
7116 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
7118 * profile.c: Likewise.
7119 * profile.h: Likewise.
7121 2018-06-05 Martin Liska <mliska@suse.cz>
7123 PR gcov-profile/84846
7124 * gcov.c (output_lines): Print working directory only
7125 in intermediate format.
7127 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
7129 * config/s390/s390-builtin-types.def: Add void function type.
7130 * config/s390/s390-builtins.def: Use the function type for the
7133 2018-06-04 Jim Wilson <jimw@sifive.com>
7135 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
7137 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
7138 and maybe_eh_return. Change regno to unsigned int. Use new args to
7139 handle EH_RETURN_DATA_REGNO registers properly.
7140 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
7141 (riscv_expand_epilogue): Update comment. Change argument name and
7142 type. Update code to use new name and type. Pass new args to
7143 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
7145 * config/riscv/riscv.md (NORMAL_RETURN): New.
7146 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
7147 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
7148 (eh_return): Call gen_eh_return_internal and emit barrier.
7149 (eh_return_internal): Call riscv_expand_epilogue.
7151 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
7153 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
7154 bit_insertion field and declare can_be_merged_into method.
7155 (merged_store_group::can_be_merged_into): New method.
7156 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
7157 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
7158 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
7160 2018-06-04 Richard Biener <rguenther@suse.de>
7162 PR tree-optimization/85955
7163 * builtins.c (fold_builtin_sincos): Convert pointers to
7164 destination to appropriate type before dereferencing.
7166 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
7168 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
7170 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
7172 * expr.c (expand_expr_real_1): Force the operand into memory if
7173 its TYPE_MODE is BLKmode and if there is no integer mode for
7174 the number of bits being extracted.
7176 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7180 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7181 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
7183 2018-06-04 Richard Biener <rguenther@suse.de>
7185 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
7186 (cleanup_tree_cfg_noloop): ... single caller. Do
7187 start_recording_case_labels later.
7189 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
7191 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
7192 to _IMMINTRIN_H_INCLUDED.
7193 * config/i386/pconfigintrin.h: Ditto.
7194 * config/i386/waitpkgintrin.h: Ditto.
7195 * config/i386/immintrin.h: Add includes for sgxintrin.h,
7196 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
7197 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
7198 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
7199 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
7200 waitpkgintrin.h and cldemoteintrin.h.
7202 2018-06-04 Richard Biener <rguenther@suse.de>
7204 PR tree-optimization/86038
7205 * tracer.c (find_best_successor): Check probability for
7206 being initialized, bail out if not.
7208 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
7211 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
7212 of bits to ignore when comparing architectures.
7214 2018-06-04 Jakub Jelinek <jakub@redhat.com>
7216 PR tree-optimization/69615
7217 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
7218 maximum or minimum of the type, try to merge it also as if
7219 range1 is + [-, x - 1] or + [x + 1, -].
7222 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
7224 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
7226 PR tree-optimization/86034
7227 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
7228 the unsigned bitfield type in a bit insertion sequence if it does not
7229 have a larger precision than the bitfield size.
7230 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
7232 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
7234 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
7236 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7238 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
7239 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
7240 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
7241 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
7243 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
7245 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
7246 Disable -fdelete-null-pointer-checks for ELF toolchain.
7248 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7249 Kito Cheng <kito.cheng@gmail.com>
7251 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
7252 (nds32le-*-*, nds32be-*-*): Integrate checking process.
7253 (nds32*-*-*): Add glibc and uclibc conditions.
7254 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
7255 (TARGET_EXCEPT_UNWIND_INFO): Define.
7256 * config/nds32/elf.h: New file.
7257 * config/nds32/linux.h: New file.
7258 * config/nds32/nds32-elf.opt: New file.
7259 * config/nds32/nds32-linux.opt: New file.
7260 * config/nds32/nds32-fp-as-gp.c
7261 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
7262 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
7264 (nds32_asm_file_end): Ditto.
7265 (nds32_print_operand): Ditto.
7266 (nds32_insert_attributes): Ditto.
7267 (nds32_init_libfuncs): New function.
7268 (TARGET_HAVE_TLS): Define.
7269 (TARGET_INIT_LIBFUNCS): Define.
7270 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
7272 (TARGET_ELF): Apply different mcmodel setting.
7273 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
7274 been migrated into elf.h and linux.h files.
7275 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
7276 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
7277 (mcmodel): The content has been migrated into nds32-elf.opt and
7278 nds32-linux.opt files.
7279 * config/nds32/t-elf: New file.
7280 * config/nds32/t-linux: New file.
7282 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
7283 Shiva Chen <shiva0217@gmail.com>
7285 * config/nds32/constants.md (unspec_volatile_element): Add
7286 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
7287 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
7289 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
7290 (make_pass_nds32_fp_as_gp): Declare.
7291 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
7293 (nds32_asm_function_end_prologue): Remove unused asm output.
7294 (nds32_asm_function_begin_epilogue): Remove unused asm output.
7295 (nds32_asm_file_start): Output necessary fp_as_gp information.
7296 (nds32_option_override): Adjust register usage.
7297 (nds32_expand_prologue): Consider fp_as_gp situation.
7298 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
7299 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
7302 (simple_return): Ditto.
7303 (omit_fp_begin): Output special directive for fp_as_gp.
7304 (omit_fp_end): Output special directive for fp_as_gp.
7305 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
7306 mforbid-fp-as-gp): New options.
7308 2018-06-01 Mark Wielaard <mark@klomp.org>
7310 * dwarf2out.c (dwarf2out_finish): Remove generation of
7311 DW_AT_loclists_base.
7313 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
7315 * gimple-ssa-store-merging.c: Include gimple-fold.h.
7316 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
7317 (struct merged_store_group): Add bit_insertion field.
7318 (dump_char_array): Use standard hexadecimal format.
7319 (merged_store_group::merged_store_group): Set bit_insertion to false.
7320 (merged_store_group::apply_stores): Use optimal buffer size. Deal
7321 with BIT_INSERT_EXPR stores. Move up code updating the mask and
7322 also print the mask in the dump file.
7323 (pass_store_merging::gate): Minor tweak.
7324 (imm_store_chain_info::coalesce_immediate): Fix wrong association
7325 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
7326 stores with INTEGER_CST stores.
7327 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
7328 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
7329 and use it throughout. Generate bit insertion sequences if need be.
7330 (pass_store_merging::process_store): Remove redundant condition.
7331 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
7333 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
7335 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
7336 the 128-bit floating point types. Fix function comment.
7338 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7340 * config/aarch64/aarch64-simd.md
7341 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
7343 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
7346 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
7348 PR tree-optimization/85989
7349 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
7351 (backprop::intersect_uses): Check it when deciding whether this
7352 is a backedge reference.
7353 (backprop::process_block): Add each phi to m_visited_phis
7354 after visiting it, then clear it at the end.
7356 2018-06-01 Richard Biener <rguenther@suse.de>
7358 * tree-vectorizer.h (vect_dr_stmt): New function.
7359 (vect_get_load_cost): Adjust.
7360 (vect_get_store_cost): Likewise.
7361 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
7362 Use vect_dr_stmt instead of DR_SMTT.
7363 (vect_record_base_alignments): Likewise.
7364 (vect_calculate_target_alignment): Likewise.
7365 (vect_compute_data_ref_alignment): Likewise and make static.
7366 (vect_update_misalignment_for_peel): Likewise.
7367 (vect_verify_datarefs_alignment): Likewise.
7368 (vector_alignment_reachable_p): Likewise.
7369 (vect_get_data_access_cost): Likewise. Pass down
7370 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
7371 (vect_get_peeling_costs_all_drs): Likewise.
7372 (vect_peeling_hash_get_lowest_cost): Likewise.
7373 (vect_enhance_data_refs_alignment): Likewise.
7374 (vect_find_same_alignment_drs): Likewise.
7375 (vect_analyze_data_refs_alignment): Likewise.
7376 (vect_analyze_group_access_1): Likewise.
7377 (vect_analyze_group_access): Likewise.
7378 (vect_analyze_data_ref_access): Likewise.
7379 (vect_analyze_data_ref_accesses): Likewise.
7380 (vect_vfa_segment_size): Likewise.
7381 (vect_small_gap_p): Likewise.
7382 (vectorizable_with_step_bound_p): Likewise.
7383 (vect_prune_runtime_alias_test_list): Likewise.
7384 (vect_analyze_data_refs): Likewise.
7385 (vect_supportable_dr_alignment): Likewise.
7386 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7387 (vect_gen_prolog_loop_niters): Likewise.
7388 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7389 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
7391 (vect_recog_mask_conversion_pattern): Likewise.
7392 (vect_try_gather_scatter_pattern): Likewise.
7393 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
7394 to vect_get_store_cost.
7395 (vect_get_store_cost): Get stmt_info instead of DR.
7396 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
7397 (vect_get_load_cost): Get stmt_info instead of DR.
7399 2018-06-01 Richard Biener <rguenther@suse.de>
7402 * gimple-fold.c (var_decl_component_p): Also allow offsetted
7403 vars wrapped in MEM_REFs.
7405 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
7407 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7408 Fix subreg tests so that we only return a choice between
7409 GENERAL_REGS and FP_REGS if the original classes included both.
7411 2018-06-01 Richard Biener <rguenther@suse.de>
7414 * tree-ssa-structalias.c (get_function_part_constraint):
7415 Handle NULL fi->decl.
7416 (find_func_aliases_for_call): Properly handle indirect
7417 fi from direct call.
7418 (find_func_clobbers): Likewise.
7419 (ipa_pta_execute): Likewise.
7420 (create_variable_info_for): For functions that are ifunc_resolver
7421 resolve to a varinfo that contains the result of the resolver call.
7422 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
7425 2018-05-31 Michael Collison <michael.collison@arm.com>
7427 * config/aarch64/aarch64.md:
7428 (*fix_to_zero_extenddfdi2): New pattern.
7429 * gcc.target/aarch64/fix_extend1.c: New testcase.
7431 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
7435 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
7436 and BUILT_IN_STRNCMP_EQ.
7437 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
7438 BUILT_IN_STRNCMP_EQ.
7439 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
7440 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7441 (gimple_fold_builtin): Likewise.
7442 * tree-ssa-strlen.c (compute_string_length): New function.
7443 (determine_min_obsize): New function.
7444 (handle_builtin_string_cmp): New function to handle calls to
7445 string compare functions.
7446 (strlen_optimize_stmt): Add handling to builtin string compare
7448 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
7449 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7450 * tree.c (build_common_builtin_nodes): Add new defines of
7451 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
7453 2018-05-31 Jakub Jelinek <jakub@redhat.com>
7456 * bb-reorder.c (pass_partition_blocks::gate): Return false for
7457 functions with naked attribute.
7459 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
7461 * config/i386/sse.md (avx_vec_concat<mode>):
7462 Substitute concat_tg_mode mode attribute with xtg_mode.
7463 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
7464 (concat_tg_mode): Remove mode attribute.
7466 2018-05-31 Martin Sebor <msebor@redhat.com>
7469 * calls.c (alloc_max_size): Correct a logic error/typo.
7470 Treat excessive arguments as infinite. Warn for invalid arguments.
7471 * doc/invoke.texi (-Walloc-size-larger-than): Update.
7473 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
7476 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
7477 and movx for Haswell.
7479 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
7480 Cesar Philippidis <cesar@codesourcery.com>
7483 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
7484 when emitting error on private/firstprivate reductions.
7485 * omp-low.c (lower_omp_target): Avoid reference-type processing
7486 on pointers for firstprivate clause.
7488 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7490 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
7493 * config/aarch64/aarch64-simd.md
7494 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
7495 (aarch64_ld1_x3_<mode>): Likewise
7496 (aarch64_st1x2<VALLDIF:mode>): Likewise
7497 (aarch64_st1_x2_<mode>): Likewise
7498 (aarch64_st1x3<VALLDIF:mode>): Likewise
7499 (aarch64_st1_x3_<mode>): Likewise
7500 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
7501 (vld1_s8_x3): Likewise.
7502 (vld1_u16_x3): Likewise.
7503 (vld1_s16_x3): Likewise.
7504 (vld1_u32_x3): Likewise.
7505 (vld1_s32_x3): Likewise.
7506 (vld1_u64_x3): Likewise.
7507 (vld1_s64_x3): Likewise.
7508 (vld1_f16_x3): Likewise.
7509 (vld1_f32_x3): Likewise.
7510 (vld1_f64_x3): Likewise.
7511 (vld1_p8_x3): Likewise.
7512 (vld1_p16_x3): Likewise.
7513 (vld1_p64_x3): Likewise.
7514 (vld1q_u8_x3): Likewise.
7515 (vld1q_s8_x3): Likewise.
7516 (vld1q_u16_x3): Likewise.
7517 (vld1q_s16_x3): Likewise.
7518 (vld1q_u32_x3): Likewise.
7519 (vld1q_s32_x3): Likewise.
7520 (vld1q_u64_x3): Likewise.
7521 (vld1q_s64_x3): Likewise.
7522 (vld1q_f16_x3): Likewise.
7523 (vld1q_f32_x3): Likewise.
7524 (vld1q_f64_x3): Likewise.
7525 (vld1q_p8_x3): Likewise.
7526 (vld1q_p16_x3): Likewise.
7527 (vld1q_p64_x3): Likewise.
7528 (vst1_s64_x2): Likewise.
7529 (vst1_u64_x2): Likewise.
7530 (vst1_f64_x2): Likewise.
7531 (vst1_s8_x2): Likewise.
7532 (vst1_p8_x2): Likewise.
7533 (vst1_s16_x2): Likewise.
7534 (vst1_p16_x2): Likewise.
7535 (vst1_s32_x2): Likewise.
7536 (vst1_u8_x2): Likewise.
7537 (vst1_u16_x2): Likewise.
7538 (vst1_u32_x2): Likewise.
7539 (vst1_f16_x2): Likewise.
7540 (vst1_f32_x2): Likewise.
7541 (vst1_p64_x2): Likewise.
7542 (vst1q_s8_x2): Likewise.
7543 (vst1q_p8_x2): Likewise.
7544 (vst1q_s16_x2): Likewise.
7545 (vst1q_p16_x2): Likewise.
7546 (vst1q_s32_x2): Likewise.
7547 (vst1q_s64_x2): Likewise.
7548 (vst1q_u8_x2): Likewise.
7549 (vst1q_u16_x2): Likewise.
7550 (vst1q_u32_x2): Likewise.
7551 (vst1q_u64_x2): Likewise.
7552 (vst1q_f16_x2): Likewise.
7553 (vst1q_f32_x2): Likewise.
7554 (vst1q_f64_x2): Likewise.
7555 (vst1q_p64_x2): Likewise.
7556 (vst1_s64_x3): Likewise.
7557 (vst1_u64_x3): Likewise.
7558 (vst1_f64_x3): Likewise.
7559 (vst1_s8_x3): Likewise.
7560 (vst1_p8_x3): Likewise.
7561 (vst1_s16_x3): Likewise.
7562 (vst1_p16_x3): Likewise.
7563 (vst1_s32_x3): Likewise.
7564 (vst1_u8_x3): Likewise.
7565 (vst1_u16_x3): Likewise.
7566 (vst1_u32_x3): Likewise.
7567 (vst1_f16_x3): Likewise.
7568 (vst1_f32_x3): Likewise.
7569 (vst1_p64_x3): Likewise.
7570 (vst1q_s8_x3): Likewise.
7571 (vst1q_p8_x3): Likewise.
7572 (vst1q_s16_x3): Likewise.
7573 (vst1q_p16_x3): Likewise.
7574 (vst1q_s32_x3): Likewise.
7575 (vst1q_s64_x3): Likewise.
7576 (vst1q_u8_x3): Likewise.
7577 (vst1q_u16_x3): Likewise.
7578 (vst1q_u32_x3): Likewise.
7579 (vst1q_u64_x3): Likewise.
7580 (vst1q_f16_x3): Likewise.
7581 (vst1q_f32_x3): Likewise.
7582 (vst1q_f64_x3): Likewise.
7583 (vst1q_p64_x3): Likewise.
7585 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7587 * config/msp430/msp430.c (msp430_output_labelref): Prepend
7588 user_label_prefix to name.
7590 * tree-core.h: Update comment about the format of NAME string
7591 passed to handler in attribute_spec.
7593 * config/msp430/msp430.md: Remove erroneous subreg expression from
7594 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
7595 zero_extend{q,h}isi2.
7597 2018-05-30 Borislav Petkov <bp@suse.de>
7599 * doc/extend.texi: Document some architecture specific
7600 constraints and sort entries.
7602 2018-05-30 Martin Sebor <msebor@redhat.com>
7605 * builtins.c (expand_builtin_stpcpy_1): New function.
7606 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
7607 only if the former succeeds.
7609 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7611 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
7614 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7616 * doc/invoke.texi (-flinker-output): Document
7618 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7620 * passes.c (ipa_write_summaries): Only modify statements if body
7622 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
7623 incrementally linking.
7624 (ipa_passes): Likewise.
7625 * lto-cgraph.c (lto_output_node): When incrementally linking do not
7626 pass down resolution info.
7627 * common.opt (flag_incremental_link): Update info.
7628 * gcc.c (plugin specs): Turn flinker-output=* to
7629 -plugin-opt=-linker-output-known
7630 * toplev.c (compile_file): Also cut compilation when doing incremental
7632 * flag-types. (enum lto_partition_model): Add
7633 LTO_LINKER_OUTPUT_NOLTOREL.
7634 (invoke.texi): Add -flinker-output docs.
7635 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
7636 link same way as WPA; do not stream in dead initializers.
7638 * dwarf2out.c (dwarf2out_die_ref_for_decl,
7639 darf2out_register_external_decl): Support incremental link.
7641 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7643 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
7645 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7647 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
7648 it down to simple_object_copy_lto_debug_sections.
7649 (run_gcc): Determine incremental LTO link time and configure
7650 lto1 into non-wpa mode, disable renaming of debug sections.
7652 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
7654 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7655 descriptions of various incorrectly documented functions.
7657 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
7660 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7661 address check not strict.
7663 2018-05-30 Richard Biener <rguenther@suse.de>
7665 PR tree-optimization/85964
7666 * tracer.c (better_p): Drop initialized count check, we only
7667 call the function with initialized counts now.
7668 (find_best_successor): Do find a best edge if one
7669 has uninitialized count.
7670 (find_best_predecessor): Likewise. Do BB frequency check only
7671 if count is initialized.
7673 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
7675 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
7676 (aarch64_ldrstr_offset_compare): New.
7677 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
7678 load/store orderings.
7679 (aarch64_gen_adjusted_ldpstp): Likewise.
7681 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
7683 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7684 Check for subset of GENERAL_REGS and FP_REGS.
7685 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
7688 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
7690 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
7691 and wi::to_poly_offset. Add the current offset and then check
7692 whether the sum fits, rather than using an unchecked addition of
7693 a checked term. Check for a shwi rather than a uhwi.
7694 * expr.c (get_bit_range): Use tree_to_poly_uint64.
7695 (store_constructor): Use poly_int_tree_p.
7696 (expand_expr_real_1): Likewise.
7697 * function.c (assign_temp): Likewise.
7698 * fold-const.c (const_binop): Use poly_int_tree_p and
7700 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
7702 * ipa-icf-gimple.c (func_checker::compare_operand): Use
7703 to_poly_offset for MEM offsets.
7704 * ipa-icf.c (sem_variable::equals): Likewise.
7705 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
7706 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
7707 wi::to_poly_offset for BIT_FIELD_REF offsets.
7708 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
7710 * var-tracking.c (emit_note_insn_var_location): Use
7711 tree_to_poly_uint64.
7713 2018-05-29 Jim Wilson <jimw@sifive.com>
7715 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
7717 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
7720 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
7721 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
7723 (sse4_1_round<mode>2): Use nonimmediate_operand
7724 for operand 1 predicate.
7726 2018-05-29 Martin Sebor <msebor@redhat.com>
7727 Richard Biener <rguenther@suse.de>
7730 * calls.c (get_size_range): Call determine_value_range instead
7731 of get_value_range..
7732 * tree-vrp.h (determine_value_range): Declared new function.
7733 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
7735 2018-05-29 Richard Biener <rguenther@suse.de>
7737 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
7738 sure to use non-pattern stmts for get_earlier_stmt arguments.
7739 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
7740 called on pattern stmts.
7741 (get_later_stmt): Likewise.
7743 2018-05-29 Martin Liska <mliska@suse.cz>
7745 PR gcov-profile/85759
7746 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
7749 2018-05-29 Jakub Jelinek <jakub@redhat.com>
7751 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
7753 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
7757 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7758 VEC_PACK_FLOAT_EXPR): New tree codes.
7759 * tree-pretty-print.c (op_code_prio): Handle
7760 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7761 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7762 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7763 * tree-inline.c (estimate_operator_cost): Likewise.
7764 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
7765 * fold-const.c (const_binop): Likewise.
7766 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
7767 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7768 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7769 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
7770 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7771 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7772 * expr.c (expand_expr_real_2): Likewise.
7773 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
7774 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
7775 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
7777 * optabs.c (expand_widen_pattern_expr): For
7778 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
7779 sign from result type rather than operand's type.
7780 (expand_binop_directly): For vec_packu_float_optab and
7781 vec_packs_float_optab allow result type to be different from operand's
7783 * optabs-tree.c (optab_for_tree_code): Handle
7784 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7785 VEC_PACK_FLOAT_EXPR. Formatting fixes.
7786 * tree-vect-generic.c (expand_vector_operations_1): Handle
7787 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7788 VEC_PACK_FLOAT_EXPR.
7789 * tree-vect-stmts.c (supportable_widening_operation): Handle
7791 (supportable_narrowing_operation): Handle FLOAT_EXPR.
7792 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
7793 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
7794 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
7795 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
7797 (vec_pack<floatprefix>_float_<mode>): New expander.
7798 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
7800 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
7801 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
7802 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
7803 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
7804 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
7806 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
7807 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
7808 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7809 VEC_PACK_FLOAT_EXPR): Document.
7811 2018-05-29 Richard Biener <rguenther@suse.de>
7813 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
7815 (stmt_vec_info_vec): Make pointer.
7816 (init_stmt_vec_info_vec): Remove.
7817 (free_stmt_vec_info_vec): Likewise.
7818 (set_stmt_vec_info_vec): New function.
7819 (free_stmt_vec_infos): Likewise.
7820 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
7821 (set_vinfo_for_stmt): Likewise.
7822 (get_earlier_stmt): Likewise.
7823 (get_later_stmt): Likewise.
7824 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
7825 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
7826 (vec_info::~vec_info): Free stmt_vec_infos.
7827 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
7828 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
7829 (pass_slp_vectorize::execute): Likewise.
7830 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
7831 (free_stmt_vec_info_vec): Likewise.
7832 (set_stmt_vec_info_vec): New function.
7833 (free_stmt_vec_infos): Likewise.
7834 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
7835 the global stmt_vec_info_vec.
7836 * tree-parloops.c (gather_scalar_reductions): Use
7837 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
7840 2018-05-29 Richard Biener <rguenther@suse.de>
7842 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
7844 2018-05-29 Martin Liska <mliska@suse.cz>
7845 David Malcolm <dmalcolm@redhat.com>
7847 * vec.c (test_reverse): New.
7848 (vec_c_tests): Add new test.
7849 * vec.h (vl_ptr>::reverse): New function.
7851 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
7853 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
7855 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
7858 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
7860 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
7862 2018-05-28 Richard Biener <rguenther@suse.de>
7864 PR tree-optimization/85933
7865 * tree-vect-data-refs.c (vect_record_base_alignments): Only
7866 look at stmts marked as vectorizable.
7868 2018-05-28 Richard Biener <rguenther@suse.de>
7870 PR tree-optimization/85934
7871 * tree-vect-generic.c (expand_vector_operations_1): Hoist
7872 vector boolean check before scalar optimization.
7874 2018-05-28 Jakub Jelinek <jakub@redhat.com>
7876 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
7879 2018-05-28 Mark Wielaard <mark@klomp.org>
7881 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
7882 if it is an expression containing a minus sign.
7884 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
7886 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
7888 2018-05-27 Paul Koning <ni1d@arrl.net>
7890 * config/pdp11/pdp11.md (truncsihi2): Remove.
7892 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
7893 Chung-Ju Wu <jasonwucj@gmail.com>
7895 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
7897 (unaligned_store_dw): Ditto.
7898 * config/nds32/nds32-memory-manipulation.c
7899 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
7900 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
7901 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
7902 (emit_setmem_word_loop): Rename to ...
7903 (emit_setmem_doubleword_loop): ... this.
7904 (nds32_gen_dup_4_byte_to_word_value): New function.
7905 (nds32_gen_dup_8_byte_to_double_word_value): New function.
7906 (nds32_expand_setmem_loop): Refine implementation.
7907 (nds32_expand_setmem_loop_v3m): Ditto.
7908 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
7911 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7913 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
7915 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7917 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
7918 (nds32_init_machine_status): Initialize machine->attr_naked_p and
7919 machine->attr_no_prologue_p.
7920 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7921 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
7922 (nds32_expand_epilogue): Consider attr_naked_p.
7923 (nds32_expand_epilogue_v3pop): Likewise.
7924 (nds32_can_use_return_insn): Likewise.
7925 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
7926 attr_no_prologue_p fields.
7927 * config/nds32/nds32.opt (mret-in-naked-func): New option.
7929 2018-05-27 Jakub Jelinek <jakub@redhat.com>
7932 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
7934 * config/i386/sse.md
7935 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
7937 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
7939 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
7941 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
7943 (*<floatsuffix>floatv2div2sf2): Rename to ...
7944 (*float<floatunssuffix>v2div2sf2): ... this.
7945 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
7946 (float<floatunssuffix>v2div2sf2_mask): ... this.
7947 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
7948 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
7949 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
7951 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
7953 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7955 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7957 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7959 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7961 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
7962 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
7963 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
7964 gen_ufix_truncv8dfv8si2.
7965 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
7966 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
7967 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
7968 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
7969 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
7970 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
7971 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
7972 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
7974 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7978 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
7980 2018-05-25 Jim Wilson <jimw@sifive.com>
7982 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
7983 * config/riscv/riscv.c (struct machine_function): Add
7984 interrupt_handler_p and attribute_checked_p fields.
7985 (riscv_attribute_table): Add interrupt.
7986 (riscv_interrupt_type_p): New.
7987 (riscv_save_reg_p): Save extra regs for interrupt handler.
7988 (riscv_use_save_libcall): Return false for interrupt handler.
7989 (riscv_first_stack_step): Add forward declaration.
7990 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
7991 for interrupt handler with large frame. Use it for saved reg list.
7992 (riscv_expand_prologue): Move flag_stack_usage_info support to
7993 eliminate duplication.
7994 (riscv_expand_epilogue): Generate mret for interrupt handler.
7995 (riscv_epilogue_uses): New.
7996 (riscv_can_use_return_insn): Return false for interrupt handler.
7997 (riscv_function_ok_for_sibcall): Likewise.
7998 (riscv_set_current_function): Add interrupt handler support.
7999 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
8000 * config/riscv/riscv.md (UNSPECV_MRET): New.
8002 (riscv_frflags, riscv_fsflags): Use tab after opcode.
8004 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
8006 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
8008 PR tree-optimization/85712
8009 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
8010 this candidate has already been replaced in-situ by a copy.
8012 2018-05-25 Jason Merrill <jason@redhat.com>
8014 PR c++/80485 - inline function non-zero address.
8015 * symtab.c (nonzero_address): Check DECL_COMDAT.
8017 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
8020 * config/alpha/alpha.md (ashlsi3): New insn pattern.
8021 (*ashlsi_se): Rename from *ashldi_se. Define as sign
8022 extension of SImode operation. Use const123_operand predicate.
8023 (*saddsi_1): Remove.
8024 (*saddl_se_1): Ditto.
8026 (*ssubl_se_1): Ditto.
8027 * config/alpha/predicates.md (const123_operand): New predicate.
8028 * config/alpha/constraints.md (P): Use IN_RANGE.
8030 2018-05-25 Richard Biener <rguenther@suse.de>
8032 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
8034 (ref_maybe_used_by_stmt_p): Likewise.
8035 (stmt_may_clobber_ref_p): Likewise.
8036 (stmt_may_clobber_ref_p_1): Likewise.
8037 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
8039 (ref_maybe_used_by_stmt_p): Likewise.
8040 (stmt_may_clobber_ref_p): Likewise.
8041 (stmt_may_clobber_ref_p_1): Likewise.
8042 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
8043 the alias oracle to disambiguate DRs with stmts DR analysis
8045 (vect_analyze_data_refs): Do not give up on not analyzable
8046 DRs for BB vectorization. Remove code truncating the dataref
8049 2018-05-25 Jakub Jelinek <jakub@redhat.com>
8052 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8053 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
8054 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
8056 2018-05-25 Richard Biener <rguenther@suse.de>
8058 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
8059 function, combining stmt data ref gathering and fatal analysis
8061 (vect_analyze_data_refs): Remove now redudnant code and simplify.
8062 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
8063 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
8064 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
8065 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
8067 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8069 PR tree-optimization/85720
8070 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
8071 SCC if all partitions are builtins.
8072 (version_loop_by_alias_check): New parameter. Generate cancelable
8073 runtime alias check if all partitions are builtins.
8074 (distribute_loop): Update call to above function.
8076 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8078 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
8079 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
8080 (parm_default_def_partition_arg): Ditto.
8081 (set_parm_default_def_partition): Ditto.
8082 (get_parm_default_def_partitions): Ditto and make it static.
8083 (get_undefined_value_partitions): Ditto and make it static.
8084 (remove_ssa_form): Refactor call to init_var_map here.
8085 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
8086 computation for loop region.
8087 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
8088 (register_default_def): Delete.
8089 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
8090 (parm_default_def_partition_arg): Ditto.
8091 (set_parm_default_def_partition): Ditto.
8092 (get_parm_default_def_partitions): Ditto and make it static.
8093 (get_undefined_value_partitions): Ditto and make it static.
8094 (coalesce_with_default, coalesce_with_default): Update comment.
8095 (create_coalesce_list_for_region): New func factored out from
8096 create_outofssa_var_map.
8097 (populate_coalesce_list_for_outofssa): New func factored out from
8098 create_outofssa_var_map and coalesce_ssa_name.
8099 (create_outofssa_var_map): Delete.
8100 (coalesce_ssa_name): Refactor to support live range computation.
8101 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
8102 (get_parm_default_def_partitions): Delete.
8103 (get_undefined_value_partitions): Ditto.
8104 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
8105 computation for loop region.
8106 (new_tree_live_info, loe_visit_block): Ditto.
8107 (live_worklist, set_var_live_on_entry): Ditto.
8108 (calculate_live_on_exit, verify_live_on_entry): Ditto.
8109 * tree-ssa-live.h (struct _var_map): New fields.
8110 (init_var_map): Change decl.
8111 (region_contains_p): New.
8113 2018-05-25 Bin Cheng <bin.cheng@arm.com>
8115 * tree-ssa-live.h (live_merge_and_clear): Delete.
8117 2018-05-25 Richard Biener <rguenther@suse.de>
8120 * tree-dump.c (dequeue_and_dump): Remove access to removed
8121 operand 2 of a SWITCH_EXPR.
8123 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8125 * doc/sourcebuild.texi (vect_double_cond_arith): Include
8126 multiplication and division.
8127 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
8128 (cond_udiv@var{m}, cond_umod@var{m}): Document.
8129 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
8130 (cond_udiv_optab, cond_umod_optab): New optabs.
8131 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
8132 (IFN_COND_RDIV): New internal functions.
8133 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
8134 TRUNC_MOD_EXPR and RDIV_EXPR.
8135 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
8136 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
8138 (SVE_INT_BINARY): Include mult.
8139 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
8140 (optab, sve_int_op): Handle mult.
8141 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
8143 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
8144 for SVE_INT_BINARY_SD.
8146 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8148 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
8149 (optab, sve_int_op): Handle div and udiv.
8150 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
8151 for SVE_INT_BINARY_SD.
8152 (*<optab><mode>3): New insn for the same.
8154 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8156 * tree-vect-patterns.c: Include predict.h.
8157 (vect_recog_divmod_pattern): Restrict check for division support
8158 to when optimizing for size.
8160 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8162 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
8163 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
8164 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
8165 (gimple_match_op::set_op): Likewise.
8166 (gimple_resimplify4): Declare.
8167 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
8168 (expr::gen_transform): Likewise.
8169 (decision_tree::gen): Generate a simplification routine for 4 operands.
8170 * gimple-match-head.c (gimple_simplify): Add an overload for
8171 4 operands. In the top-level function, handle up to 4 call
8172 arguments and call gimple_resimplify4.
8173 (gimple_resimplify4): New function.
8174 (build_call_internal): Pass a fourth operand.
8175 (maybe_push_to_seq): Likewise.
8176 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
8177 Fold VEC_COND_EXPRs of an operation and a default value into
8178 an IFN_COND_* function if possible.
8179 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
8181 (SVE_COND_FP_BINARY): Include them.
8182 (optab, sve_fp_op): Handle them.
8183 (SVE_INT_BINARY_REV): New code iterator.
8184 (SVE_COND_FP_BINARY_REV): New int iterator.
8185 (commutative): New int attribute.
8186 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
8188 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
8190 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
8191 (*cond_<optab><mode>): New patterns for reversed operands.
8193 2018-05-25 Richard Biener <rguenther@suse.de>
8195 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
8196 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
8197 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
8198 (STMT_VINFO_GROUPED_ACCESS): Adjust.
8199 * tree-vect-data-refs.c (everywhere): Adjust users.
8200 * tree-vect-loop.c (everywhere): Likewise.
8201 * tree-vect-slp.c (everywhere): Likewise.
8202 * tree-vect-stmts.c (everywhere): Likewise.
8203 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8205 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8207 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
8209 (gcc_cv_as_section_exclude): ... this.
8210 Try Solaris as #exclude syntax.
8211 * configure: Regenerate.
8212 * config.in: Regenerate.
8213 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
8215 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
8216 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
8218 * varasm.c (default_elf_asm_named_section): Don't check if
8219 HAVE_GAS_SECTION_EXCLUDE is defined.
8221 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8223 * doc/md.texi: Update the documentation of the cond_* optabs
8224 to mention the new final operand. Fix GET_MODE_NUNITS call.
8225 Describe the scalar case too.
8226 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
8227 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
8229 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
8230 (get_conditional_internal_fn): Update comment.
8231 * tree-vect-loop.c (vectorizable_reduction): Pass the original
8232 accumulator value as a final argument to conditional functions.
8233 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
8234 a define_expand and add an "else" operand. Assert for now that
8235 the else operand is equal to operand 2. Use SVE_INT_BINARY and
8236 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
8237 (*cond_<optab><mode>): New patterns.
8238 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
8239 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8240 (UNSPEC_COND_EOR): Delete.
8241 (optab): Remove associated mappings.
8242 (SVE_INT_BINARY): New code iterator.
8243 (sve_int_op): Remove int attribute and add "minus" to the code
8245 (SVE_COND_INT_OP): Delete.
8246 (SVE_COND_FP_OP): Rename to...
8247 (SVE_COND_FP_BINARY): ...this.
8249 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
8251 * optabs.c (can_reuse_operands_p): New function.
8252 (maybe_legitimize_operands): Try to reuse the results for
8255 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8257 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
8258 Add {q} suffix to insn mnemonic.
8260 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8262 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
8263 (msp430_warn_func_return): New.
8265 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
8267 * fold-const.c (tree_nonzero_bits): New function.
8268 * fold-const.h (tree_nonzero_bits): Likewise.
8269 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
8270 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
8272 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
8276 * varasm.c (assemble_alias): Check ifunc_resolver only on
8279 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
8282 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
8283 when memory input operand is handled.
8285 2018-05-24 Luis Machado <luis.machado@linaro.org>
8287 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
8289 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
8291 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
8293 * match.pd: Delay FMA folds until after vectorization.
8295 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
8298 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8299 address check not strict.
8301 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
8303 * gimple-match.h (gimple_match_op): New class.
8304 (mprts_hook): Replace parameters with a gimple_match_op *.
8305 (maybe_build_generic_op): Likewise.
8306 (gimple_simplified_result_is_gimple_val): Replace parameters with
8307 a const gimple_match_op *.
8308 (gimple_simplify): Replace code_helper * and tree * parameters with
8309 a gimple_match_op * parameter.
8310 (gimple_resimplify1): Replace code_helper *, tree and tree *
8311 parameters with a gimple_match_op * parameter.
8312 (gimple_resimplify2): Likewise.
8313 (gimple_resimplify3): Likewise.
8314 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
8315 parameters with a gimple_match_op * parameter.
8316 * gimple-match-head.c (gimple_simplify): Change prototypes of
8317 auto-generated functions to take a gimple_match_op * instead of
8318 separate code_helper * and tree * parameters. Make the same
8319 change in the top-level overload and update calls to the
8320 gimple_resimplify routines. Update calls to the auto-generated
8321 functions and to maybe_push_res_to_seq in the publicly-facing
8322 operation-specific gimple_simplify overloads.
8323 (gimple_match_op::MAX_NUM_OPS): Define.
8324 (gimple_resimplify1): Replace rcode and ops with a single res_op
8325 parameter. Update call to gimple_simplify.
8326 (gimple_resimplify2): Likewise.
8327 (gimple_resimplify3): Likewise.
8328 (mprts_hook): Replace parameters with a gimple_match_op *.
8329 (maybe_build_generic_op): Likewise.
8330 (build_call_internal): Replace type, nargs and ops with
8331 a gimple_match_op *.
8332 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
8333 with a single gimple_match_op *. Update calls to mprts_hook,
8334 build_call_internal and gimple_simplified_result_is_gimple_val.
8335 Factor out code that is common to the tree_code and combined_fn cases.
8336 * genmatch.c (expr::gen_transform): Replace tem_code and
8337 tem_ops with a gimple_match_op called tem_op. Update calls
8338 to the gimple_resimplify functions and maybe_push_res_to_seq.
8339 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
8340 res_ops. Update call to the gimple_resimplify functions.
8341 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
8342 (decision_tree::gen): Make the functions take a gimple_match_op *
8343 called res_op instead of separate res_code and res_ops parameters.
8344 Update call accordingly.
8345 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
8346 and ops with a single res_op parameter. Update calls to
8347 maybe_build_generic_op and maybe_push_res_to_seq.
8348 (fold_stmt_1): Update calls to gimple_simplify and
8349 replace_stmt_with_simplification.
8350 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
8351 and gimple_simplified_result_is_gimple_val.
8352 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
8354 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
8355 with a gimple_match_op *.
8356 (vn_nary_build_or_lookup): Likewise. Update call to
8357 vn_nary_build_or_lookup_1.
8358 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
8359 gimple_match_op *. Update calls to the gimple_resimplify routines
8360 and to gimple_simplified_result_is_gimple_val.
8361 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
8362 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
8363 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
8364 (visit_nary_op): Likewise.
8365 (visit_reference_op_load): Likewise.
8367 2018-05-23 Luis Machado <luis.machado@linaro.org>
8369 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
8370 modifier for printing the step amount.
8372 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
8375 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
8378 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
8380 * doc/sourcebuild.texi (Endianness): New subsubsection.
8382 2018-05-23 Luis Machado <luis.machado@linaro.org>
8384 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8385 <prefetch_dynamic_strides>: New const bool field.
8386 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8387 prefetch_dynamic_strides.
8388 (exynosm1_prefetch_tune): Likewise.
8389 (thunderxt88_prefetch_tune): Likewise.
8390 (thunderx_prefetch_tune): Likewise.
8391 (thunderx2t99_prefetch_tune): Likewise.
8392 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
8394 (aarch64_override_options_internal): Update to set
8395 PARAM_PREFETCH_DYNAMIC_STRIDES.
8396 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8397 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8398 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8399 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8400 prefetch-dynamic-strides setting.
8402 2018-05-23 Luis Machado <luis.machado@linaro.org>
8404 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8405 <minimum_stride>: New const int field.
8406 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8407 minimum_stride field defaulting to -1.
8408 (exynosm1_prefetch_tune): Likewise.
8409 (thunderxt88_prefetch_tune): Likewise.
8410 (thunderx_prefetch_tune): Likewise.
8411 (thunderx2t99_prefetch_tune): Likewise.
8412 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
8413 <default_opt_level>: Set to 3.
8414 (aarch64_override_options_internal): Update to set
8415 PARAM_PREFETCH_MINIMUM_STRIDE.
8416 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8417 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8418 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8419 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8420 stride is constant and is below the minimum stride threshold.
8422 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8424 * config/arm/arm-cpus.in (mode26): Delete.
8425 (armv4): Delete mode26 reference.
8426 * config/arm/arm.c (arm_configure_build_target): Delete use of
8429 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8431 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
8433 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
8434 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
8436 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
8437 (floatunsdidf2): Ditto.
8439 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8441 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
8442 (fixuns_trunc<mode>si2_avx512f): Ditto.
8443 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
8444 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
8445 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
8447 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8449 PR rtl-optimization/79985
8450 * df-scan.c (df_insn_refs_collect): Remove special case for
8451 global registers and asm statements.
8453 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8455 * extend.texi (Global Register Variables): Rewrite the bullet list.
8456 Note that the register is available for allocation. Note that access
8457 via inline asm must use constraints. Add note about async-signal
8458 handlers. Remove paragraph about automagic register selection.
8460 2018-05-23 Richard Biener <rguenther@suse.de>
8462 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
8463 of fixed offset from memset VN.
8465 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
8467 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
8469 (alloc_cand_and_find_basis): Initialize first_interp field.
8470 (slsr_process_mul): Modify first_interp field.
8471 (slsr_process_add): Likewise.
8472 (slsr_process_cast): Modify first_interp field for each new
8474 (slsr_process_copy): Likewise.
8475 (dump_candidate): Dump first_interp field.
8476 (replace_mult_candidate): Process all interpretations, not just
8478 (replace_rhs_if_not_dup): Likewise.
8479 (replace_one_candidate): Likewise.
8481 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
8483 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
8485 (aarch64_needs_frame_chain): New function.
8486 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
8488 2018-05-23 Sudakshina Das <sudi.das@arm.com>
8491 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
8492 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
8493 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
8494 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
8495 as true for strict-align.
8496 (aarch64_can_inline_p): Perform checks even when callee has no
8497 attributes to check for strict alignment.
8498 * doc/extend.texi (AArch64 Function Attributes): Document
8500 * doc/invoke.texi: (AArch64 Options): Likewise.
8502 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
8504 PR tree-optimization/85853
8505 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
8506 the handling of the root of the node to...
8507 (vect_slp_analyze_node_operations_1): ...this new function,
8508 and run the whole thing with the child nodes' def types
8509 set according to their SLP node's def type.
8511 2018-05-23 Richard Biener <rguenther@suse.de>
8514 * tree-data-ref.c (create_runtime_alias_checks): Defer
8515 and ignore overflow warnings.
8517 2018-05-23 Yury Gribov <tetra2005@gmail.com>
8519 PR tree-optimization/85822
8520 * tree-vrp.c (is_masked_range_test): Fix handling of negative
8523 2018-05-23 Richard Biener <rguenther@suse.de>
8525 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8526 memset constants via native_interpret_expr.
8528 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
8531 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
8533 (cgraph_node::create_alias): Likewise.
8534 (cgraph_node::get_availability): Check ifunc_resolver instead
8535 of looking up ifunc attribute.
8536 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8537 * varasm.c (do_assemble_alias): Likewise.
8538 (assemble_alias): Likewise.
8539 (default_binds_local_p_3): Likewise.
8540 * cgraph.h (cgraph_node): Add ifunc_resolver.
8541 (cgraph_node::only_called_directly_or_aliased_p): Return false
8543 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
8545 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
8546 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
8547 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
8548 instead of looking up ifunc attribute.
8550 2018-05-22 Luis Machado <luis.machado@linaro.org>
8552 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
8554 2018-05-22 Martin Sebor <msebor@redhat.com>
8557 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
8558 only when expasion succeeds.
8559 (expand_builtin_strcmp): Same.
8560 (expand_builtin_strncmp): Same.
8562 2018-05-22 Martin Sebor <msebor@redhat.com>
8564 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
8566 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8567 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8569 * config/aarch64/aarch64-ldpstp.md: Replace uses of
8570 aarch64_mem_pair_operand with memory_operand and delete operand swapping
8572 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8573 Add check for legitimate_address.
8574 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
8575 (aarch64_swap_ldrstr_operands): New.
8576 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
8579 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8580 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8582 * config/aarch64/aarch64.md: New patterns to generate stp
8584 (store_pair_sw, store_pair_dw): New patterns to generate stp for
8585 single words and double words.
8586 (load_pair_sw, load_pair_dw): Likewise.
8587 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
8589 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
8591 * config/aarch64/aarch64-ldpstp.md: Modify peephole
8592 for different mode ldpstp and add peephole for merged zero stores.
8594 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8596 (aarch64_gen_store_pair): Rename calls to match new patterns.
8597 (aarch64_gen_load_pair): Rename calls to match new patterns.
8598 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
8599 (load_pair<DREG:mode><DREG2:mode>): ... This.
8600 (store_pair<mode>): Rename to...
8601 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
8602 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
8604 (V_INT_EQUIV): Handle SImode.
8605 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
8608 2018-05-22 Martin Sebor <msebor@redhat.com>
8611 * calls.c (maybe_warn_nonstring_arg): Use string length to set
8612 or ajust the presumed bound on an operation to avoid unnecessary
8615 2018-05-22 Martin Sebor <msebor@redhat.com>
8617 PR tree-optimization/85826
8618 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
8619 assuming that a DECL necesarily has a constant size.
8621 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
8624 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
8626 2018-05-22 Richard Biener <rguenther@suse.de>
8628 PR tree-optimization/85834
8629 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
8630 non-constant and non-zero memset arguments.
8632 2018-05-22 Martin Liska <mliska@suse.cz>
8635 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
8637 2018-05-22 Richard Biener <rguenther@suse.de>
8639 PR tree-optimization/85863
8640 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
8641 comparisons when vectype is specified.
8642 (vectorizable_condition): Do not specify vectype for
8643 vect_is_simple_cond when SLP vectorizing.
8645 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
8648 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
8649 define __ibm128 as long double.
8650 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
8651 as a distinct type when IEEE 128-bit support is enabled.
8652 (init_float128_ieee): Fix up conversions between IFmode and IEEE
8653 128-bit types to use the correct functions.
8654 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
8655 convert between 128-bit floating point types that have different
8656 modes but the same representation, instead of using gen_lowpart to
8658 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
8660 (IFKF_reg): New attributes to give the register constraints for
8662 (extend<mode>tf2_internal): New insns to mark an explicit
8663 conversion between 128-bit floating point types that have a
8664 different mode but share the same representation.
8666 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
8668 PR tree-optimization/85814
8669 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
8670 a null return from get_strinfo when unsharing the next
8671 strinfo in the chain.
8673 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8676 * varasm.c (weak_finish): Clean up weak_decls.
8678 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8680 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
8681 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
8682 UNSPEC_UADALP values.
8683 * config/aarch64/iterators.md (ABAL): New int iterator.
8686 (sur): Add mappings for the above.
8687 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
8689 (aarch64_<sur>abal<mode>_4): Likewise.
8690 (aarch64_<sur>adalp<mode>_3): Likewise.
8691 (<sur>sadv16qi): New define_expand.
8693 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
8695 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
8696 (*movdf_internal): Ditto.
8697 (*rcpsf2_sse): Ditto.
8698 (*rsqrtsf2_sse): Ditto.
8699 (*sqrt<mode>2_sse): Ditto.
8701 2018-05-21 Tamar Christina <tamar.christina@arm.com>
8703 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
8705 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
8706 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
8707 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
8708 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
8710 * config/aarch64/arm_neon.h: Likewise.
8711 * config/aarch64/iterators.md (VQ_I): New.
8713 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
8715 * config.gcc: Add arc/t-multilib-linux to tmake_file for
8717 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
8720 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8722 * config/nds32/constraints.md (S): New constraint.
8723 * config/nds32/nds32.md (call_internal): Use constraint S.
8724 (call_value_internal): Likewise.
8725 (sibcall_internal): Likewise.
8726 (sibcall_value_internal): Likewise.
8728 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8729 Chung-Ju Wu <jasonwucj@gmail.com>
8731 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
8734 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8735 Chung-Ju Wu <jasonwucj@gmail.com>
8737 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
8738 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
8739 (nds32_rtx_costs_impl): Simplify.
8740 (nds32_address_cost_impl): Simplify.
8741 (nds32_init_rtx_costs): New function.
8742 (nds32_rtx_costs_speed_prefer): Likewise.
8743 (nds32_rtx_costs_size_prefer): Likewise.
8744 (nds32_address_cost_speed_prefer): Likewise.
8745 (nds32_address_cost_speed_fwprop): Likewise.
8746 (nds32_address_cost_size_prefer): Likewise.
8747 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
8748 * config/nds32/nds32.c (nds32_option_override): Use
8749 nds32_init_rtx_costs function.
8751 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8753 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
8754 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
8755 (TARGET_PIPELINE_N8): Likewise.
8756 (TARGET_PIPELINE_N10): Likewise.
8757 (TARGET_PIPELINE_N13): Likewise.
8758 (TARGET_PIPELINE_GRAYWOLF): Likewise.
8760 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
8762 * config/nds32/nds32-fpu.md: Update copyright year.
8764 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8766 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
8768 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8770 * config/nds32/nds32.c
8771 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
8772 * config/nds32/nds32.opt (minline-asm-r15): New option.
8774 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8776 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
8778 * config/nds32/nds32.md (abssi2): New pattern.
8780 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
8782 * config/i386/i386.md (rex64namesuffix): New mode attribute.
8783 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
8784 Merge insn pattern from sse_cvtsi2ss<round_name> and
8785 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
8786 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
8787 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
8788 using SWI48 mode iterator.
8789 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
8790 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
8791 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
8792 pattern from sse_cvttss2si<round_saeonly_name>
8793 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
8794 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
8795 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
8796 using SWI48 mode iterator.
8797 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
8798 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
8799 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
8800 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
8801 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
8802 using SWI48 mode iterator.
8803 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
8804 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
8805 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
8806 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
8807 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
8808 SWI48 mode iterator.
8809 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
8810 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
8811 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
8812 pattern from sse_cvttsd2si<round_saeonly_name>
8813 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
8815 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8817 * config/nds32/nds32-md-auxiliary.c
8818 (nds32_valid_smw_lwm_base_p): Refine.
8819 (nds32_output_smw_single_word): Refine.
8820 (nds32_output_smw_double_word): New.
8821 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
8823 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8825 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
8826 (nds32_output_stack_pop): Refine.
8827 (nds32_expand_unaligned_load): Refine.
8828 (nds32_expand_unaligned_store): Refine.
8830 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8831 Chung-Ju Wu <jasonwucj@gmail.com>
8833 * config/nds32/constants.md: Add TP_REGNUM constant.
8834 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
8835 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
8837 * config/nds32/nds32-doubleword.md: Consider flag_pic.
8838 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
8839 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
8840 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
8841 and PIC code generation.
8842 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
8844 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
8846 * config/nds32/nds32.md: Support TLS and PIC.
8847 * config/nds32/nds32.c: Support TLS and PIC.
8848 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
8849 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
8852 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8854 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
8855 mode with E_ prefix.
8857 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8858 Chung-Ju Wu <jasonwucj@gmail.com>
8860 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
8861 * config/nds32/nds32-md-auxiliary.c
8862 (symbolic_reference_mentioned_p): New.
8863 (nds32_legitimize_ict_address): New.
8864 (nds32_expand_ict_move): New.
8865 (nds32_indirect_call_referenced_p): New.
8866 (nds32_symbol_binds_local_p): Delete.
8867 (nds32_long_call_p): Modify.
8868 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
8869 * config/nds32/nds32-protos.h
8870 (symbolic_reference_mentioned_p): Declare.
8871 (nds32_legitimize_ict_address): Declare.
8872 (nds32_expand_ict_move): Declare.
8873 (nds32_indirect_call_referenced_p): Declare.
8874 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
8875 (nds32_relax_group): Use nds32_ict_const_p as condition.
8876 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
8877 (nds32_asm_file_start): Output ict_model directive in asm code.
8878 (nds32_legitimate_address_p): Consider indirect call.
8879 (nds32_print_operand): Consider indirect call.
8880 (nds32_print_operand_address): Consider indirect call.
8881 (nds32_insert_attributes): Handle "indirect_call" attribute.
8882 (TARGET_LEGITIMATE_ADDRESS_P): Define.
8883 (TARGET_LEGITIMATE_CONSTANT_P): Define.
8884 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
8885 (TARGET_DELEGITIMIZE_ADDRESS): Define.
8886 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
8887 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
8888 (TARGET_ICT_MODEL_SMALL): Define.
8889 (TARGET_ICT_MODEL_LARGE): Define.
8890 * config/nds32/nds32.md (movsi): Consider ict model.
8891 (call, call_value): Consider ict model.
8892 (sibcall, sibcall_value): Consider ict model.
8893 * config/nds32/nds32.opt (mict-model): New option.
8894 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
8897 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
8898 Monk Chiang <sh.chiang04@gmail.com>
8899 Jim Wilson <jimw@sifive.com>
8901 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
8902 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
8903 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
8904 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
8905 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
8906 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
8907 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
8908 compute save_libcall_adjustment properly.
8909 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
8910 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
8911 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
8912 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
8913 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
8914 (ABI_SPEC): Handle mabi=ilp32e.
8915 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
8916 (RVE): Add RVE mask.
8917 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
8918 <-march>: Add rv32e as an example.
8920 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
8923 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
8924 (intra_create_variable_infos): Handle C++ constructors.
8926 2018-05-18 Martin Liska <mliska@suse.cz>
8928 * passes.def: Remove a redundant pass.
8930 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8933 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
8935 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8937 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
8939 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
8941 (armv2, armv2a, armv3, armv3m): Delete architectures.
8942 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
8943 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
8944 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
8946 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
8947 (*mulsidi3adddi): Likewise.
8948 (mulsidi3): Likewise.
8949 (*mulsidi3_nov6): Likewise.
8950 (umulsidi3): Likewise.
8951 (umulsidi3_nov6): Likewise.
8952 (umaddsidi4): Likewise.
8953 (*umulsidi3adddi): Likewise.
8954 (smulsi3_highpart): Likewise.
8955 (*smulsi3_highpart_nov6): Likewise.
8956 (umulsi3_highpart): Likewise.
8957 (*umulsi3_highpart_nov6): Likewise.
8958 * config/arm/arm.h (arm_arch3m): Delete.
8959 * config/arm/arm.c (arm_arch3m): Delete.
8960 (arm_option_override_internal): Update armv3-related comment.
8961 (arm_configure_build_target): Delete use of isa_bit_mode32.
8962 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
8963 (arm_rtx_costs_internal): Delete check of arm_arch3m.
8964 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
8966 (mulusa3): Likewise.
8967 * config/arm/arm-protos.h (arm_arch3m): Delete.
8968 * config/arm/arm-tables.opt: Regenerate.
8969 * config/arm/arm-tune.md: Likewise.
8970 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
8971 deleted architectures.
8973 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8975 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
8976 (armv5t, armv5te): New features.
8977 (ARMv5, ARMv5e): Delete fgroups.
8978 (ARMv5t, ARMv5te): Adjust for above changes.
8980 (armv5, armv5e): Delete arches.
8981 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
8983 (*call_reg_arm): Likewise.
8984 (*call_value_reg_armv5): Likewise.
8985 (*call_value_reg_arm): Likewise.
8986 (*call_symbol): Likewise.
8987 (*call_value_symbol): Likewise.
8988 (*sibcall_insn): Likewise.
8989 (*sibcall_value_insn): Likewise.
8991 (prefetch): Likewise.
8992 (define_split and define_peephole2 dependent on arm_arch5):
8994 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
8996 (TARGET_ARM_QBIT): Likewise.
8997 (TARGET_DSP_MULTIPLY): Likewise.
8998 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
8999 (arm_arch5, arm_arch5e): Delete.
9000 (arm_arch5t, arm_arch5te): Declare.
9001 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
9002 (arm_arch5t): Declare.
9003 (arm_option_reconfigure_globals): Update for the above.
9004 (arm_options_perform_arch_sanity_checks): Update comment, replace
9005 use of arm_arch5 with arm_arch5t.
9006 (use_return_insn): Likewise.
9007 (arm_emit_call_insn): Likewise.
9008 (output_return_instruction): Likewise.
9009 (arm_final_prescan_insn): Likewise.
9010 (arm_coproc_builtin_available): Likewise.
9011 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
9012 arm_arch5e with arm_arch5t and arm_arch5te.
9013 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
9014 (arm_arch5t, arm_arch5te): Declare.
9015 * config/arm/arm-tables.opt: Regenerate.
9016 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
9017 * config/arm/t-multilib: Likewise.
9018 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
9019 instead of arm_arch5.
9020 (*call_reg_thumb1): Likewise.
9021 (*call_value_reg_thumb1_v5): Likewise.
9022 (*call_value_reg_thumb1): Likewise.
9023 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
9025 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
9027 2018-05-18 Martin Liska <mliska@suse.cz>
9029 PR gcov-profile/84846
9030 * doc/gcov.texi: Document -t option of gcov tool.
9032 2018-05-18 Martin Liska <mliska@suse.cz>
9034 PR gcov-profile/84846
9035 * gcov.c (print_usage): Add new -t option.
9036 (process_args): Handle the option.
9037 (generate_results): Use stdout as output when requested by
9040 2018-05-18 Martin Liska <mliska@suse.cz>
9042 PR gcov-profile/84846
9043 * coverage.c (coverage_init): Write PWD to .gcno file.
9044 * doc/gcov.texi: Document how working directory is printed.
9045 * gcov-dump.c (dump_gcov_file): Print PWD.
9046 * gcov.c (output_intermediate_file): Likewise.
9047 (read_graph_file): Read PWD string.
9048 (output_lines): Print PWD.
9050 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9053 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
9054 for retval and return false if all args to phi are zero.
9056 2018-05-18 Richard Biener <rguenther@suse.de>
9058 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
9060 (evrp_dom_walker::before_dom_children): Call it.
9062 2018-05-18 Richard Biener <rguenther@suse.de>
9064 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
9065 results when processing array refs with variable index.
9067 2018-05-18 Toon Moene <toon@moene.org>
9069 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9070 directly after that of -floop-interchange. Indicate that both
9071 options are enabled by default when specifying -O3.
9073 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9075 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
9076 iterator. Delete separate integer-mode vec_set<mode> expander.
9077 (aarch64_simd_vec_setv2di): Delete.
9078 (vec_setv2di): Delete.
9079 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
9080 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
9081 the "w, r" alternative.
9083 2018-05-18 Martin Liska <mliska@suse.cz>
9085 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
9086 * tree-pass.h (make_pass_lower_switch_O0): New function.
9087 * tree-switch-conversion.c (node_has_low_bound): Remove.
9088 (node_has_high_bound): Likewise.
9089 (node_is_bounded): Likewise.
9090 (class pass_lower_switch): Make it a template type and create
9092 (pass_lower_switch::execute): Add template argument.
9093 (make_pass_lower_switch): New function.
9094 (make_pass_lower_switch_O0): New function.
9095 (do_jump_if_equal): Remove.
9096 (emit_case_nodes): Simplify to just handle all 3 cases and leave
9097 all the hard work to tree optimization passes.
9099 2018-05-18 Martin Liska <mliska@suse.cz>
9101 * dbgcnt.c (limit_low): Renamed from limit.
9102 (limit_high): New variable.
9103 (dbg_cnt_is_enabled): Check for upper limit.
9104 (dbg_cnt): Adjust dumping.
9105 (dbg_cnt_set_limit_by_index): Add new argument for high
9107 (dbg_cnt_set_limit_by_name): Likewise.
9108 (dbg_cnt_process_single_pair): Parse new format.
9109 (dbg_cnt_process_opt): Use strtok.
9110 (dbg_cnt_list_all_counters): Remove 'value' and add
9112 * doc/invoke.texi: Document changes.
9114 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
9116 * doc/sourcebuild.texi (scalar_all_fma): Document.
9117 * tree.def (FMA_EXPR): Delete.
9118 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
9119 * internal-fn.c (ternary_direct): New macro.
9120 (expand_ternary_optab_fn): Likewise.
9121 (direct_ternary_optab_supported_p): Likewise.
9122 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
9123 * builtins.c (fold_builtin_fma): Delete.
9124 (fold_builtin_3): Don't call it.
9125 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
9126 * expr.c (expand_expr_real_2): Likewise.
9127 * fold-const.c (operand_equal_p): Likewise.
9128 (fold_ternary_loc): Likewise.
9129 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
9130 * gimple.c (DEFTREECODE): Likewise.
9131 * gimplify.c (gimplify_expr): Likewise.
9132 * optabs-tree.c (optab_for_tree_code): Likewise.
9133 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9134 * tree-eh.c (operation_could_trap_p): Likewise.
9135 (stmt_could_throw_1_p): Likewise.
9136 * tree-inline.c (estimate_operator_cost): Likewise.
9137 * tree-pretty-print.c (dump_generic_node): Likewise.
9138 (op_code_prio): Likewise.
9139 * tree-ssa-loop-im.c (stmt_cost): Likewise.
9140 * tree-ssa-operands.c (get_expr_operands): Likewise.
9141 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
9142 * fold-const-call.h (fold_fma): Delete.
9143 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
9144 CFN_FNMA and CFN_FNMS.
9146 * genmatch.c (combined_fn): New enum.
9147 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
9148 (commutative_op): New function.
9149 (commutate): Use it. Handle more than 2 operands.
9150 (dt_operand::gen_gimple_expr): Use commutative_op.
9151 (parser::parse_expr): Allow :c to be used with non-binary
9152 operators if the commutative operand is known.
9153 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
9154 CFN_FMS, CFN_FNMA and CFN_FNMS.
9155 (backprop::process_assign_use): Remove FMA_EXPR handling.
9156 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
9157 (gen_hsa_fma): New function.
9158 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
9159 IFN_FNMA and IFN_FNMS.
9160 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
9161 * gimple-fold.h (follow_all_ssa_edges): Declare.
9162 * gimple-fold.c (follow_all_ssa_edges): New function.
9163 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
9164 gimple_build interface and use follow_all_ssa_edges to fold the result.
9165 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
9166 instead of checking for optabs directly.
9167 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
9168 rather than FMA_EXPRs.
9169 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
9170 call to IFN_FMA instead of an FMA_EXPR.
9172 2018-05-17 Jim Wilson <jimw@sifive.com>
9174 * expr.c (do_tablejump): When converting index to Pmode, if we have a
9175 sign extended promoted subreg, and the range does not have the sign bit
9176 set, then do a sign extend.
9178 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
9179 test, check for sign extended subreg and/or constant operands, and
9180 do a sign extend in that case.
9182 2018-05-17 Steve Ellcey <sellcey@cavium.com>
9184 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
9185 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
9187 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
9188 Change logics_shift_reg to logics_shift_imm.
9189 (thunderx2t99_fp_loadpair_basic): Delete.
9190 (thunderx2t99_fp_storepair_basic): Delete.
9191 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
9192 (thunderx2t99_asimd_polynomial): Delete.
9193 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
9194 and neon_fp_mul_d_scalar_q.
9195 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
9196 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
9197 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
9198 (thunderx2t99_asimd_lut): Add missing tbl types.
9199 (thunderx2t99_asimd_ext): Delete.
9200 (thunderx2t99_asimd_load1_1_mult): Delete.
9201 (thunderx2t99_asimd_load1_2_mult): Delete.
9202 (thunderx2t99_asimd_load1_ldp): New.
9203 (thunderx2t99_asimd_load1): New.
9204 (thunderx2t99_asimd_load2): Add missing *load2* types.
9205 (thunderx2t99_asimd_load3): New.
9206 (thunderx2t99_asimd_load4): New.
9207 (thunderx2t99_asimd_store1_1_mult): Delete.
9208 (thunderx2t99_asimd_store1_2_mult): Delete.
9209 (thunderx2t99_asimd_store2_mult): Delete.
9210 (thunderx2t99_asimd_store2_onelane): Delete.
9211 (thunderx2t99_asimd_store_stp): New.
9212 (thunderx2t99_asimd_store1): New.
9213 (thunderx2t99_asimd_store2): New.
9214 (thunderx2t99_asimd_store3): New.
9215 (thunderx2t99_asimd_store4): New.
9217 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
9219 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
9220 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
9222 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
9223 Segher Boessenkool <segher@kernel.crashing.org>
9226 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
9229 2018-05-17 Richard Biener <rguenther@suse.de>
9231 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
9232 for pruning loop and prune defs feeding only already visited PHIs.
9234 2018-05-17 Richard Biener <rguenther@suse.de>
9236 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
9238 2018-05-17 Bin Cheng <bin.cheng@arm.com>
9239 Richard Biener <rguenther@suse.de>
9241 PR tree-optimization/85793
9242 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
9243 for VMAT_ELEMENTWISE.
9245 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9247 * internal-fn.h (lookup_internal_fn): Declare
9248 * internal-fn.c (lookup_internal_fn): New function.
9249 * gimple.c (gimple_build_call_from_tree): Handle calls to
9251 * gimple-pretty-print.c (dump_gimple_call): Print "." before
9252 internal function names.
9253 * tree-pretty-print.c (dump_generic_node): Likewise.
9254 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
9256 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
9258 * gimple-fold.h (gimple_build): Make the function forms take
9259 combined_fn rather than built_in_function.
9260 (gimple_simplify): Likewise.
9261 * gimple-match-head.c (gimple_simplify): Likewise.
9262 * gimple-fold.c (gimple_build): Likewise.
9263 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
9264 rather than gimple_build_call_internal.
9265 (get_initial_defs_for_reduction): Likewise.
9266 (vect_create_epilog_for_reduction): Likewise.
9267 (vectorizable_live_operation): Likewise.
9269 2018-05-17 Martin Liska <mliska@suse.cz>
9271 * gimple-ssa-sprintf.c (format_directive): Do not use
9272 space in between 'G_' and '('.
9274 2018-05-17 Jakub Jelinek <jakub@redhat.com>
9277 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
9278 even if the mask is not all ones.
9281 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
9283 (ix86_gimple_fold_builtin): Likewise.
9286 * config/i386/i386.c: Include tree-vector-builder.h.
9287 (ix86_vector_shift_count): New function.
9288 (ix86_fold_builtin): Fold shift builtins by scalar count.
9289 (ix86_gimple_fold_builtin): Likewise.
9291 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
9292 _mm512_setzero): New intrinsics.
9294 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
9295 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9297 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
9298 code generation for cases where splatting a value is not useful.
9299 * simplify-rtx.c (simplify_ternary_operation): Simplify
9300 vec_merge across a vec_duplicate and a paradoxical subreg forming
9301 a vector mode to a vec_concat.
9303 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
9305 * config.gcc: Support "goldmont-plus".
9306 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
9308 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9309 PROCESSOR_GOLDMONT_PLUS.
9310 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
9311 (processor_target_table): Add "goldmont-plus".
9312 (PTA_GOLDMONT_PLUS): Define.
9313 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
9314 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
9315 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
9316 (fold_builtin_cpu): Add "goldmont-plus".
9317 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
9318 (ix86_option_override_internal): Add "goldmont-plus".
9319 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
9320 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
9321 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
9322 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
9324 2018-05-17 Richard Biener <rguenther@suse.de>
9326 PR tree-optimization/85757
9327 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
9328 remove defs that only feed that PHI from further processing.
9330 2018-05-16 Jim Wilson <jimw@sifive.com>
9332 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
9334 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
9336 2018-05-16 Mark Wielaard <mark@klomp.org>
9338 * dwarf2out.c (count_index_strings): New function.
9339 (output_indirect_strings): Call count_index_strings and generate
9340 header for dwarf_version >= 5.
9342 2018-05-16 Mark Wielaard <mark@klomp.org>
9344 * dwarf2out.c (dwarf_FORM): New function.
9345 (set_indirect_string): Use dwarf_FORM.
9346 (reset_indirect_string): Likewise.
9347 (size_of_die): Likewise.
9348 (value_format): Likewise.
9349 (output_die): Likewise.
9350 (add_skeleton_AT_string): Likewise.
9351 (output_macinfo_op): Likewise.
9352 (index_string): Likewise.
9353 (output_index_string_offset): Likewise.
9354 (output_index_string): Likewise.
9355 (count_index_strings): Likewise.
9357 2018-05-16 Carl Love <cel@us.ibm.com>
9359 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9360 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
9362 2018-05-16 Martin Jambor <mjambor@suse.cz>
9364 * ipa-prop.c (ipa_free_all_edge_args): Remove.
9365 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
9367 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
9369 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
9370 (fnma<mode>4): Likewise.
9371 (fms<mode>4): Likewise.
9372 (fnms<mode>4): Likewise.
9373 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
9374 (aarch64_fnma<mode>4): Likewise.
9375 (aarch64_fms<mode>4): Likewise.
9376 (aarch64_fnms<mode>4): Likewise.
9377 (aarch64_fnmadd<mode>4): Likewise.
9379 2018-05-16 Jason Merrill <jason@redhat.com>
9381 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
9383 2018-05-16 Richard Biener <rguenther@suse.de>
9385 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
9386 (dump_stmt_cost): Declare.
9387 (add_stmt_cost): Dump cost we add.
9388 (add_stmt_costs): New function.
9389 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
9391 (vect_analyze_stmt): Adjust prototype.
9392 (vectorizable_condition): Likewise.
9393 (vectorizable_live_operation): Likewise.
9394 (vectorizable_reduction): Likewise.
9395 (vectorizable_induction): Likewise.
9396 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
9397 cost vector to pass to vectorizable_ and record afterwards.
9398 (vect_model_reduction_cost): Take cost vector argument and adjust.
9399 (vect_model_induction_cost): Likewise.
9400 (vectorizable_reduction): Likewise.
9401 (vectorizable_induction): Likewise.
9402 (vectorizable_live_operation): Likewise.
9403 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
9404 SLP_TREE_NUMBER_OF_VEC_STMTS.
9405 (vect_analyze_slp_cost_1): Remove.
9406 (vect_analyze_slp_cost): Likewise.
9407 (vect_slp_analyze_node_operations): Take visited args and
9408 a target cost vector. Avoid processing already visited stmt sets.
9409 (vect_slp_analyze_operations): Use a local cost vector to gather
9410 costs and register those of non-discarded instances.
9411 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
9412 (vect_schedule_slp_instance): Remove copying of
9413 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
9415 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
9416 adding cost. Record cost entry location.
9417 (vect_prologue_cost_for_slp_op): Function to compute cost of
9418 a constant or invariant generated for SLP vect in the prologue,
9419 split out from vect_analyze_slp_cost_1.
9420 (vect_model_simple_cost): Make static. Adjust for SLP costing.
9421 (vect_model_promotion_demotion_cost): Likewise.
9422 (vect_model_store_cost): Likewise, make static.
9423 (vect_model_load_cost): Likewise.
9424 (vectorizable_bswap): Add cost vector arg and adjust.
9425 (vectorizable_call): Likewise.
9426 (vectorizable_simd_clone_call): Likewise.
9427 (vectorizable_conversion): Likewise.
9428 (vectorizable_assignment): Likewise.
9429 (vectorizable_shift): Likewise.
9430 (vectorizable_operation): Likewise.
9431 (vectorizable_store): Likewise.
9432 (vectorizable_load): Likewise.
9433 (vectorizable_condition): Likewise.
9434 (vectorizable_comparison): Likewise.
9435 (can_vectorize_live_stmts): Likewise.
9436 (vect_analyze_stmt): Likewise.
9437 (vect_transform_stmt): Adjust calls to vectorizable_*.
9438 * tree-vectorizer.c: Include gimple-pretty-print.h.
9439 (dump_stmt_cost): New function.
9441 2018-05-16 Richard Biener <rguenther@suse.de>
9443 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
9444 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
9445 * tree-ssa-dse.c: Include tree-ssa-loop.h.
9446 (check_name): New callback.
9447 (dse_classify_store): Track cycles via a visited bitmap of PHI
9448 defs and simplify handling of in-loop and across loop dead stores
9449 and properly fail for loop-variant refs. Handle byte-tracking with
9450 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
9453 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
9455 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
9456 (vect_get_mask_type_for_stmt): Likewise.
9457 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
9459 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
9460 to determine the statement's vector type and the vector type that
9461 should be used for calculating nunits. Deal with cases in which
9462 the type has to be deferred.
9463 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
9464 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
9465 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
9466 (vect_determine_vf_for_stmt): New functions, split out from...
9467 (vect_determine_vectorization_factor): ...here.
9468 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
9469 (vect_get_mask_type_for_stmt): New functions, split out from
9470 vect_determine_vectorization_factor.
9472 2018-05-16 Richard Biener <rguenther@suse.de>
9474 * tree-cfg.c (verify_gimple_assign_ternary): Properly
9475 verify the [VEC_]COND_EXPR embedded comparison.
9477 2018-05-15 Martin Sebor <msebor@redhat.com>
9479 PR tree-optimization/85753
9480 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
9481 RECORD_TYPE in addition to ARRAY_TYPE.
9483 2018-05-15 Martin Sebor <msebor@redhat.com>
9486 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
9488 2018-05-15 Richard Biener <rguenther@suse.de>
9490 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
9491 add by_clobber_p one. Change algorithm to collect all defs
9492 representing uses we need to walk and try reducing them to
9493 a single one before failing.
9494 (dse_dom_walker::dse_optimize_stmt): Adjust.
9496 2018-05-13 Mark Wielaard <mark@klomp.org>
9498 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
9499 (size_of_loc_descr): Likewise.
9500 (output_loc_operands): Likewise.
9501 (output_loc_operands_raw): Likewise.
9502 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
9503 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
9504 (hash_loc_operands): Likewise.
9505 (compare_loc_operands): Likewise.
9507 2018-05-14 Mark Wielaard <mark@klomp.org>
9509 * dwarf2out.c (count_index_addrs): New function.
9510 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
9512 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9514 PR tree-optimization/83648
9515 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
9516 return value as malloc candidate.
9518 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9521 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
9522 param as true in call to suggest_attribute.
9524 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
9526 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
9527 -mreadonly-in-sdata.
9529 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9531 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
9533 (aarch64_crypto_aesd_fused): Likewise.
9535 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
9537 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
9538 (movsi_aarch64): Likewise.
9539 (load_pairsi): Likewise.
9540 (load_pairdi): Likewise.
9541 (store_pairsi): Likewise.
9542 (store_pairdi): Likewise.
9543 (load_pairsf): Likewise.
9544 (load_pairdf): Likewise.
9545 (store_pairsf): Likewise.
9546 (store_pairdf): Likewise.
9547 (zero_extend): Likewise.
9548 (trunc): Swap alternatives.
9549 (fcvt_target): Add '?' to prefer w over r.
9551 2018-05-14 Jakub Jelinek <jakub@redhat.com>
9554 * config/i386/i386.md: Disallow non-commutative arithmetics in
9555 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
9556 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
9557 in the peephole2 before it.
9559 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
9561 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
9562 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
9563 (ix86_handle_option): Handle -mcldemote.
9564 * config.gcc: New header.
9565 * config/i386/cldemoteintrin.h: New file.
9566 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
9567 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
9569 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9570 OPTION_MASK_ISA_CLDEMOTE.
9571 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
9572 (ix86_valid_target_attribute_inner_p): Ditto.
9573 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
9574 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
9575 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
9576 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
9577 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
9579 * config/i386/i386.opt: Add -mcldemote.
9580 * config/i386/x86intrin.h: New header.
9581 * doc/invoke.texi: Add -mcldemote.
9583 2018-05-14 Richard Biener <rguenther@suse.de>
9585 * doc/match-and-simplify.texi: Adjust :s documentation.
9587 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
9589 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
9590 intended memcpy size.
9591 (REORDER_45): Likewise.
9593 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
9595 * sort.cc: New file.
9596 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
9597 * vec.c (qsort_chk): Use gcc_qsort.
9598 * Makefile.in (OBJS-libcommon): Add sort.o.
9599 (build/sort.o): New target. Use it...
9600 (BUILD_RTL): ... here, and...
9601 (build/gencfn-macros): ... here, and...
9602 (build/genmatch): ... here.
9604 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9605 Chung-Ju Wu <jasonwucj@gmail.com>
9607 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
9608 * config/nds32/nds32-graywolf.md: New file.
9609 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
9610 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
9612 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
9613 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
9614 * config/nds32/nds32.md (pipeline_model): Add graywolf.
9615 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
9616 * config/nds32/pipelines.md: Include n15 settings.
9618 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9619 Chung-Ju Wu <jasonwucj@gmail.com>
9621 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
9622 * config/nds32/nds32-n13.md: New file.
9623 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
9624 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
9626 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
9627 * config/nds32/nds32.md (pipeline_model): Add n13.
9628 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
9629 * config/nds32/pipelines.md: Include n13 settings.
9631 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9632 Chung-Ju Wu <jasonwucj@gmail.com>
9634 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
9635 * config/nds32/nds32-n10.md: New file.
9636 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
9637 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
9639 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
9640 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
9641 * config/nds32/nds32.md (pipeline_model): Add n10.
9642 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
9643 * config/nds32/pipelines.md: Include n10 settings.
9645 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
9646 Kito Cheng <kito.cheng@gmail.com>
9647 Chung-Ju Wu <jasonwucj@gmail.com>
9649 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
9650 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9651 Add enum values for DSP extension instructions.
9652 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
9654 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
9655 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
9657 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
9658 * config/nds32/nds32-dspext.md: New file for DSP implementation.
9659 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
9660 * config/nds32/nds32-intrinsic.md: Likewise.
9661 * config/nds32/nds32_intrinsic.h: Likewise.
9662 * config/nds32/nds32-md-auxiliary.c: Likewise.
9663 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
9664 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
9665 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
9666 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
9667 * config/nds32/nds32-protos.h: New declarations for DSP extension.
9668 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
9669 TYPE_DMAC in switch statement.
9670 * config/nds32/nds32.c: New checking and implementation for DSP
9671 extension instructions.
9672 * config/nds32/nds32.h: Likewise.
9673 * config/nds32/nds32.md: Likewise.
9674 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
9675 * config/nds32/predicates.md: Implement new predicates for DSP
9678 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9680 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
9681 Reformat alternatives and attributes so it is easier to identify
9682 which constraints/attributes go with which instruction.
9683 (mov<mode>_hardfloat32, FMOVE64): Likewise.
9684 (mov<mode>_softfloat32, FMOVE64): Likewise.
9685 (mov<mode>_hardfloat64, FMOVE64): Likewise.
9686 (mov<mode>_softfloat64, FMOVE64): Likewise.
9688 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
9690 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9692 (Basic PowerPC Built-in Functions): The new name of the
9693 subsection previously known as "PowerPC Built-in Functions".
9694 (Basic PowerPC Built-in Functions Available on all Configurations):
9696 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
9697 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
9698 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
9699 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
9701 2018-05-11 Martin Jambor <mjambor@suse.cz>
9704 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
9707 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9710 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
9712 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
9714 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
9715 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
9716 (ix86_handle_option): Handle -mwaitpkg.
9717 * config.gcc: New header.
9718 * config/i386/cpuid.h (bit_WAITPKG): New bit.
9719 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
9720 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
9722 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9723 OPTION_MASK_ISA_WAITPKG.
9724 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
9725 (ix86_option_override_internal): Add PTA_WAITPKG.
9726 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
9727 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
9728 IX86_BUILTIN_TPAUSE.
9729 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
9730 __builtin_ia32_umwait and __builtin_ia32_tpause.
9731 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
9732 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
9733 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
9734 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
9735 UNSPECV_TPAUSE): New.
9736 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
9737 * config/i386/i386.opt: Add -mwaitpkg.
9738 * config/i386/waitpkgintrin.h: New file.
9739 * config/i386/x86intrin.h: New header.
9740 * doc/invoke.texi: Add -mwaitpkg.
9742 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9745 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
9747 (cortex-m0): Use armv6s-m isa.
9748 (cortex-m0plus): Likewise.
9749 (cortex-m1): Likewise.
9750 (cortex-m0.small-multiply): Likewise.
9751 (cortex-m0plus.small-multiply): Likewise.
9752 (cortex-m1.small-multiply): Likewise.
9754 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
9755 Jakub Jelinek <jakub@redhat.com>
9757 PR tree-optimization/85692
9758 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
9759 source permute as well.
9761 2018-05-11 Martin Liska <mliska@suse.cz>
9764 * doc/extend.texi: Document LLVM style format for no_sanitize
9767 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9769 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
9770 mode_supports_vsx_dform_quad to mode_supports_dq_form.
9771 (mode_supports_vsx_dform_quad): Likewise.
9772 (mode_supports_vmx_dform): Move these functions to be next to the
9773 other mode_supports functions.
9774 (mode_supports_dq_form): Likewise.
9775 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
9776 mode_supports_dq_form.
9777 (reg_offset_addressing_ok_p): Likewise.
9778 (offsettable_ok_by_alignment): Likewise.
9779 (rs6000_legitimate_offset_address_p): Likewise.
9780 (legitimate_lo_sum_address_p): Likewise.
9781 (rs6000_legitimize_address): Likewise.
9782 (rs6000_legitimize_reload_address): Likewise.
9783 (rs6000_secondary_reload_inner): Likewise.
9784 (rs6000_preferred_reload_class): Likewise.
9785 (rs6000_output_move_128bit): Likewise.
9787 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9789 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
9790 Generate SImode target register for null target.
9791 <case IX86_BUILTIN_XGETBV>: Ditto.
9792 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
9793 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
9795 2018-05-10 Carl Love <cel@us.ibm.com>
9797 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9798 dcbtt and dcbtstt if operands[2] is 0.
9800 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9803 * config/i386/sse.md (usadv64qi): New expander.
9805 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
9807 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
9808 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
9809 -maltivec=be support.
9810 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
9811 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
9812 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
9813 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
9814 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
9815 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
9816 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
9817 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
9818 altivec_vsumsws): Adjust.
9819 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
9820 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
9821 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
9823 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
9824 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
9825 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
9826 (altivec_lve<VI_char>x): Delete expand.
9827 (*altivec_lve<VI_char>x_internal): Rename to...
9828 (altivec_lve<VI_char>x): ... this.
9829 (altivec_lvxl_<mode>): Delete expand.
9830 (*altivec_lvxl_<mode>_internal): Rename to ...
9831 (altivec_lvxl_<mode>): ... this.
9832 (altivec_stvxl_<mode>): Delete expand.
9833 (*altivec_stvxl_<mode>_internal): Rename to ...
9834 (altivec_stvxl_<mode>): ... this.
9835 (altivec_stve<VI_char>x): Delete expand.
9836 (*altivec_stve<VI_char>x_internal): Rename to ...
9837 (altivec_stve<VI_char>x): ... this.
9838 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
9839 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
9840 reduc_plus_scal_<mode>): Adjust.
9841 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
9843 (rs6000_cpu_cpp_builtins): Adjust.
9844 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
9845 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
9846 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
9847 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
9848 -maltivec=be support.
9849 (rs6000_split_vec_extract_var): Adjust.
9850 (rs6000_split_v4si_init): Adjust.
9851 (swap_selector_for_mode): Delete.
9852 (altivec_expand_lvx_be, altivec_expand_stvx_be,
9853 altivec_expand_stvex_be): Delete.
9854 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
9855 -maltivec=be support.
9856 (rs6000_gimple_fold_builtin): Ditto.
9857 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
9859 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
9860 (TARGET_DIRECT_MOVE_64BIT): Adjust.
9861 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
9862 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
9863 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
9864 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
9865 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
9866 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
9867 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
9868 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
9869 anonymous split): Adjust.
9870 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
9871 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
9873 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
9875 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
9876 when --with-gxx-include-dir is also specified.
9877 * configure: Regenerate.
9879 2018-05-09 Jim Wilson <jimw@sifive.com>
9882 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
9883 * config/riscv/t-withmultilib: New.
9884 * config/riscv/withmultilib.h: New.
9885 * doc/install.texi: Document RISC-V --with-multilib-list support.
9887 2018-05-09 Richard Biener <rguenther@suse.de>
9889 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
9891 (vect_bb_vectorization_profitable_p): Adjust. Compute
9892 actual scalar cost using the cost vector and the add_stmt_cost
9895 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9897 PR rtl-optimization/85645
9898 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
9899 in the REG_CFA_REGISTER note for LR, don't leave it empty.
9901 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9903 PR rtl-optimization/85645
9904 * shrink-wrap.c (spread_components): Return a boolean saying if
9905 anything was changed.
9906 (try_shrink_wrapping_separate): Iterate spread_components until
9907 nothing changes anymore.
9909 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9911 PR rtl-optimization/85645
9912 * regrename.c (build_def_use): Also kill the chains that include the
9913 destination of a REG_CFA_REGISTER note.
9915 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9917 PR rtl-optimization/85645
9918 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
9919 insn that has a REG_CFA_REGISTER note.
9921 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
9923 * cfgexpand.c (expand_clobber): New function.
9924 (expand_gimple_stmt_1): Use it.
9925 * tree-vect-stmts.c (vect_clobber_variable): New function,
9927 (vectorizable_simd_clone_call): ...here.
9928 (vectorizable_store): Emit a clobber either side of an
9929 IFN_STORE_LANES sequence.
9930 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
9932 2018-05-09 Tom de Vries <tom@codesourcery.com>
9935 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
9936 (define_insn "trap_if_false"): Add exit after trap.
9938 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
9940 PR rtl-optimization/85638
9941 * bb-reorder.c: Include common/common-target.h.
9942 (create_forwarder_block): New function extracted from...
9943 (fix_up_crossing_landing_pad): ...here. Rename into...
9944 (dw2_fix_up_crossing_landing_pad): ...this.
9945 (sjlj_fix_up_crossing_landing_pad): New function.
9946 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
9947 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
9948 from both partitions and exit the loop after one iteration.
9950 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9953 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9955 (Basic PowerPC Built-in Functions): The new name of the
9956 subsection previously known as "PowerPC Built-in Functions".
9957 (Basic PowerPC Built-in Functions Available on all Configurations):
9959 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9961 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9963 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9965 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9968 2018-05-08 Jim Wilson <jimw@sifive.com>
9970 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
9971 (LD_EMUL_SUFFIX): New.
9972 (LINK_SPEC): Use it.
9974 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9976 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9978 (Basic PowerPC Built-in Functions): The new name of the
9979 subsection previously known as "PowerPC Built-in Functions".
9980 (Basic PowerPC Built-in Functions Available on all Configurations):
9982 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9984 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9986 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9988 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9991 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9994 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
9995 after cmpelim optimization.
9997 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
9999 * config.gcc: Support "goldmont".
10000 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
10001 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10002 PROCESSOR_GOLDMONT.
10003 * config/i386/i386.c (m_GOLDMONT): Define.
10004 (processor_target_table): Add "goldmont".
10005 (PTA_GOLDMONT): Define.
10006 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
10007 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
10008 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
10009 (fold_builtin_cpu): Add "goldmont".
10010 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
10011 (ix86_option_override_internal): Add "goldmont".
10012 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
10013 (processor_type): Add PROCESSOR_GOLDMONT.
10014 * config/i386/i386.md: Add CPU "glm".
10015 * config/i386/glm.md: New file.
10016 * config/i386/x86-tune.def: Add m_GOLDMONT.
10017 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
10019 2018-05-08 Jakub Jelinek <jakub@redhat.com>
10022 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
10024 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
10025 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
10026 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
10029 * config/i386/i386.c (ix86_fold_builtin): Handle
10030 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
10033 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
10034 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
10036 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
10039 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
10040 (check_arch): Likewise.
10041 (check_fpu): Return the result rather than printing it.
10042 (end arch): Fix operator precedence.
10043 (end cpu): Likewise.
10044 (END): Print the result from check_fpu.
10046 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10047 Alan Hayward <alan.hayward@arm.com>
10048 David Sherwood <david.sherwood@arm.com>
10050 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
10051 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
10052 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
10053 (*fcmuo<mode>_and): New patterns.
10055 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
10057 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
10058 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
10059 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
10060 (cmp_op, sve_imm_con): New code attributes.
10061 (SVE_COND_INT_CMP, imm_con): Delete.
10062 (cmp_op): Remove above unspecs from int attribute.
10063 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
10065 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
10066 comparison-specific unspecs.
10067 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
10068 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
10069 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
10070 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
10071 (*vec_fcm<cmp_op><mode>): Rename to...
10072 (*fcm<cmp_op><mode>): ...this and adjust likewise.
10073 (*vec_fcmuo<mode>): Rename to...
10074 (*fcmuo<mode>): ...this and adjust likewise.
10075 (*pred_fcm<cmp_op><mode>): New pattern.
10076 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
10077 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
10079 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
10081 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
10082 (aarch64_emit_sve_predicated_cond): New function.
10083 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
10084 (aarch64_emit_unspec_cond_or): Replace with...
10085 (aarch64_emit_sve_or_conds): ...this new function. Use
10086 aarch64_emit_sve_ptrue_op for the individual comparisons and
10087 aarch64_emit_binop to OR them together.
10088 (aarch64_emit_inverted_unspec_cond): Replace with...
10089 (aarch64_emit_sve_inverted_cond): ...this new function. Use
10090 aarch64_emit_sve_ptrue_op for the comparison and
10091 aarch64_emit_unop to invert the result.
10092 (aarch64_expand_sve_vec_cmp_float): Update after the above
10093 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
10095 2018-05-07 Nathan Sidwell <nathan@acm.org>
10097 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
10098 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
10099 (Backwards Compatibility): Likewise.
10101 2018-05-07 Luis Machado <luis.machado@linaro.org>
10105 2018-05-07 Luis Machado <luis.machado@linaro.org>
10107 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10108 <prefetch_dynamic_strides>: New const bool field.
10109 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10110 prefetch_dynamic_strides.
10111 (exynosm1_prefetch_tune): Likewise.
10112 (thunderxt88_prefetch_tune): Likewise.
10113 (thunderx_prefetch_tune): Likewise.
10114 (thunderx2t99_prefetch_tune): Likewise.
10115 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10117 (aarch64_override_options_internal): Update to set
10118 PARAM_PREFETCH_DYNAMIC_STRIDES.
10119 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10120 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10121 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10122 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10123 prefetch-dynamic-strides setting.
10125 2018-05-07 Luis Machado <luis.machado@linaro.org>
10127 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10128 <minimum_stride>: New const int field.
10129 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10130 minimum_stride field.
10131 (exynosm1_prefetch_tune): Likewise.
10132 (thunderxt88_prefetch_tune): Likewise.
10133 (thunderx_prefetch_tune): Likewise.
10134 (thunderx2t99_prefetch_tune): Likewise.
10135 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10136 (aarch64_override_options_internal): Update to set
10137 PARAM_PREFETCH_MINIMUM_STRIDE.
10138 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10139 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10140 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10141 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10142 stride is constant and is below the minimum stride threshold.
10144 2018-05-07 Luis Machado <luis.machado@linaro.org>
10146 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
10149 2018-05-07 Luis Machado <luis.machado@linaro.org>
10151 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10152 <prefetch_dynamic_strides>: New const bool field.
10153 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10154 prefetch_dynamic_strides.
10155 (exynosm1_prefetch_tune): Likewise.
10156 (thunderxt88_prefetch_tune): Likewise.
10157 (thunderx_prefetch_tune): Likewise.
10158 (thunderx2t99_prefetch_tune): Likewise.
10159 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
10161 (aarch64_override_options_internal): Update to set
10162 PARAM_PREFETCH_DYNAMIC_STRIDES.
10163 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10164 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10165 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10166 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10167 prefetch-dynamic-strides setting.
10169 2018-05-07 Luis Machado <luis.machado@linaro.org>
10171 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10172 <minimum_stride>: New const int field.
10173 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10174 minimum_stride field.
10175 (exynosm1_prefetch_tune): Likewise.
10176 (thunderxt88_prefetch_tune): Likewise.
10177 (thunderx_prefetch_tune): Likewise.
10178 (thunderx2t99_prefetch_tune): Likewise.
10179 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
10180 (aarch64_override_options_internal): Update to set
10181 PARAM_PREFETCH_MINIMUM_STRIDE.
10182 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10183 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10184 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10185 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10186 stride is constant and is below the minimum stride threshold.
10188 2018-05-06 Jakub Jelinek <jakub@redhat.com>
10191 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
10192 the type is addressable. Don't force op into register if it has
10195 2018-05-05 Roland McGrath <mcgrathr@google.com>
10198 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
10199 any section for which we don't know a specific type it should have,
10200 regardless of name. Previously this was done only for the exact
10201 names ".init_array", ".fini_array", and ".preinit_array".
10202 (default_elf_asm_named_section): Add comment about
10203 relationship with default_section_type_flags and SECTION_NOTYPE.
10204 (get_section): Don't consider it a type conflict if one side has
10205 SECTION_NOTYPE and the other doesn't, as long as neither has the
10206 SECTION_BSS et al used in the default_section_type_flags logic.
10208 2018-05-05 Tom de Vries <tom@codesourcery.com>
10211 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
10212 (workaround_barsyncs): New function.
10213 (nvptx_reorg): Use workaround_barsyncs.
10214 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10215 (define_expand "nvptx_membar_cta"): New define_expand.
10216 (define_insn "*nvptx_membar_cta"): New insn.
10218 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
10220 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
10221 To improve optimization opportunities.
10222 * builtin-types.def: The new needed builtin types for the above.
10224 2018-05-04 Richard Biener <rguenther@suse.de>
10226 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
10227 * gimple-ssa-store-merging.c
10228 (imm_store_chain_info::output_merged_store): Remove redundant create,
10229 release split_store vector contents on failure.
10230 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
10231 scalar stmt vector on cache hit.
10233 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
10235 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
10237 * config.gcc (powerpc-xilinx-eabi*): Remove.
10238 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
10240 (fusion_addis_mem_combo_load): Ditto.
10241 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
10243 (rs6000_cpu_cpp_builtins): Ditto.
10244 * config/rs6000/rs6000-linux.c
10245 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
10246 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
10247 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
10249 (rs6000_setup_reg_addr_masks): Ditto.
10250 (rs6000_init_hard_regno_mode_ok): Ditto.
10251 (rs6000_option_override_internal): Ditto.
10252 (legitimate_lo_sum_address_p): Ditto.
10253 (rs6000_legitimize_address): Ditto.
10254 (rs6000_legitimize_reload_address): Ditto.
10255 (rs6000_legitimate_address_p): Ditto.
10256 (abi_v4_pass_in_fpr): Ditto.
10257 (setup_incoming_varargs): Ditto.
10258 (rs6000_gimplify_va_arg): Ditto.
10259 (rs6000_split_multireg_move): Ditto.
10260 (rs6000_savres_strategy): Ditto.
10261 (rs6000_emit_prologue_components): Ditto.
10262 (rs6000_emit_epilogue_components): Ditto.
10263 (rs6000_emit_prologue): Ditto.
10264 (rs6000_emit_epilogue): Ditto.
10265 (rs6000_elf_file_end): Ditto.
10266 (rs6000_function_value): Ditto.
10267 (rs6000_libcall_value): Ditto.
10268 * config/rs6000/rs6000.h: Ditto.
10269 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
10270 (TARGET_MINMAX): ... this. New.
10271 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
10272 * config/rs6000/rs6000.md: Remove Xilinx FP support.
10273 (*movsi_internal1_single): Delete.
10274 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
10275 mfpu=, mxilinx-fpu): Delete.
10276 * config/rs6000/singlefp.h: Delete.
10277 * config/rs6000/sysv4.h: Remove Xilinx FP support.
10278 * config/rs6000/t-rs6000: Ditto.
10279 * config/rs6000/t-xilinx: Delete.
10280 * config/rs6000/titan.md: Adjust for fp_type removal.
10281 * config/rs6000/vsx.md: Remove Xilinx FP support.
10282 (VStype_simple): Delete.
10283 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
10284 * config/rs6000/xfpu.h: Delete.
10285 * config/rs6000/xfpu.md: Delete.
10286 * config/rs6000/xilinx.h: Delete.
10287 * config/rs6000/xilinx.opt: Delete.
10288 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
10289 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
10291 2018-05-04 Tom de Vries <tom@codesourcery.com>
10294 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
10297 2018-05-04 Richard Biener <rguenther@suse.de>
10299 PR middle-end/85627
10300 * tree-complex.c (update_complex_assignment): We are always in SSA form.
10301 (expand_complex_div_wide): Likewise.
10302 (expand_complex_operations_1): Likewise.
10303 (expand_complex_libcall): Preserve EH info of the original stmt.
10304 (tree_lower_complex): Handle removed blocks.
10305 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
10306 on complex multiplication and division libcall builtins.
10308 2018-05-04 Richard Biener <rguenther@suse.de>
10310 PR middle-end/85574
10311 * fold-const.c (negate_expr_p): Restrict negation of operand
10312 zero of a division to when we know that can happen without
10314 (fold_negate_expr_1): Likewise.
10316 2018-05-04 Jakub Jelinek <jakub@redhat.com>
10319 * real.h (real_nextafter): Declare.
10320 * real.c (real_nextafter): New function.
10321 * fold-const-call.c (fold_const_nextafter): New function.
10322 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
10323 CASE_CFN_NEXTTOWARD.
10324 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
10325 even when arg1_mode is different from arg0_mode.
10327 2018-05-03 Nathan Sidwell <nathan@acm.org>
10329 * doc/extend.texi (Deprecated Features): Remove
10330 -ffriend-injection.
10331 (Backwards Compatibility): Likewise.
10332 * doc/invoke.texi (C++ Language Options): Likewise.
10333 (C++ Dialect Options): Likewise.
10335 2018-05-03 Jakub Jelinek <jakub@redhat.com>
10338 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
10339 _mm512_mask_mullox_epi64): New intrinsics.
10341 2018-05-03 Tom de Vries <tom@codesourcery.com>
10344 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10345 dump files): Add offload-tree.
10347 2018-05-03 Richard Biener <rguenther@suse.de>
10349 PR tree-optimization/85615
10350 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
10351 to loops not nested in BBs loop father to avoid creating multi-entry
10354 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10356 PR tree-optimization/70291
10357 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
10358 arguments. Change return type to tree. Emit libcall as a new
10359 statement rather than replacing existing one when inplace_p is true.
10360 (expand_complex_multiplication_components): New function.
10361 (expand_complex_multiplication): Expand floating-point complex
10362 multiplication using the above.
10363 (expand_complex_division): Rename inner_type parameter to type.
10364 Update expand_complex_libcall call-site.
10365 (expand_complex_operations_1): Update expand_complex_multiplication
10366 and expand_complex_division call-sites.
10368 2018-05-02 Jakub Jelinek <jakub@redhat.com>
10371 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
10372 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
10373 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
10374 the highest significant bit of the shift count mask is clear. In
10375 check whether and[sq]i3 is needed verify that all significant bits
10376 of the shift count other than the highest are set.
10378 2018-05-02 Tom de Vries <tom@codesourcery.com>
10381 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
10382 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
10383 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
10384 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
10385 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
10386 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
10387 __builtin_goacc_parlevel_size.
10389 2018-05-02 Richard Biener <rguenther@suse.de>
10391 PR tree-optimization/85597
10392 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
10393 do not use split vect_get_vec_defs call but call vect_get_slp_defs
10396 2018-05-02 Tom de Vries <tom@codesourcery.com>
10399 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10400 dump files): Add ltrans-tree.
10402 2018-05-02 Tom de Vries <tom@codesourcery.com>
10405 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10406 dump files): Add wpa-ipa.
10408 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
10410 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
10411 powerpc*-*-linux*paired* target.
10412 * config/rs6000/750cl.h: Delete.
10413 * config/rs6000/paired.h: Delete.
10414 * config/rs6000/paired.md: Delete.
10415 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
10417 * config/rs6000/rs6000-builtin.def: Remove paired float support.
10418 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
10419 comment. Remove paired float support.
10420 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
10421 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
10423 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
10424 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
10426 * config/rs6000/rs6000.c: Remove paired float support.
10427 (paired_expand_vector_init, paired_expand_vector_move,
10428 paired_emit_vector_compare, paired_emit_vector_cond_expr,
10429 (paired_expand_lv_builtin, paired_expand_stv_builtin,
10430 paired_expand_builtin, paired_expand_predicate_builtin,
10431 paired_init_builtins): Delete.
10432 * config/rs6000/rs6000.h: Remove paired float support.
10433 * config/rs6000/rs6000.md: Remove paired float support.
10434 (move_from_CR_ov_bit): Delete.
10435 * config/rs6000/rs6000.opt (mpaired): Delete.
10436 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
10437 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
10439 2018-05-02 Richard Biener <rguenther@suse.de>
10441 PR middle-end/85567
10442 * gimplify.c (gimplify_save_expr): When in SSA form allow
10443 SAVE_EXPRs to compute to SSA vars.
10445 2018-05-02 Jakub Jelinek <jakub@redhat.com>
10448 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
10449 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
10450 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
10451 clobber operands[2], instead use a new pseudo. Formatting fixes.
10453 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
10455 PR tree-optimization/85586
10456 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
10457 exit early for statements in the same group if the accesses are
10460 2018-05-02 Tom de Vries <tom@codesourcery.com>
10463 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
10466 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
10468 PR tree-optimization/85143
10469 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
10471 2018-05-01 Tom de Vries <tom@codesourcery.com>
10474 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
10475 not found" error message.
10477 2018-05-01 Tom de Vries <tom@codesourcery.com>
10480 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
10481 * vec.c (test_ordered_remove_if): New function.
10482 (vec_c_tests): Call test_ordered_remove_if.
10483 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
10484 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
10485 * tree-vect-patterns.c (vect_pattern_recog_1): Use
10486 VEC_ORDERED_REMOVE_IF.
10488 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10490 PR tree-optimization/82665
10491 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
10492 pointer subtraction where arguments come from a memchr call.
10494 2018-05-01 Jakub Jelinek <jakub@redhat.com>
10496 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
10497 --push-state --as-needed and --pop-state instead of --as-needed and
10498 --no-as-needed if ld supports it.
10499 * configure: Regenerated.
10502 * doc/install.texi2html: Replace _002d with - and _002a with * in
10503 generated html files using sed.
10505 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10508 * gcc-rich-location.c (blank_line_before_p): New function.
10509 (use_new_line): New function.
10510 (gcc_rich_location::add_fixit_insert_formatted): New function.
10511 * gcc-rich-location.h
10512 (gcc_rich_location::add_fixit_insert_formatted): New function.
10514 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10516 * selftest.c (assert_streq): Rename "expected" and "actual" to
10517 "val1" and "val2". Extend NULL-handling to cover both inputs
10518 symmetrically, while still requiring both to be non-NULL for a pass.
10519 * selftest.h (assert_streq): Rename "expected" and "actual" to
10521 (ASSERT_EQ): Likewise.
10522 (ASSERT_EQ_AT): Likewise.
10523 (ASSERT_KNOWN_EQ): Likewise.
10524 (ASSERT_KNOWN_EQ_AT): Likewise.
10525 (ASSERT_NE): Likewise.
10526 (ASSERT_MAYBE_NE): Likewise.
10527 (ASSERT_MAYBE_NE_AT): Likewise.
10528 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
10529 the assertion to pass.
10530 (ASSERT_STREQ_AT): Likewise.
10532 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10534 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
10535 interaction with -pie.
10537 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10539 * selftest.h: Fix alphabetization of per-source-file selftest
10542 2018-04-30 Jason Merrill <jason@redhat.com>
10544 PR c++/61982 - dead stores to destroyed objects.
10545 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
10548 2018-04-30 Jason Merrill <jason@redhat.com>
10550 * tree.c (build_clobber): New.
10551 * tree.h: Declare it.
10552 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
10554 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10556 * diagnostic-show-locus.c (layout::layout): Update for
10557 location_get_source_line returning a char_span.
10558 (struct char_span): Move to input.h.
10559 (struct correction): Update for fields in char_span becoming
10561 (struct source_line): Update for location_get_source_line
10562 returning a char_span.
10563 (layout::print_line): Likewise.
10564 * edit-context.c (edited_file::print_content): Likewise.
10565 (edited_file::print_diff_hunk): Likewise.
10566 (edited_file::print_run_of_changed_lines): Likewise.
10567 (edited_file::get_num_lines): Likewise.
10568 (edited_line::edited_line): Likewise.
10569 * final.c (asm_show_source): Likewise.
10570 * input.c (location_get_source_line): Convert return type
10571 from const char * to char_span, losing the final "line_len"
10573 (dump_location_info): Update for the above.
10574 (get_substring_ranges_for_loc): Likewise. Use a char_span
10575 when handling the literal within the line.
10576 (test_reading_source_line): Update for location_get_source_line
10577 returning a char_span.
10578 * input.h (class char_span): Move here from
10579 diagnostic-show-locus.c, converting from a struct to a class.
10580 Make data members private.
10581 (char_span::operator bool): New.
10582 (char_span::length): New.
10583 (char_span::get_buffer): New.
10584 (char_span::operator[]): New.
10585 (char_span::subspan): Make const.
10586 (char_span::xstrdup): New.
10587 (location_get_source_line): Convert return type from const char *
10588 to char_span, losing the final "line_size" param.
10590 2018-04-30 Jan Hubicka <jh@suse.cz>
10592 * lto-wrapper.c (ltrans_priorities): New static var.
10593 (cmp_priority): New.
10594 (run_gcc): Read priorities and if doing parallel build order
10595 the Makefile by them.
10597 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10599 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
10601 2018-04-30 Richard Biener <rguenther@suse.de>
10603 * tree-cfg.c (verify_address): Remove base argument, add
10604 flag whether to check TREE_ADDRESSABLE and do that.
10605 (verify_expr): Remove.
10606 (verify_types_in_gimple_reference): Add pieces from verify_expr.
10607 (verify_gimple_assign_single): Likewise.
10608 (verify_gimple_switch): Likewise.
10609 (verify_expr_location_1): Dereference tp once. Add (disabled)
10610 piece from verify_expr.
10611 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
10613 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10615 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
10617 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10619 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
10620 (small_data_pattern): Likewise.
10621 (arc_rewrite_small_data): Likewise.
10622 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
10623 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
10624 (get_symbol_alignment): New function.
10625 (legitimate_small_data_address_p): Likewise.
10626 (legitimate_scaled_address): Update, call
10627 legitimate_small_data_address_p.
10628 (output_sdata): New static variable.
10629 (arc_print_operand): Update how we handle small data operands.
10630 (arc_print_operand_address): Likewise.
10631 (arc_legitimate_address_p): Update, use
10632 legitimate_small_data_address_p.
10633 (arc_rewrite_small_data_p): Remove.
10634 (arc_rewrite_small_data_1): Likewise.
10635 (arc_rewrite_small_data): Likewise.
10636 (small_data_pattern): Likewise.
10637 (compact_sda_memory_operand): Update to use
10638 legitimate_small_data_address_p and get_symbol_alignment.
10639 (prepare_move_operands): Don't rewite sdata pattern.
10640 (prepare_extend_operands): Remove.
10641 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
10643 (zero_extendqisi2): Likewise.
10644 (zero_extendhisi2): Likewise.
10645 (extendqihi2): Likewise.
10646 (extendqisi2): Likewise.
10647 (extendhisi2): Likewise.
10648 (addsi3): Likewise.
10649 (subsi3): Likewise.
10650 (andsi3): Likewise.
10651 * config/arc/constraints.md (Usd): Change it to memory constraint.
10653 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10655 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
10656 as source of std instructions.
10657 * config/arc/arc.md (movsi_insn): Update pattern predicate to
10658 allow 6-bit constants as source for store instructions.
10659 (movdi_insn): Update instruction pattern to allow 6-bit constants
10660 as source for store instructions.
10662 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10664 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
10666 2018-04-30 Nathan Sidwell <nathan@acm.org>
10667 Sandra Loosemore <sandra@codesourcery.com>
10669 * dumpfile.c (dump_open): Allow '-' for stdout.
10670 * doc/invoke.texi (Developer Options): Document dump filename
10671 determination early. Document stdin/stdout selection.
10673 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
10675 Microblaze Target: PIC data text relative
10677 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
10678 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
10680 * config/microblaze/microblaze.h (microblaze_constant_address_p):
10681 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
10682 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
10683 New addressing mode for data-text relative position indepenedent code.
10684 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
10685 'ADDRESS_SYMBOLIC_TXT_REL'.
10686 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
10687 (microblaze_legitimate_pic_operand): Exclude function calls from
10688 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
10689 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
10691 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10692 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10693 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
10694 for 'address + offset'.
10695 (microblaze_expand_prologue): Add new function prologue call for
10697 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
10698 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
10699 table in case of TARGET_PIC_DATA_TEXT_REL.
10700 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
10701 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
10702 Add new macros 'UNSPEC_TEXT',
10703 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
10704 + exclude function calls from 'UNSPEC_PLT' in case of data text
10706 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10707 new target hook for generating address diff vector tables in case of
10709 * doc/tm.texi : Regenerate.
10710 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
10711 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
10712 of addr diff vector generation.
10713 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10714 target hook definition.
10715 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
10716 Add default function for generate_pic_addr_diff_vec -> flag_pic.
10717 * doc/invoke.texi (Add new pic option): Add new microblaze pic
10718 option for data text relative.
10720 2018-04-30 Richard Biener <rguenther@suse.de>
10722 * tree-chrec.h (evolution_function_is_constant_p): Remove
10724 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
10726 2018-04-30 Richard Biener <rguenther@suse.de>
10729 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
10731 2018-04-30 Richard Biener <rguenther@suse.de>
10733 PR tree-optimization/28364
10734 PR tree-optimization/85275
10735 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
10736 copying first exit test.
10738 2018-04-28 Mark Wielaard <mark@klomp.org>
10740 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
10741 dwarf_version >= 5.
10742 (dwarf_AT): Handle DW_AT_addr_base.
10743 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
10745 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
10748 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
10749 (*ashl<dwi>3_doubleword_mask_1): Ditto.
10750 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
10751 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
10753 2018-04-28 Richard Biener <rguenther@suse.de>
10755 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
10756 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
10757 to reflect use. Only add interesting stmts.
10759 2018-04-27 Martin Jambor <mjambor@suse.cz>
10762 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
10763 the jump function allows for passing through aggregate values.
10765 2018-04-27 David Malcolm <dmalcolm@redhat.com>
10767 * input.h (in_system_header_at): Convert from macro to inline
10769 (from_macro_expansion_at): Likewise.
10770 (from_macro_definition_at): Likewise.
10772 2018-04-27 Jeff Law <law@redhat.com>
10774 * config.gcc: Mark tile* targets as deprecated/obsolete.
10776 2018-04-27 Richard Biener <rguenther@suse.de>
10778 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
10781 2018-04-27 Richard Biener <rguenther@suse.de>
10783 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
10785 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
10787 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
10788 with Yd constraint. Set "preferred_for_speed" attribute from
10789 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
10790 with Yd constraint.
10791 (*movdi_internal): Ditto.
10792 (movti_interunit splitters): Remove
10793 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
10794 (movdi_interunit splitters): Ditto.
10795 * config/i386/constraints.md (Ye): Remove.
10796 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
10798 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10801 * config/aarch64/constraints.md (Usg): Limit to 31.
10802 (Usj): Limit to 63.
10804 2018-04-27 Jakub Jelinek <jakub@redhat.com>
10806 PR tree-optimization/85529
10807 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
10808 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
10809 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
10810 zero extension or masking of the MSB bit.
10811 (optimize_range_tests): Add FIRST_BB argument, pass it through
10812 to optimize_range_tests_var_bound.
10813 (maybe_optimize_range_tests, reassociate_bb): Adjust
10814 optimize_range_tests callers.
10816 2018-04-26 Richard Biener <rguenther@suse.de>
10817 Jakub Jelinek <jakub@redhat.com>
10819 * cgraph.h (symbol_table): Just declare debug method here.
10820 * symtab.c (symbol_table::debug): Define.
10822 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
10824 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
10826 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
10828 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
10829 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
10830 (*movdi_internal): Substitute Yi and Yj constraint with x
10831 and Ym and Yn constraint with y constraint. Update "isa"
10832 attribute and set "preferred_for_speed" attribute from
10833 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
10834 (*movsi_internal): Ditto.
10835 (*movdf_internal): Ditto.
10836 (*movsf_internal): Ditto.
10837 (*zero_extendsidi2): Ditto.
10838 * config/i386/sse.md (vec_set<mode>_0): Ditto.
10839 (sse2_loadld): Ditto.
10840 (*vec_extract<ssevecmodelower>_0): Ditto.
10841 (*vec_extractv4si_0_zext_sse4): Ditto.
10842 (vec_concatv2di): Ditto.
10843 (*vec_dup<mode>): Ditto.
10844 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
10845 * config/i386/constraints.md (Yi): Remove.
10850 2018-04-26 Nathan Sidwell <nathan@acm.org>
10852 * dumpfile.c (dump_open): New.
10853 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
10854 (dump_finish): Detect stdio/stderr by value not name.
10856 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
10858 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
10860 2018-04-26 Tom de Vries <tom@codesourcery.com>
10863 * config/nvptx/nvptx.c (verify_neutering_jumps)
10864 (verify_neutering_labels): New function
10865 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
10867 2018-04-26 Tom de Vries <tom@codesourcery.com>
10870 * config/nvptx/nvptx.c (needs_neutering_p): New function.
10871 (nvptx_single): Use needs_neutering_p to skip over insns that do not
10874 2018-04-26 Richard Biener <rguenther@suse.de>
10875 Tom de Vries <tom@codesourcery.com>
10878 * lto-streamer-out.c (output_function): Fixup loops if required to match
10879 discovery done in the reader.
10881 2018-04-26 Richard Biener <rguenther@suse.de>
10883 PR tree-optimization/85116
10884 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
10885 have a loop exit from the single latch predecessor. Remove
10886 case of header with just condition.
10887 (ch_base::copy_headers): Exclude infinite loops from any
10889 (pass_ch::execute): Record exits.
10891 2018-04-26 Richard Biener <rguenther@suse.de>
10893 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
10894 prologue cost vector and pass it to vect_get_load_cost.
10895 (vect_get_peeling_costs_all_drs): Likewise.
10896 (vect_peeling_hash_get_lowest_cost): Likewise.
10897 (vect_enhance_data_refs_alignment): Likewise.
10899 2018-04-26 Richard Biener <rguenther@suse.de>
10901 PR middle-end/85450
10902 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
10903 checking of integer<->pointer conversions.
10904 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
10905 sign-/zero-extending pointer types.
10906 (expand_omp_for_static_chunk): Likewise.
10908 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
10909 Jean Lee <xiaoyur347@gmail.com>
10911 * config/mips/mips.c (mips_asan_shadow_offset): New function.
10912 (TARGET_ASAN_SHADOW_OFFSET): Define.
10913 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
10914 true for -fsanitize=address.
10916 2018-04-25 Mark Wielaard <mark@klomp.org>
10918 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
10921 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10923 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
10924 than "alu", remove explicit "memory" and "imm_disp" attributes.
10925 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
10927 PR middle-end/85414
10928 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
10929 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
10930 gen_lowpart_no_emit.
10932 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
10935 * config/i386/i386.c (ix86_expand_builtin): Change memory
10936 operand to XI, extend p0 to Pmode.
10937 * config/i386/i386.md: Change unspec volatile and operand
10938 1 mode to XI, change operand 0 mode to P.
10940 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10942 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
10943 GET_MODE_MASK before any checking.
10944 (nds32_can_use_bset_p): Likewise.
10945 (nds32_can_use_btgl_p): Likewise.
10947 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10949 * config/nds32/nds32-doubleword.md: New define_split pattern for
10950 illegal register number.
10952 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10954 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10956 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10958 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
10960 2018-04-25 Richard Biener <rguenther@suse.de>
10962 * lto-streamer.h (LTO_major_version): Bump to 8.
10964 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10966 * BASE-VER: Set to 9.0.0.
10968 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
10970 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
10971 in __abskf2 and __powikf2.
10973 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10976 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10977 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
10978 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
10979 Use the above on operand 2. Reindent.
10980 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
10982 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10985 * common/config/i386/i386-common.c (ix86_handle_option): Don't
10987 * config/i386/i386.opt (mcet): Removed.
10988 * doc/install.texi: Remove -mcet documentation.
10989 * doc/invoke.texi: Likewise.
10991 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10994 * doc/install.texi: Remove -mcet from bootstrap-cet.
10996 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10999 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
11000 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
11004 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
11005 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
11006 containing a CONST_VECTOR.
11008 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
11010 * doc/install.texi: Update newlib dependency for nvptx.
11012 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11015 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
11016 instead of INTVAL when shifting x left.
11018 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
11020 PR tree-optimization/85478
11021 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
11022 vect_grouped_store_supported for single element vectors.
11024 2018-04-24 Richard Biener <rguenther@suse.de>
11027 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
11028 load cost increase to the case of non-constant step.
11030 2018-04-24 Jakub Jelinek <jakub@redhat.com>
11033 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
11034 destination if any_malformed_asm.
11036 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
11038 PR middle-end/85496
11039 * expr.c (store_field): In the bitfield case, if the value comes from
11040 a function call and is returned in registers by means of a PARALLEL,
11041 do not change the mode of the temporary unless BLKmode and VOIDmode.
11043 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
11045 PR rtl-optimization/85423
11046 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
11047 dependencies to debug insns when the previous insn is non-debug.
11049 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11051 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
11052 enums into a single definition.
11053 (fls): Fix predicates and printing.
11056 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
11058 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
11059 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
11060 and short u6 immediate.
11061 (check_if_valid_sleep_operand): Remove.
11062 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
11064 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11066 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
11067 flag_always_save_lp condition.
11068 * config/nds32/nds32.opt (malways-save-lp): New option.
11070 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11072 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
11073 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
11074 * config/nds32/nds32.h
11075 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
11076 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
11078 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11080 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
11081 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
11083 2018-04-22 Shiva Chen <shiva0217@gmail.com>
11084 Chung-Ju Wu <jasonwucj@gmail.com>
11086 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
11088 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
11089 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
11091 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11093 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
11095 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11097 * config/nds32/nds32-protos.h (nds32_data_alignment,
11098 nds32_local_alignment): Declare.
11099 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
11100 nds32_local_alignment): New functions.
11101 (TARGET_CONSTANT_ALIGNMENT): Define.
11102 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
11104 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11106 * config/nds32/nds32.c
11107 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
11108 (TARGET_MODES_TIEABLE_P): Likewise.
11110 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
11112 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
11113 level Ofast and Og.
11115 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
11116 Chung-Ju Wu <jasonwucj@gmail.com>
11118 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11119 for unaligned access.
11120 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11122 * config/nds32/nds32-intrinsic.md: Likewise.
11123 * config/nds32/nds32_intrinsic.h: Likewise.
11124 * config/nds32/nds32.h (nds32_builtins): Likewise.
11125 * config/nds32/nds32.opt (munaligned-access): New option.
11126 * config/nds32/nds32.c (nds32_asm_file_start): Display
11127 flag_unaligned_access status.
11129 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11131 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
11132 -mno-relax is present.
11133 * config/riscv/linux.h (LINK_SPEC): Ditto.
11135 2018-04-20 Martin Sebor <msebor@redhat.com>
11138 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
11140 (gimple_fold_builtin_stxcpy_chk): Same.
11141 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
11143 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
11146 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
11147 __powikf2 when long double is IEEE 128-bit.
11149 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
11151 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
11152 step to make sure stack always aligned.
11154 2018-04-20 Carl Love <cel@us.ibm.com>
11157 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
11158 size check for arg0.
11160 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
11161 Tom de Vries <tom@codesourcery.com>
11164 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
11165 Emit insns for calls too.
11166 (nvptx_find_par): Always look for worker-level predecessor insn.
11167 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
11169 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
11170 (nvptx_process_pars): Propagate frames for calls.
11172 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
11175 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
11177 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11178 (ix86_handle_option): Don't handle OPT_mibt.
11179 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
11181 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
11183 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
11184 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
11185 (ix86_target_macros): Define __CET__ with flag_cf_protection
11186 for -fcf-protection.
11187 * config/i386/i386.c (isa2_opts): Remove -mibt.
11188 * config/i386/i386.h (TARGET_IBT): Removed.
11189 (TARGET_IBT_P): Likewise.
11190 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
11191 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
11192 * config/i386/i386.opt (mcet): Update help message.
11193 (mshstk): Likewise.
11195 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
11196 -mcet as an alias for -mshstk.
11198 2018-04-20 Richard Biener <rguenther@suse.de>
11200 PR middle-end/85475
11201 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
11202 complexity by forcing a single use of the multiply operand.
11204 2018-04-20 Martin Jambor <mjambor@suse.cz>
11207 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
11208 recursion dependency to only apply to non-clones.
11210 2018-04-20 Martin Jambor <mjambor@suse.cz>
11213 * ipa-cp.c (create_specialized_node): Check that clones of
11214 self-recursive edges exist during IPA-CP.
11216 2018-04-19 Toon Moene <toon@moene.org>
11218 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
11221 2018-04-19 Jakub Jelinek <jakub@redhat.com>
11223 PR tree-optimization/85467
11224 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
11225 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
11226 VECTOR_CST element to type.
11228 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11231 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
11232 * config/i386/i386.md (builtin_setjmp_setup): Removed.
11233 (builtin_longjmp): Likewise.
11234 (save_stack_nonlocal): New pattern.
11235 (restore_stack_nonlocal): Likewise.
11237 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11240 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11241 Replace ASM_OUTPUT_LABEL with fprintf.
11243 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
11246 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
11247 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
11248 * config/i386/i386-c.c (ix86_target_macros_internal): Also
11249 define __IBT__ and __SHSTK__ for -fcf-protection.
11250 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
11252 (ix86_trampoline_init): Likewise.
11253 (x86_output_mi_thunk): Likewise.
11254 (ix86_notrack_prefixed_insn_p): Likewise.
11255 (ix86_option_override_internal): Don't disallow -fcf-protection.
11256 * config/i386/i386.md (rdssp<mode>): Also enable for
11258 (incssp<mode>): Likewise.
11259 (nop_endbr): Likewise.
11260 * config/i386/i386.opt (mcet): Change help message to built-in
11263 (mshstk): Likewise.
11264 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
11265 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
11266 enable CET built-in functions.
11268 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
11270 * common/config/i386/i386-common.c
11271 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
11272 OPTION_MASK_ISA_MOVDIRI_UNSET,
11273 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
11274 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
11275 * config.gcc (movdirintrin.h): New header.
11276 * config/i386/cpuid.h (bit_MOVDIRI,
11277 bit_MOVDIR64B): New bits.
11278 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
11280 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
11281 (VOID, PVOID, PCVOID)): New function types.
11282 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
11283 __builtin_ia32_directstoreu_u64,
11284 __builtin_ia32_movdir64b): New builtins.
11285 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
11286 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
11288 (ix86_valid_target_attribute_inner_p): Ditto.
11289 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
11290 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
11291 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
11292 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
11293 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
11294 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
11295 (movdiri<mode>, movdir64b_<mode>): New.
11296 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
11297 * config/i386/immintrin.h: Include movdirintrin.h.
11298 * config/i386/movdirintrin.h: New file.
11299 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
11301 2018-04-19 Richard Biener <rguenther@suse.de>
11303 PR middle-end/85455
11304 * cfg.c (clear_bb_flags): When loop state says we have
11305 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
11307 2018-04-19 Richard Biener <rguenther@suse.de>
11309 PR tree-optimization/84737
11310 * tree-vect-data-refs.c (vect_copy_ref_info): New function
11311 copying restrict info.
11312 (vect_setup_realignment): Use it.
11313 * tree-vectorizer.h (vect_copy_ref_info): Declare.
11314 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
11315 the first DR to all generated stores.
11316 (vectorizable_load): Likewise for loads.
11318 2018-04-19 Jakub Jelinek <jakub@redhat.com>
11320 PR tree-optimization/85446
11321 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
11322 the integral and pointer types to have the same precision.
11324 * doc/install.texi: Document --disable-cet being the default and
11327 2018-04-18 Martin Liska <mliska@suse.cz>
11329 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
11332 2018-04-18 Martin Liska <mliska@suse.cz>
11335 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11338 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11339 arguments if they are comparable.
11341 2018-04-18 Martin Liska <mliska@suse.cz>
11344 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11347 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11350 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
11353 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
11354 ENDBR after calling __morestack.
11356 2018-04-18 David Malcolm <dmalcolm@redhat.com>
11359 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
11360 by using gcc_base_ver to generate a gcc_driver_version, and use
11361 it when generating GCC_DRIVER_NAME.
11362 * configure: Regenerate.
11364 2018-04-18 Jakub Jelinek <jakub@redhat.com>
11367 * config.gcc: Obsolete powerpc*-*-*spe*.
11369 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11372 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
11373 (stabstr_D): Change type of unum from unsigned int to
11374 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
11377 2018-04-17 Jim Wilson <jimw@sifive.com>
11380 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
11381 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
11382 Set arg_pointer_offset after using pretend_args_size.
11384 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11386 PR rtl-optimization/85431
11387 * dse.c (record_store): Ignore zero width stores.
11390 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
11391 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
11392 __builtin_stack_restore rather than after it.
11393 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
11394 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
11395 argument instead of virtual_dynamic_stack_rtx.
11397 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
11399 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
11401 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
11402 Add note to error message to explain internal mapping of overloaded
11403 built-in function name to non-overloaded built-in function name.
11404 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
11407 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11410 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
11411 where the inputs overlap with the output.
11413 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11415 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
11416 (=v, v) alternative and explicit "memory" attribute.
11417 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
11418 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11420 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
11421 "sselog1" type instead of "sselog".
11422 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
11423 "sselog". Remove explicit "memory" attribute.
11424 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
11425 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11427 (vec_extract_hi_v32hi): Merge all alternatives into one, use
11428 "sselog1" type instead of "sselog". Remove explicit "memory"
11430 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
11431 use "sselog1" type instead of "sselog". Remove explicit "memory"
11433 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
11434 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11436 (vec_extract_hi_v64qi): Merge all alternatives into one, use
11437 "sselog1" type instead of "sselog". Remove explicit "memory"
11439 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
11440 use "sselog1" type instead of "sselog". Remove explicit "memory"
11444 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
11446 PR middle-end/85414
11447 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
11450 2018-04-17 Martin Jambor <mjambor@suse.cz>
11453 * ipa-cp.c (create_specialized_node): Call
11454 expand_all_artificial_thunks if necessary.
11456 2018-04-17 Martin Liska <mliska@suse.cz>
11459 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
11460 in message, remote space in between '_G' and '('.
11462 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11465 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
11466 avx512f_vmcmp<mode>3<round_saeonly_name>,
11467 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
11468 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
11469 avx512f_rndscale<mode><round_saeonly_name>,
11470 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
11471 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
11472 Use %<iptr>2 instead of %2 for -masm=intel.
11473 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
11474 avx512f_vcvttss2usi<round_saeonly_name>,
11475 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
11477 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
11478 avx512f_vcvttsd2usi<round_saeonly_name>,
11479 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
11480 Use %q1 instead of %1 for -masm=intel.
11481 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
11482 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
11483 of %3 for -masm=intel.
11484 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
11486 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
11488 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
11490 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
11491 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
11493 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
11495 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
11496 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
11497 %g1 and one with %0 and %1.
11498 (avx512er_vmrcp28<mode><round_saeonly_name>,
11499 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
11500 %1 for -masm=intel.
11501 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
11502 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
11503 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
11504 of %0 and %{%4%} for -masm=intel.
11505 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
11506 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
11507 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
11508 order of %0 and %{%5%}%{z%} for -masm=intel.
11510 2018-04-17 Jan Hubicka <jh@suse.cz>
11513 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
11515 2018-04-17 Martin Liska <mliska@suse.cz>
11518 * multiple_target.c (create_dispatcher_calls): Set apostrophes
11519 for target_clone error message. Make default implementation
11520 clone to be a local declaration.
11521 (separate_attrs): Add new argument and check for an empty
11523 (expand_target_clones): Handle it.
11524 (ipa_target_clone): Make redirection just for target_clones
11527 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
11528 Tom de Vries <tom@codesourcery.com>
11530 PR middle-end/84955
11531 * omp-expand.c (expand_oacc_for): Add dummy false branch for
11532 tiled basic blocks without omp continue statements.
11534 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
11537 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
11538 vec_extract expression as having side effects to make sure it gets
11541 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
11544 * config/i386/i386.c (get_builtin_code_for_version): Check
11547 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
11550 * config.gcc: Support "skylake".
11551 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11553 * config/i386/i386.c (m_SKYLAKE): Define.
11554 (processor_target_table): Add "skylake".
11555 (ix86_option_override_internal): Add "skylake".
11556 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
11557 PROCESSOR_CANNONLAKE.
11558 (get_builtin_code_for_version): Fix priority for
11559 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
11560 PROCESSOR_SKYLAKE-AVX512.
11561 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
11562 (processor_type): Add PROCESSOR_SKYLAKE.
11564 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
11565 Jason Merrill <jason@redhat.com>
11568 * convert.c (convert_to_integer_1): Use direct recursion for
11569 enumeral types and types with a precision less than the number
11570 of bits in their mode.
11572 2018-04-16 Julia Koval <julia.koval@intel.com>
11575 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
11576 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
11578 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
11581 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
11582 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
11583 and -mno-direct-move.
11585 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
11588 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
11589 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
11590 Ensure negative shifts result in {0}.
11592 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
11594 PR rtl-optimization/79916
11595 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
11596 regs (if any) to define how to gnerate SD moves when LRA is in
11599 2018-04-13 Jakub Jelinek <jakub@redhat.com>
11601 PR rtl-optimization/85393
11602 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
11603 * except.c (expand_dw2_landing_pad_for_region): Make static.
11604 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
11605 a label and unconditional jump to old_bb, rather than
11606 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
11609 PR rtl-optimization/85376
11610 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
11611 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
11612 instead of a specific value.
11614 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11615 Bin Cheng <bin.cheng@arm.com>
11617 PR tree-optimization/82965
11618 PR tree-optimization/83991
11619 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
11620 by_profile_only parameter.
11621 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
11622 information if the loop was predicted to iterate too many times.
11623 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
11625 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11628 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
11631 2018-04-13 Martin Liska <mliska@suse.cz>
11632 Jakub Jelinek <jakub@redhat.com>
11634 PR middle-end/81657
11635 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
11636 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
11637 * builtins.c (expand_builtin_memory_copy_args): Use
11638 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
11639 handle dest_addr == pc_rtx.
11641 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
11644 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
11645 asked to not generate direct moves.
11646 (fix_trunc<mode>si2_stfiwx): Similar.
11647 (fix_trunc<mode>si2_internal): Similar.
11649 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11652 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
11653 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
11654 lookup if dest in some wider mode is known to be const0_rtx and
11655 if so, record permanent equivalence for it to be ZERO_EXTEND of
11656 the narrower mode destination.
11658 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11660 * lto-streamer-out.c (output_function): Revert 259346.
11661 * omp-expand.c (expand_oacc_for): Likewise.
11663 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
11665 PR rtl-optimization/85354
11666 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
11667 * sel-sched.c (sel_global_init): ... here.
11669 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
11672 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
11673 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11674 mode for PE-COFF targets.
11675 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
11676 (i386_pe_asm_lto_end): Likewise.
11677 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
11678 (TARGET_ASM_LTO_END): Likewise.
11679 * config/i386/winnt.c (saved_debug_info_level): New static variable.
11680 (i386_pe_asm_lto_start): New function.
11681 (i386_pe_asm_lto_end): Likewise.
11683 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11684 Richard Biener <rguenther@suse.de>
11686 PR middle-end/84955
11687 * lto-streamer-out.c (output_function): Fix CFG loop state before
11689 * omp-expand.c (expand_oacc_for): Handle calls to internal
11690 functions like regular functions.
11692 2018-04-12 Richard Biener <rguenther@suse.de>
11695 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
11696 for the early LTO debug to properly generate references to it
11697 during DIE emission. Do not re-use that for the skeleton for
11699 (dwarf2out_early_finish): Likewise.
11701 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11704 * config/i386/sse.md
11705 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
11706 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
11707 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
11708 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
11709 and output is a reg, avoid creating invalid lowpart subreg, but
11710 instead split into a 512-bit move. Don't split if not AVX512VL,
11711 input is xmm16+ reg and output is a mem.
11712 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
11713 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
11714 xmm16+ reg and output is a mem.
11716 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11718 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
11719 also for flag_dwarf2_cfi_asm.
11721 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11723 PR rtl-optimization/85342
11724 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
11725 a bool scalar var inside of the loop instead. Don't try to update
11726 recog_data.operand after failed apply_change_group.
11728 2018-04-12 Tom de Vries <tom@codesourcery.com>
11731 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
11732 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
11733 array with flexible array member as array without given dimension.
11734 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
11735 argument for undefined param to true.
11737 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
11740 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11741 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
11742 from PowerPC section.
11743 * config/rs6000/sysv4.opt (mcall-): Improve help text.
11744 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
11745 help text that is too long.
11746 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
11747 help text that is too long.
11748 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
11749 help text that is too long.
11751 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
11753 * config/alpha/alpha.md (stack_probe_internal): Rename
11754 from "probe_stack". Update all callers.
11756 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11758 PR rtl-optimization/84566
11759 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
11760 sched_macro_fuse_insns.
11762 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11765 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
11766 (compute_block_dependences): ... from here.
11768 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11770 PR tree-optimization/85331
11771 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
11772 from int to HOST_WIDE_INT.
11774 2018-04-11 Martin Jambor <mjambor@suse.cz>
11777 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
11778 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
11779 not the same as the source val.
11780 (cgraph_edge_brings_value_p): New parameter.
11781 (gather_edges_for_value): Pass destination value to
11782 cgraph_edge_brings_value_p.
11783 (perhaps_add_new_callers): Likewise.
11784 (get_info_about_necessary_edges): Likewise and exclude values brought
11785 only by self-recursive edges.
11786 (create_specialized_node): Redirect only clones of self-calling edges.
11787 (+self_recursive_pass_through_p): New function.
11788 (find_more_scalar_values_for_callers_subset): Use it.
11789 (find_aggregate_values_for_callers_subset): Likewise.
11790 (known_aggs_to_agg_replacement_list): Removed.
11791 (decide_whether_version_node): Re-calculate known constants for all
11792 remaining context clones.
11794 2018-04-11 Richard Biener <rguenther@suse.de>
11797 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
11798 from early DWARF output.
11799 (dwarf2out_early_finish): Output line info unconditionally into
11800 early DWARF and add reference to it.
11802 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11805 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
11806 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
11807 other than V2DFmode using iptr mode attribute.
11808 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
11810 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11812 PR rtl-optimization/84659
11813 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
11815 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11818 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
11820 (output_loc_list): Pass address of a dummy size variable even in the
11821 locview handling loop.
11822 (index_location_lists): Add comment on why skip_loc_list_entry can't
11825 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
11828 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
11831 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
11834 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11835 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
11836 and -mstring-compare-inline-limit.
11838 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11841 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
11842 for stack clash protection in a register whenever we need it to be in
11845 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11847 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
11848 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
11850 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11853 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
11855 (mlong-double-): Ditto.
11856 * config/rs6000/sysv4.opt (msdata=): Ditto.
11857 (mtls-size=): Ditto.
11859 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11861 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11862 erroneous entries for
11863 "vector int vec_ldl (int, long int *)", and
11864 "vector unsigned int vec_ldl (int, unsigned long int *)".
11865 Add comments and entries for
11866 "vector bool char vec_ldl (int, bool char *)",
11867 "vector bool short vec_ldl (int, bool short *)",
11868 "vector bool int vec_ldl (int, bool int *)",
11869 "vector bool long long vec_ldl (int, bool long long *)",
11870 "vector pixel vec_ldl (int, pixel *)",
11871 "vector long long vec_ldl (int, long long *)",
11872 "vector unsigned long long vec_ldl (int, unsigned long long *)".
11873 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
11874 type tree bool_long_long_type_node and correct definition of
11875 bool_V2DI_type_node to make reference to this new type tree.
11876 (rs6000_mangle_type): Replace erroneous reference to
11877 bool_long_type_node with bool_long_long_type_node.
11878 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
11879 comments to emphasize sign distinctions for char and int types and
11880 replace RS6000_BTI_bool_long constant with
11881 RS6000_BTI_bool_long_long constant. Also add comment to restrict
11882 use of RS6000_BTI_pixel.
11883 (bool_long_type_node): Remove this macro definition.
11884 (bool_long_long_type_node): New macro definition
11886 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11888 PR rtl-optimization/85300
11889 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
11890 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
11891 simplify_unary_operation fails.
11893 2018-04-10 Martin Liska <mliska@suse.cz>
11895 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
11896 cgraph_edge and ipa_ref.
11898 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11902 * config/i386/sse.md
11903 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
11904 computation of the VEC_MERGE selector from mask.
11905 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
11906 Fix decoding of the VEC_MERGE selector into mask.
11908 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11910 PR tree-optimization/85286
11911 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
11913 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11915 * final.c (final_1): Set insn_last_address as well as
11916 insn_current_address.
11918 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11921 * explow.c (emit_stack_probe): Call validize_mem on memory location
11922 before passing it to gen_probe_stack. Create address operand and
11923 legitimize it for the probe_stack_address case.
11925 2018-04-09 Jan Hubicka <jh@suse.cz>
11928 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
11929 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
11930 * tree.c (free_lang_data_in_type): Fix handling of binfos;
11932 (free_lang_data): Rebuild type inheritance graph.
11934 2018-04-09 Martin Sebor <msebor@redhat.com>
11936 * invoke.texi (-finline-small-functions): Mention other optimization
11938 (-findirect-inlining, -fpartial-inlining): Same.
11939 (-finline-functions-called-once): Same.
11940 (-freorder-blocks-and-partition): Same.
11942 2018-04-09 Jan Hubicka <jh@suse.cz>
11945 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
11946 jumps; choose last target that matches the criteria (i.e.
11947 no partition changes for non-crossing jumps).
11948 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
11949 support for redirecting crossing jumps to non-crossing.
11951 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
11953 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
11954 also for naked functions.
11956 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
11958 * config/arc/arc.md (add_shift): New pattern.
11959 (add_shift2): Likewise.
11960 (sub_shift): Likewise.
11961 (sub_shift_cmp0_noout): Likewise.
11962 (compare_si_ashiftsi): Likewise.
11963 (xbfu_cmp0_noout): New combine pattern.
11964 (xbfu_cmp0"): Likewise.
11965 (movsi_set_cc_insn): Place the predicable variant first.
11966 (commutative_binary_cmp0_noout): Remove clobber.
11967 (commutative_binary_cmp0): New pattern.
11968 (noncommutative_binary_cmp0): Likewise.
11969 (noncommutative_binary_cmp0_noout): Likewise.
11970 (noncommutative_binary_comparison_result_used): Removed.
11971 (rsub_cmp0): New pattern.
11972 (rsub_cmp0_noout): Likewise.
11973 (extzvsi): Changed, keep only meaningful variants.
11974 (SQH, SEZ): New iterators.
11975 (SQH_postfix): New mode attribute.
11976 (SEZ_prefix): New code attribute.
11977 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
11978 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
11979 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
11980 of numerical value.
11981 (noncommutative_operator): Check the availability of barrel
11984 2018-04-09 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/85284
11987 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
11988 Only use the niter constraining form of simple_iv when the exit
11989 is always executed.
11991 2018-04-09 Tom de Vries <tom@codesourcery.com>
11994 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11995 (define_expand "*memory_barrier"): New define_expand.
11996 (define_insn "memory_barrier"): New insn.
11998 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12000 PR rtl-optimization/80463
12001 PR rtl-optimization/83972
12002 PR rtl-optimization/83480
12004 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
12005 correct producer for the insn.
12006 (tidy_control_flow): Fixup seqnos in case of debug insns.
12008 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12010 PR rtl-optimization/83913
12012 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
12013 different sched-times when merging exprs.
12015 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12017 PR rtl-optimization/83962
12019 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
12020 tidy_fallthru_edge and tidy_control_flow.
12022 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
12024 PR rtl-optimization/83530
12026 * sel-sched.c (force_next_insn): New global variable.
12027 (remove_insn_for_debug): When force_next_insn is true, also leave only
12028 next insn in the ready list.
12029 (sel_sched_region): When the region wasn't scheduled, make another pass
12030 over it with force_next_insn set to 1.
12032 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12034 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
12036 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12037 for interrupt control.
12038 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12039 functions for interrupt control.
12040 * config/nds32/nds32-intrinsic.md: Likewise.
12041 * config/nds32/nds32_intrinsic.h: Likewise.
12042 * config/nds32/nds32.h (nds32_builtins): Likewise.
12044 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12046 * config/nds32/nds32.c (nds32_init_machine_status,
12047 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
12048 strict_aligned_p field.
12049 (nds32_expand_to_rtl_hook): New function.
12050 (TARGET_EXPAND_TO_RTL_HOOK): Define.
12051 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
12053 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12054 Chung-Ju Wu <jasonwucj@gmail.com>
12056 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
12057 * config/nds32/nds32-n7.md: New file.
12058 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
12059 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
12061 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
12062 * config/nds32/nds32.md (pipeline_model): Add n7.
12063 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
12064 * config/nds32/pipelines.md: Include n7 settings.
12066 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12067 Chung-Ju Wu <jasonwucj@gmail.com>
12069 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
12070 * config/nds32/nds32-e8.md: New file.
12071 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
12072 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
12074 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
12075 * config/nds32/nds32.md (pipeline_model): Add e8.
12076 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
12077 * config/nds32/pipelines.md: Include e8 settings.
12079 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12080 Chung-Ju Wu <jasonwucj@gmail.com>
12082 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
12083 * config/nds32/nds32-n8.md: New file.
12084 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
12085 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
12087 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
12088 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
12089 * config/nds32/nds32.md (pipeline_model): Add n8.
12090 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
12091 * config/nds32/pipelines.md: Include n8 settings.
12093 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
12094 Chung-Ju Wu <jasonwucj@gmail.com>
12096 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
12097 * config/nds32/nds32-n9-2r1w.md: New file.
12098 * config/nds32/nds32-n9-3r2w.md: New file.
12099 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
12100 nds32_register_ports): New or modify for cpu n9.
12101 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
12103 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
12104 * config/nds32/nds32-utils.c: New file.
12105 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
12106 TARGET_MUL_SLOW): Define.
12107 * config/nds32/nds32.md (pipeline_model): New attribute.
12108 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
12109 New options that support cpu n9.
12110 * config/nds32/pipelines.md: Include n9 settings.
12111 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
12113 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
12115 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
12116 information if necessary.
12117 (output_cond_branch_compare_zero): Likewise.
12118 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
12119 (nds32_target_alignment): Refine for alignment.
12120 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
12121 (FUNCTION_BOUNDARY): Modify.
12122 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
12124 * config/nds32/nds32.opt (malways-align, malign-functions): New.
12126 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
12128 * config/nds32/constants.md (unspec_volatile_element): Add values for
12129 TLB operation and data prefetch.
12130 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
12131 functions for TLB operation and data prefetch.
12132 * config/nds32/nds32-intrinsic.md: Likewise.
12133 * config/nds32/nds32_intrinsic.h: Likewise.
12134 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
12135 (nds32_print_operand): Likewise.
12136 * config/nds32/nds32.h (nds32_builtins): Likewise.
12138 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
12139 Andrew Pinski <pinsika@gcc.gnu.org>
12141 PR middle-end/82976
12142 * match.pd: Use constant_boolean_node of correct type instead of
12143 boolean_true_node or boolean_false_node for simplifying
12144 pointer comparisons to zero.
12146 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12148 PR tree-optimization/80021
12149 * tree.c (verify_type_variant): Make error call in verify_variant_match
12150 translatable and remove final full stop.
12152 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12154 * config/nds32/constants.md (unspec_volatile_element): Add
12155 UNSPEC_VOLATILE_EH_RETURN.
12156 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
12157 nds32_output_stack_pop): Support dwarf exception handling process.
12158 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
12159 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
12160 exception handling process.
12161 (nds32_compute_stack_frame): Likewise.
12162 (nds32_return_addr_rtx): Likewise.
12163 (nds32_initial_elimination_offset): Likewise.
12164 (nds32_expand_prologue): Likewise.
12165 (nds32_expand_epilogue): Likewise.
12166 (nds32_dynamic_chain_address): New function.
12167 * config/nds32/nds32.h (machine_function): Add fields for dwarf
12168 exception handling.
12169 (DYNAMIC_CHAIN_ADDRESS): Define.
12170 (EH_RETURN_DATA_REGNO): Define.
12171 (EH_RETURN_STACKADJ_RTX): Define.
12172 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
12173 patterns for dwarf exception handling.
12175 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12177 * config/nds32/nds32.h: Clean up obsolete macros.
12179 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12181 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12182 Add enum values for particular instructions.
12183 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12184 particular intrinsic functions.
12185 * config/nds32/nds32-intrinsic.md: Likewise.
12186 * config/nds32/nds32_intrinsic.h: Likewise.
12187 * config/nds32/nds32.h (nds32_builtins): Likewise.
12188 * config/nds32/nds32.md (type): Add pbsad and pbsada.
12189 (btst, ave): New patterns for particular instructions.
12191 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12193 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12194 Add enum values for atomic load/store and memory sync.
12195 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
12197 * config/nds32/nds32-intrinsic.md: Likewise.
12198 * config/nds32/nds32_intrinsic.h: Likewise.
12199 * config/nds32/nds32.h (nds32_builtins): Likewise.
12201 2018-04-07 Jakub Jelinek <jakub@redhat.com>
12203 PR tree-optimization/85257
12204 * fold-const.c (native_encode_vector): If not all elts could fit
12205 and off is -1, return 0 rather than offset.
12206 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
12207 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
12208 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
12209 adjust buffer in native_interpret_expr call.
12211 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12213 * config/nds32/constants.md (unspec_volatile_element): Add cache
12214 control enum values.
12215 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
12216 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
12217 * config/nds32/nds32.c (nds32_cctl_names): New.
12218 (nds32_print_operand): Handle cache control register names.
12219 * config/nds32/nds32.h (nds32_builtins): New enum values.
12220 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
12222 * config/nds32/nds32.md (type): Add mmu.
12223 * config/nds32/pipelines.md (simple_insn): Add mmu.
12225 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12227 * config/nds32/nds32.md (type): Remove call.
12228 * config/nds32/pipelines.md (simple_insn): Likewise.
12230 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12232 * config/nds32/constants.md (unspec_volatile_element): Add
12233 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
12234 UNSPEC_VOLATILE_FMFCFG.
12235 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
12236 description for fmfcfg and fmfcsr.
12237 (bdesc_1arg): Add fmtcsr.
12238 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
12239 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
12240 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
12241 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
12242 unspec_fmfcfg): New patterns.
12243 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
12244 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
12245 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
12246 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
12247 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
12248 __nds32__fmfcfg): Define.
12250 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
12252 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
12253 intrinsic register names.
12254 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
12255 intrinsic register enum values and macros.
12257 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
12259 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
12260 for load/store addressing form.
12261 (nds32_print_operand_address): Likewise.
12263 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
12266 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
12267 based on LABEL_REF. Remove useless assertion.
12268 (pic_address_needs_scratch): Fix formatting.
12269 (sparc_legitimize_pic_address): Minor tweaks.
12270 (sparc_delegitimize_address): Adjust assertion accordingly.
12271 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
12272 into symbolic_operand.
12273 (movsi_high_pic_label_ref): Likewise.
12274 (movsi_lo_sum_pic_label_ref): Likewise.
12275 (movdi_pic_label_ref): Likewise.
12276 (movdi_high_pic_label_ref): Likewise.
12277 (movdi_lo_sum_pic_label_ref): Likewise.
12279 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
12281 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
12282 custom LIB_SPEC setup.
12284 2018-04-06 Ruslan Bukin <br@bsdpad.com>
12285 Kito Cheng <kito.cheng@gmail.com>
12287 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
12288 * config/riscv/freebsd.h: New.
12290 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12292 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
12293 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
12296 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12297 Kito Cheng <kito.cheng@gmail.com>
12299 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
12300 nds32_output_call, nds32_symbol_binds_local_p): New functions.
12301 * config/nds32/nds32-protos.h (nds32_output_call,
12302 nds32_output_return): Declare.
12303 * config/nds32/nds32.md: Refine all the call and return patterns.
12305 2018-04-06 Jakub Jelinek <jakub@redhat.com>
12308 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
12309 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
12311 PR rtl-optimization/84872
12312 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
12313 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
12314 EDGE_CROSSING edge.
12316 2018-04-06 Tamar Christina <tamar.christina@arm.com>
12318 * expr.c (copy_blkmode_to_reg): Revert 254862.
12319 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
12321 2018-04-06 Richard Biener <rguenther@suse.de>
12323 PR middle-end/85244
12324 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
12325 after seeing a component reference with an adjacent field. Treat
12326 refs to arrays at struct end of external decls similar to
12327 refs to unconstrained commons.
12329 2018-04-06 Jakub Jelinek <jakub@redhat.com>
12332 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
12333 look through SAVE_EXPRs with non-side-effects argument. Adjust
12335 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
12338 2018-04-06 Richard Biener <rguenther@suse.de>
12340 PR middle-end/85180
12341 * alias.c (find_base_term): New wrapper around find_base_term
12342 unwinding CSELIB_VAL_PTR changes.
12343 (find_base_term): Do not restore CSELIB_VAL_PTR during the
12346 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12348 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
12350 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
12351 constant definitions.
12352 ("nop"): lr 0,0 -> nopr r0
12353 ("nop_lr0", "nop_lr1"): New insn definitions.
12355 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
12357 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
12358 NDS32_V3PUSH_AVAILABLE_P macro.
12360 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
12361 Chung-Ju Wu <jasonwucj@gmail.com>
12363 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
12364 (nds32*-*-*): Add float and fpu_config into supported_defaults.
12365 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
12366 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
12367 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
12368 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
12369 * config/nds32/constraints.md: New constraints and checking for hard
12370 float configuration.
12371 * config/nds32/iterators.md: New mode iterator and attribute for hard
12372 float configuration.
12373 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
12375 * config/nds32/nds32-fpu.md: New file.
12376 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
12377 deal with hard float code generation.
12378 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
12380 (abi_type, float_reg_number): New enum type.
12381 * config/nds32/nds32-predicates.c: New predicates for hard float.
12382 * config/nds32/nds32-protos.h: Declare functions for hard float.
12383 * config/nds32/nds32.c: Implementation for hard float configuration.
12384 * config/nds32/nds32.h: Definitions for hard float configuration.
12385 * config/nds32/nds32.md: Include hard float machine description and
12386 modify patterns for hard float configuration.
12387 * config/nds32/nds32.opt: New options for hard float configuration.
12388 * config/nds32/predicates.md: New predicates for hard float
12391 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12393 * common/config/nds32/nds32-common.c
12394 (nds32_option_optimization_table): Enable -mreleax-hint by default.
12396 2018-04-05 Jakub Jelinek <jakub@redhat.com>
12398 PR middle-end/85195
12399 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
12400 CONSTRUCTOR_ELT (ctor, ...)->value.
12402 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
12405 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
12407 2018-04-05 Tom de Vries <tom@codesourcery.com>
12410 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
12413 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12414 Kito Cheng <kito.cheng@gmail.com>
12416 * config/nds32/constraints.md (U33): Fine-tune checking condition.
12417 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
12418 * config/nds32/nds32.h (nds32_16bit_address_type): Add
12419 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
12421 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12422 Kito Cheng <kito.cheng@gmail.com>
12424 * config/nds32/constraints.md (Ufe): New memory constraint.
12425 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
12426 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
12427 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
12429 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
12430 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
12432 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12434 * config/nds32/nds32.md: Use optimize_size in the condition for
12435 alu-shift instructions.
12437 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12439 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
12441 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12443 * config/nds32/nds32.md (negsi2): Refine pattern.
12445 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12446 Chung-Ju Wu <jasonwucj@gmail.com>
12448 * config/nds32/iterators.md (shift_rotate): New code iterator.
12449 (shift): New code attribute.
12450 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
12451 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
12452 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
12453 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
12454 bit-wise operations.
12455 (andsi3, *andsi3): Ditto.
12456 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
12457 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
12458 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
12459 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
12460 nds32_ior_operand, nds32_xor_operand): New predicates.
12462 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12464 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
12465 (addsi3, subsi3): ... this.
12467 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12469 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
12471 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12473 * config/nds32/nds32.md: Adjust indention.
12475 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12477 * config/nds32/nds32.md (feature): New attribute.
12479 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12481 * config/nds32/nds32.md (subtype): New attribute.
12483 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
12486 * config/arm/arm-builtins.c (arm_expand_builtin): Change
12487 expansion to perform a bitwise AND of the argument followed by a
12488 boolean negation of the result.
12490 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
12492 PR rtl-optimization/84878
12493 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
12494 the basic block. Assert the use reference is not artificial and that
12495 it has an associated insn.
12497 2018-04-04 Michael Matz <matz@suse.de>
12499 * builtins.c (compute_objsize): Pass correct operand
12500 to array_at_struct_end_p.
12502 2018-04-04 Richard Biener <rguenther@suse.de>
12505 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
12506 from contexts for DINFO_LEVEL_TERSE and below.
12508 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12510 * config/nds32/nds32-doubleword.md (move_<mode>): Require
12511 resiter_operand condition.
12512 * config/nds32/nds32.md (*move<mode>): Ditto.
12514 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12515 Monk Chiang <sh.chiang04@gmail.com>
12517 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
12519 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12521 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
12523 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12524 Kito Cheng <kito.cheng@gmail.com>
12526 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
12527 nds32_cond_code_str, output_cond_branch,
12528 output_cond_branch_compare_zero, nds32_expand_cbranch,
12529 nds32_expand_cstore, nds32_expand_movcc,
12530 nds32_output_cbranchsi4_equality_zero,
12531 nds32_output_cbranchsi4_equality_reg,
12532 nds32_output_cbranchsi4_equality_reg_or_const_int,
12533 nds32_output_cbranchsi4_greater_less_zero: New functions.
12534 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
12535 nds32_expand_cstore, nds32_expand_movcc,
12536 nds32_output_cbranchsi4_equality_zero,
12537 nds32_output_cbranchsi4_equality_reg,
12538 nds32_output_cbranchsi4_equality_reg_or_const_int,
12539 nds32_output_cbranchsi4_greater_less_zero): Declare.
12540 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
12541 nds32_rimm11s_operand): New predicates.
12542 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
12543 * config/nds32/nds32.md: Rewrite all the branch and conditional move
12546 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12548 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
12549 * config/nds32/nds32.md: Ditto.
12550 * config/nds32/pipelines.md: Ditto.
12552 2018-04-04 Richard Biener <rguenther@suse.de>
12554 PR tree-optimization/85168
12555 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
12556 propagating abnormals.
12558 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12560 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
12562 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12563 Kito Cheng <kito.cheng@gmail.com>
12565 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
12566 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
12567 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
12568 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
12569 * config/nds32/nds32.md (sibcall_internal): New.
12570 (sibcall_register): Remove.
12571 (sibcall_immediate): Remove.
12572 (sibcall_value_internal): New.
12573 (sibcall_value_register): Remove.
12574 (sibcall_value_immediate): Remove.
12575 * config/nds32/predicates.md (nds32_general_register_operand): New.
12576 (nds32_call_address_operand): New.
12578 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12580 PR rtl-optimization/85167
12581 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
12582 bb_defs if *split_p, instead preinitialize it to NULL.
12584 PR tree-optimization/85156
12585 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
12586 evaluating the argument multiple times.
12588 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
12590 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
12592 (_mm_cvtpd_ps): Likewise.
12593 (_mm_cvttpd_epi32): Likewise.
12594 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
12595 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
12596 vector, pixel, and bool following altivec.h include.
12598 2018-04-03 Martin Sebor <msebor@redhat.com>
12600 * doc/extend.texi (Common Function Attributes): Clarify.
12601 (const attribute): Likewise.
12602 (pure attribute): Likewise.
12604 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12607 * config/i386/i386.c (ix86_expand_vector_set): Use
12608 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
12610 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
12612 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
12613 instructions when changing rounding bits to preserve precision bits
12614 in the x87 control word.
12616 2018-04-03 Martin Liska <mliska@suse.cz>
12618 PR tree-optimization/82491
12619 * rtl.h (strip_offset_and_add): Replace += suboffset with
12620 poly_uint64 () + suboffset.
12622 2018-03-29 Martin Liska <mliska@suse.cz>
12623 Martin Jambor <mjambor@suse.cz>
12626 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
12627 param_type is not an integral or pointer type.
12629 2018-04-03 Richard Biener <rguenther@suse.de>
12631 * sese.h (recompute_all_dominators): Remove.
12633 2018-04-02 Martin Sebor <msebor@redhat.com>
12635 * doc/invoke.texi (-Wrestrict): Fix typos.
12637 2018-04-02 Jim Wilson <jimw@sifive.com>
12639 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
12640 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
12641 (<optab>di3, <optab>si3_extend): Likewise.
12642 (<optab>si3_mask, <optab>si3_mask_1): New.
12643 (<optab>di3_mask, <optab>di3_mask_1): New.
12644 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
12645 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
12646 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
12648 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
12650 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
12653 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
12655 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
12656 (nds32_canonicalize_comparison): New function.
12658 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12659 Kito Cheng <kito.cheng@gmail.com>
12660 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12662 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
12663 * config/nds32/constants.md (unspec_volatile_element): Add
12664 UNSPEC_VOLATILE_RELAX_GROUP.
12665 * config/nds32/nds32-relax-opt.c: New file.
12666 * config/nds32/nds32-predicates.c
12667 (nds32_symbol_load_store_p): New function.
12668 * config/nds32/nds32-protos.h
12669 (nds32_symbol_load_store_p): Declare function.
12670 (make_pass_nds32_relax_opt): Declare new rtl pass function.
12671 * config/nds32/nds32.c
12672 (nds32_register_pass): New function to register pass.
12673 (nds32_register_passes): New function to register passes.
12674 * config/nds32/nds32.md (relax_group): New pattern.
12675 * config/nds32/nds32.opt (mrelax-hint): New option.
12676 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
12678 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
12680 * config/nds32/t-nds32: Modify files dependency.
12682 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12684 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
12685 (PROFILE_HOOK): Define its implementation.
12687 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12689 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
12690 type and 32-bit size.
12692 2018-04-01 Jakub Jelinek <jakub@redhat.com>
12694 PR middle-end/85090
12695 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
12696 (V_128_256): New mode iterator.
12697 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
12698 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
12699 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
12701 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
12702 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
12704 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
12707 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
12708 NaN inputs correctly.
12710 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
12713 * config/rs6000/vsx.md (??r): New mode attribute.
12714 (*vsx_mov<mode>_64bit): Use it.
12715 (*vsx_mov<mode>_32bit): Likewise.
12717 2018-03-30 Martin Sebor <msebor@redhat.com>
12719 PR tree-optimization/84818
12720 * builtins.c (check_access): Use warning_n.
12722 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12725 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
12727 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
12730 2018-03-30 Julia Koval <julia.koval@intel.com>
12733 * x86-tune.def (movx, partial_reg_dependency): Enable for
12736 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
12738 PR inline-asm/84985
12739 * lra-constraints.c (process_alt_operands): Move setting
12740 this_alternative_matches below.
12742 2018-03-29 Martin Liska <mliska@suse.cz>
12745 * doc/invoke.texi: Document how LTO works with debug info.
12746 Describe auto-load support of binutils. Mention 'x86-64'
12747 as valid option value of -march option.
12749 2018-03-29 Jakub Jelinek <jakub@redhat.com>
12751 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
12754 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
12755 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
12756 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
12759 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12762 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
12763 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
12764 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
12765 for RS6000_BTM_POWERPC64.
12766 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
12767 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
12768 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
12771 (DIVDEU): Likewise.
12773 2018-03-28 Carl Love <cel@us.ibm.com>
12776 2017-09-27 Carl Love <cel@us.ibm.com>
12778 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
12779 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
12780 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
12783 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12785 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
12786 instead of __vector bool.
12787 (_mm_max_pu8): Likewise.
12788 (_mm_min_pi16): Likewise.
12790 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12793 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
12794 (DIVWEUO): Likewise.
12795 (DIVDEO): Likewise.
12796 (DIVDEUO): Likewise.
12797 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
12798 DIVWEUO and DIVDEUO.
12799 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
12800 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
12801 (div_extend): Likewise.
12802 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
12804 (__builtin_divweuo): Likewise.
12805 (__builtin_divdeo): Likewise.
12806 (__builtin_divdeuo): Likewise.
12808 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12811 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
12812 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
12814 PR tree-optimization/82004
12815 * gimple-match-head.c (optimize_pow_to_exp): New function.
12816 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
12817 Don't fold to exp if optimize_pow_to_exp is false.
12819 2018-03-28 Martin Liska <mliska@suse.cz>
12822 * calls.c (initialize_argument_information): Fix trailing space.
12823 * common.opt: Fix typo and provide better explanation for
12824 -fsanitize-coverage option.
12825 * config/i386/i386.opt: Fix typo.
12827 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12828 Martin Liska <mliska@suse.cz>
12831 * gimplify.c (asan_poison_variable): Don't do the check for
12832 gimplify_omp_ctxp here.
12833 (gimplify_decl_expr): Do it here.
12834 (gimplify_target_expr): Likewise.
12836 2018-03-28 Martin Liska <mliska@suse.cz>
12839 * config/i386/i386.c (ix86_function_arg_advance): Do not call
12840 chkp_type_bounds_count if MPX is not enabled.
12842 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
12844 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
12846 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
12849 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
12850 function to create the function decl for complex long double
12851 multiply and divide for -mabi=ieeelongdouble.
12852 (init_float128_ieee): Call it.
12854 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
12857 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
12858 -fcf-protection=branch -mibt.
12859 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
12861 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12864 * config/arm/arm.c (arm_valid_symbolic_address): Handle
12865 arm_word_relocations.
12867 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
12870 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
12871 extern array declarations.
12873 2018-03-27 Richard Biener <rguenther@suse.de>
12875 PR middle-end/84067
12876 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
12877 explicit single_use checks.
12879 2018-03-27 Richard Biener <rguenther@suse.de>
12881 PR tree-optimization/85082
12882 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
12885 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12887 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
12888 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
12889 Turn on fasynchronous-unwind-tables and funwind-tables.
12891 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
12894 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
12895 (*bmi_blsr_<mode>_ccz): Ditto.
12897 2018-03-26 Tom de Vries <tom@codesourcery.com>
12899 PR tree-optimization/85063
12900 * omp-general.c (offloading_function_p): New function. Factor out
12902 * omp-offload.c (pass_omp_target_link::gate): ... here.
12903 * omp-general.h (offloading_function_p): Declare.
12904 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
12905 with attribute omp declare target for offloading functions.
12907 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
12909 PR tree-optimization/84005
12910 * tree-data-ref.h (get_base_for_alignment): Declare.
12911 * tree-data-ref.c (get_base_for_alignment_1): New function.
12912 (get_base_for_alignment): Likewise.
12913 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
12914 get_base_for_alignment to find a suitable base object, instead
12915 of always using drb->base_address.
12917 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12919 PR inline-asm/85022
12920 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
12921 known size by default.
12923 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
12925 PR inline-asm/85030
12926 * lra-constraints.c (process_alt_operands): Don't match BLKmode
12927 and non BLKmode operands.
12929 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12932 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
12933 Clean up attributes.
12935 2018-03-23 Richard Biener <rguenther@suse.de>
12938 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
12939 we are going to emit early debug for LTO.
12941 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12943 PR inline-asm/85034
12944 * function.c (match_asm_constraints_1): Don't optimize if input
12945 doesn't satisfy general_operand predicate for output's mode.
12947 PR inline-asm/85022
12948 * alias.c (write_dependence_p): Don't require for x_canonicalized
12949 non-VOIDmode if x has VOIDmode.
12952 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
12953 just don't try to optimize it rather than assert it never happens.
12955 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12957 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
12958 macro expansions for definition of ST_INTERNAL_<mode> and
12959 LD_INTERNAL_<mode> builtins.
12960 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
12962 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
12964 (altivec_expand_st_builtin): Likewise.
12965 (altivec_expand_builtin): Remove calls to deleted functions.
12966 (rs6000_address_for_altivec): Delete this function.
12967 * config/rs6000/vector.md: Remove expands for
12968 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
12970 2018-03-22 Sudakshina Das <sudi.das@arm.com>
12973 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
12974 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
12975 re-computing once computed.
12976 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
12977 (arm_init_machine_status): Initialize
12978 machine->static_chain_stack_bytes.
12980 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12983 * doc/extend.texi: Add four new prototypes for vec_ld.
12984 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
12985 definitions for more logical presentation.
12986 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
12987 entries for V1TI variants of __builtin_altivec_ld builtin.
12988 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
12989 handling of V1TI variant of LVX icode pattern.
12990 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
12991 (rs6000_gimple_fold_builtin): Likewise.
12992 (altivec_init_builtins): Add code to define
12993 __builtin_altivec_lvx_v1ti function.
12995 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12997 PR inline-asm/84941
12998 * function.c (match_asm_constraints_1): Don't do the optimization
12999 if input isn't a REG, SUBREG, MEM or constant.
13001 2018-03-22 Tom de Vries <tom@codesourcery.com>
13003 PR tree-optimization/84956
13004 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
13005 bb_has_abnormal_pred.
13007 2018-03-22 Jakub Jelinek <jakub@redhat.com>
13010 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
13011 DECL_INITIAL (decl) to decl at the end.
13012 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
13013 adjust the comment.
13015 2018-03-21 Joseph Myers <joseph@codesourcery.com>
13017 * doc/extend.texi (__builtin_tgmath): Document when complex
13018 integer types are treated as _Complex _Float64.
13020 2018-03-21 Tom de Vries <tom@codesourcery.com>
13022 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
13024 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13026 PR tree-optimization/84960
13027 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
13028 if it is ENTRY block, move them into single succ of ENTRY in that case.
13030 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
13032 PR tree-optimization/84811
13033 * poly-int.h (poly_span_traits): Remove the T3 parameter and
13034 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
13035 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
13036 (known_subrange_p): Update accordingly. Cast each value involved
13037 in the size comparison, rather than casting the result of the
13040 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13042 PR tree-optimization/84982
13043 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
13044 by flipping the least significant bit rather than all bits from
13045 bitpos to bitpos + bitsize - 1.
13047 2018-03-21 Nathan Sidwell <nathan@acm.org>
13049 * doc/extend.texi (Deprecated Features): Remove mention of
13050 long-deleted deprecations.
13052 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13055 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
13056 * configure: Regenerate.
13058 2018-03-21 Tom de Vries <tom@codesourcery.com>
13060 PR tree-optimization/83126
13061 * tree-parloops.c (num_phis): New function.
13062 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
13064 2018-03-21 Nathan Sidwell <nathan@acm.org>
13066 * doc/extend.texi (Deprecated Features): Update deprecated flags,
13067 mention anon-struct/union members and trailing attributes.
13069 2018-03-21 Bin Cheng <bin.cheng@arm.com>
13071 PR tree-optimization/84969
13072 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
13073 builtin memset partitions if they set different rhs values.
13075 2018-03-21 Jakub Jelinek <jakub@redhat.com>
13077 PR rtl-optimization/84989
13078 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
13079 VEC_DUPLICATE with scalar result mode.
13081 2018-03-21 Martin Liska <mliska@suse.cz>
13084 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
13085 not intended return statement.
13087 2018-03-21 Martin Liska <mliska@suse.cz>
13090 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
13091 (chkp_find_bound_slots_1): Limit number of iterations.
13093 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
13096 * Minor grammar fixes for x86 options.
13098 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13101 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
13102 holding REG_CFA_RESTORE notes, instead turn them into a USE.
13104 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
13107 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
13108 (altivec_lvx_<mode>_1op): Likewise.
13109 (altivec_stvx_<mode>_2op): Likewise.
13110 (altivec_stvx_<mode>_1op): Likewise.
13111 (altivec_lvx_<VM2:mode>): New define_expand.
13112 (altivec_stvx_<VM2:mode>): Likewise.
13113 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
13114 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13115 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
13116 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
13117 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
13118 (rs6000_gen_lvx): Likewise.
13119 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
13120 (altivec_expand_stv_builtin): Likewise.
13121 (altivec_expand_builtin): Likewise.
13122 * config/rs6000/vector.md: Likewise.
13124 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13127 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
13130 2018-03-20 Richard Biener <rguenther@suse.de>
13133 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
13134 sign-conversions as zero, fall back to standard scalar_stmt
13137 2018-03-20 Martin Liska <mliska@suse.cz>
13140 * predict.c (rebuild_frequencies): Handle case when we have
13141 PROFILE_ABSENT, but flag_guess_branch_prob is false.
13143 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13146 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
13147 flag_section_anchors.
13148 * varasm.c (use_blocks_for_decl_p): Remove hack for
13149 dw2_force_const_mem.
13152 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
13154 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
13155 be created, use lowpart_subreg of operands[0] rather than operands[0]
13157 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
13158 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
13159 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
13160 and n constraint instead of aarch64_shift_imm_di and Usd.
13161 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
13162 (*aarch64_<optab>_reg_minus<mode>3): ... this.
13164 2018-03-20 Sudakshina Das <sudi.das@arm.com>
13167 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
13168 to favor GPR over NEON registers.
13169 (<shift>di3_neon): Likewise.
13171 2018-03-20 Tom de Vries <tom@codesourcery.com>
13174 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
13175 (nvptx_process_pars): Emit bar.sync asap and alap.
13177 2018-03-20 Tom de Vries <tom@codesourcery.com>
13180 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
13181 seen_label if seen_label is already set.
13183 2018-03-20 Jakub Jelinek <jakub@redhat.com>
13186 * config/i386/i386.c (fold_builtin_cpu): For features above 31
13187 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
13188 Use 1U instead of 1. Formatting fixes.
13191 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
13192 instead of TREE_TYPE (s1) for the return value.
13194 2018-03-19 Jakub Jelinek <jakub@redhat.com>
13196 PR tree-optimization/84946
13197 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
13198 bitsize + bitsize in poly_uint64 rather than poly_int64.
13201 * dwarf2asm.c: Include fold-const.c.
13202 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
13203 of decl rather than decl itself.
13205 PR rtl-optimization/84643
13206 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
13208 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
13211 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
13212 calling assemble_variable.
13214 2018-03-19 Sudakshina Das <sudi.das@arm.com>
13217 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
13218 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
13220 2018-03-19 Jim Wilson <jimw@sifive.com>
13223 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
13224 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
13225 (riscv_first_stack_step): Likewise.
13226 (riscv_option_override): Use STACK_BOUNDARY instead of
13227 MIN_STACK_BOUNDARY.
13228 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
13229 MIN_STACK_BOUNDARY.
13230 (BIGGEST_ALIGNMENT): Set to 128.
13231 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
13232 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
13235 2018-03-19 Richard Biener <rguenther@suse.de>
13237 PR tree-optimization/84933
13238 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
13239 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
13241 2018-03-19 Richard Biener <rguenther@suse.de>
13243 PR tree-optimization/84859
13244 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
13245 (cond_if_else_store_replacement): Perform sinking operation on
13246 single-store BBs regardless of MAX_STORES_TO_SINK setting.
13247 Generalize what a BB with a single eligible store is.
13249 2018-03-19 Richard Biener <rguenther@suse.de>
13251 PR tree-optimization/84929
13252 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
13253 chrec_is_positive against non-chrec arg.
13255 2018-03-19 Tamar Christina <tamar.christina@arm.com>
13258 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
13260 2018-03-18 Martin Liska <mliska@suse.cz>
13262 PR rtl-optimization/84635
13263 * regrename.c (build_def_use): Use matches_mode only when
13266 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
13268 PR tree-optimization/84913
13269 * tree-vect-loop.c (vectorizable_reduction): Don't try to
13270 vectorize chains of COND_EXPRs.
13272 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13274 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
13276 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13278 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
13280 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
13282 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
13284 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
13285 Kito Cheng <kito.cheng@gmail.com>
13287 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
13288 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
13289 (nds32_adjust_reg_alloc_order): New function.
13290 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
13292 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
13294 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
13295 nds32_print_operand, nds32_print_operand_address): Use
13296 HOST_WIDE_INT_PRINT_DEC instead.
13298 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
13300 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
13302 2018-03-17 Jakub Jelinek <jakub@redhat.com>
13305 * config/i386/i386.c (initial_ix86_tune_features,
13306 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
13307 unsigned long long.
13308 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
13309 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
13310 rather than 1u << ix86_tune. Formatting fix.
13311 (ix86_option_override_internal): Change ix86_arch_mask from
13312 unsigned int to unsigned HOST_WIDE_INT, initialize to
13313 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
13314 (ix86_function_specific_restore): Likewise.
13316 2018-03-16 Jakub Jelinek <jakub@redhat.com>
13319 * postreload.c (reload_combine_recognize_pattern): Perform
13320 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
13321 truncate_int_for_mode the result for the destination's mode.
13324 * hsa-gen.c (mem_type_for_type): Fix comment typo.
13325 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
13327 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13330 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
13333 * lra-assigns.c (lra_split_hard_reg_for): Don't use
13334 regno_allocno_class_array and sorted_pseudos.
13335 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
13336 insns where regno is used.
13338 2018-03-16 Martin Liska <mliska@suse.cz>
13341 * multiple_target.c (create_dispatcher_calls): Redirect
13342 reference in the symbol table.
13344 2018-03-16 Martin Liska <mliska@suse.cz>
13347 * multiple_target.c (create_dispatcher_calls): Redirect also
13350 2018-03-16 Jakub Jelinek <jakub@redhat.com>
13354 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
13355 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
13356 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
13358 2018-03-16 Julia Koval <julia.koval@intel.com>
13360 * doc/invoke.texi (Skylake Server): Add CLWB.
13361 Cannonlake): Remove CLWB.
13363 2018-03-16 Jakub Jelinek <jakub@redhat.com>
13365 PR tree-optimization/84841
13366 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
13368 (FLOAT_ONE_CONST_TYPE): Define.
13369 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
13370 (sort_by_operand_rank): Put entries with higher constant_type last
13371 rather than first to match comments.
13373 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
13375 * config/nios2/nios2.md (movsi_internal): Fix thinko in
13378 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13381 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
13382 check and use address of target always.
13384 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
13387 * config/i386/i386.c (indirect_thunk_needed): Update comments.
13388 (indirect_thunk_bnd_needed): Likewise.
13389 (indirect_thunks_used): Likewise.
13390 (indirect_thunks_bnd_used): Likewise.
13391 (indirect_return_needed): New.
13392 (indirect_return_bnd_needed): Likewise.
13393 (output_indirect_thunk_function): Add a bool argument for
13395 (output_indirect_thunk_function): Don't generate alias for
13396 function return thunk.
13397 (ix86_code_end): Call output_indirect_thunk_function to generate
13398 function return thunks.
13399 (ix86_output_function_return): Set indirect_return_bnd_needed
13400 and indirect_return_needed instead of indirect_thunk_bnd_needed
13401 and indirect_thunk_needed.
13403 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
13405 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
13406 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
13407 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
13409 2018-03-15 David Malcolm <dmalcolm@redhat.com>
13410 Paul Hua <paul.hua.gm@gmail.com>
13413 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
13415 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
13417 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
13418 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
13419 resp. SFmode cases.
13421 2018-03-15 Tamar Christina <tamar.christina@arm.com>
13424 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13425 instead of GET_MODE_SIZE when comparing Units.
13427 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13430 * varasm.c (hash_section): Return an unchangeble hash value
13431 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
13432 Return !aarch64_can_use_per_function_literal_pools_p ().
13434 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13437 * optabs.c (emit_conditional_move): Pass address of cmode's copy
13438 rather than address of cmode as last argument to prepare_cmp_insn.
13440 2018-03-15 Julia Koval <julia.koval@intel.com>
13442 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
13443 F_AVX512VNNI, F_AVX512BITALG): New.
13445 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
13448 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
13449 insn for floating-point loads and stores.
13451 2018-03-14 Carl Love <cel@us.ibm.com>
13453 * config/rs6000/rs6000-c.c: Add macro definitions for
13454 ALTIVEC_BUILTIN_VEC_PERMXOR.
13455 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
13456 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
13457 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
13458 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
13460 * config/doc/extend.texi: Add prototypes for vec_permxor.
13462 2018-03-14 David Malcolm <dmalcolm@redhat.com>
13465 * diagnostic-show-locus.c (class layout_point): Convert m_line
13466 from int to linenum_type.
13467 (line_span::comparator): Use linenum "compare" function when
13468 comparing line numbers.
13469 (test_line_span): New function.
13470 (layout_range::contains_point): Convert param "row" from int to
13472 (layout_range::intersects_line_p): Likewise.
13473 (layout::will_show_line_p): Likewise.
13474 (layout::print_source_line): Likewise.
13475 (layout::should_print_annotation_line_p): Likewise.
13476 (layout::print_annotation_line): Likewise.
13477 (layout::print_leading_fixits): Likewise.
13478 (layout::annotation_line_showed_range_p): Likewise.
13479 (struct line_corrections): Likewise for field m_row.
13480 (line_corrections::line_corrections): Likewise for param "row".
13481 (layout::print_trailing_fixits): Likewise.
13482 (layout::get_state_at_point): Likewise.
13483 (layout::get_x_bound_for_row): Likewise.
13484 (layout::print_line): Likewise.
13485 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
13486 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
13487 * input.c (selftest::test_linenum_comparisons): New function.
13488 (selftest::input_c_tests): Call it.
13489 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
13490 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
13491 * selftest.h (ASSERT_GT): New macro.
13492 (ASSERT_GT_AT): New macro.
13493 (ASSERT_LT): New macro.
13494 (ASSERT_LT_AT): New macro.
13496 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
13498 PR rtl-optimization/84780
13499 * combine.c (distribute_links): Don't make a link based on pc_rtx.
13501 2018-03-14 Martin Liska <mliska@suse.cz>
13503 * tree.c (record_node_allocation_statistics): Use
13504 get_stats_node_kind.
13505 (get_stats_node_kind): New function extracted from
13506 record_node_allocation_statistics.
13507 (free_node): Use get_stats_node_kind.
13509 2018-03-14 Richard Biener <rguenther@suse.de>
13511 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
13512 that the value-set of ANTIC_IN doesn't grow.
13515 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13517 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13518 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13520 2018-03-14 Julia Koval <julia.koval@intel.com>
13522 * config.gcc (icelake-client, icelake-server): New.
13524 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
13525 (initial_ix86_arch_features): Ditto.
13526 (PTA_SKYLAKE): Add SGX.
13527 (PTA_ICELAKE): Remove.
13528 (PTA_ICELAKE_CLIENT): New.
13529 (PTA_ICELAKE_SERVER): New.
13530 (ix86_option_override_internal): Split up icelake on icelake client and
13532 (get_builtin_code_for_version): Ditto.
13533 (fold_builtin_cpu): Ditto.
13534 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
13535 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
13536 * config/i386/i386.h (processor_type): Ditto.
13537 * doc/invoke.texi: Ditto.
13539 2018-03-14 Jakub Jelinek <jakub@redhat.com>
13542 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
13543 INTEGER_CST offset, add it together with bitpos / 8 and
13544 sign extend based on POINTER_SIZE.
13548 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
13551 * config/i386/constraints.md (Yc): New register constraint.
13552 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
13553 Use Yc constraint for alternative 2 of operand 0. Remove
13554 preferred_for_speed attribute.
13556 2018-03-14 Richard Biener <rguenther@suse.de>
13558 PR tree-optimization/84830
13559 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
13560 with the old one to avoid oscillations.
13562 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
13565 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
13567 (assign_by_spills): Return a flag of reload assignment failure.
13568 Do not process the reload assignment failures. Do not spill other
13569 reload pseudos if they has the same reg class. Update n if
13571 (lra_assign): Add a return arg. Set up from the result of
13572 assign_by_spills call.
13573 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13574 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13575 usage_insns if it is not NULL.
13576 (spill_hard_reg_in_range): New function.
13577 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13578 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13579 function prototypes.
13580 (lra_assign): Change prototype.
13581 * lra.c (lra): Add code to deal with fails by splitting hard reg
13584 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
13586 * config/riscv/riscv.opt (mrelax): New option.
13587 * config/riscv/riscv.c (riscv_file_start): Emit ".option
13588 "norelax" when riscv_mrelax is disabled.
13589 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
13591 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13594 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
13595 reassociation for int modes.
13597 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13599 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
13600 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
13602 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
13603 * config/aarch64/aarch64-sve.md
13604 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
13605 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
13606 (*extend<mode><Vwide>2): Rename to...
13607 (aarch64_sve_extend<mode><Vwide>2): ...this.
13608 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
13609 renaming the old pattern to...
13610 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
13612 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
13613 define_expand, renaming the old pattern to...
13614 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
13615 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
13616 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
13617 account when deciding which SVE instruction the optab should use.
13618 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
13620 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13622 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
13623 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
13624 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
13625 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
13626 (tlsdesc_small_<mode>): Turn a define_expand and use
13627 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
13628 (tlsdesc_small_advsimd_<mode>): ...this.
13629 (tlsdesc_small_sve_<mode>): New pattern.
13631 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13633 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
13634 (UNSPEC_UMUL_HIGHPART): New constants.
13635 (MUL_HIGHPART): New int iteraor.
13636 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
13637 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
13639 (*<su>mul<mode>3_highpart): New define_insn.
13641 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13644 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13647 2018-03-13 Martin Liska <mliska@suse.cz>
13650 * (sem_item_optimizer::sem_item_optimizer): Initialize new
13652 (sem_item_optimizer::~sem_item_optimizer): Release it.
13653 (sem_item_optimizer::merge_classes): Register variable aliases.
13654 (sem_item_optimizer::fixup_pt_set): New function.
13655 (sem_item_optimizer::fixup_points_to_sets): Likewise.
13656 * ipa-icf.h: Declare new variables and functions.
13658 2018-03-13 Jakub Jelinek <jakub@redhat.com>
13660 PR middle-end/84834
13661 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
13662 integer_pow2p@2 and test integer_pow2p in condition.
13663 (A < 0 ? C : 0): Similarly for @1.
13665 PR middle-end/84831
13666 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
13667 characters starting at p contain '\0' character, don't look beyond
13671 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
13672 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
13675 * reg-stack.c (change_stack): Change update_end var from int to
13676 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
13677 also call set_block_for_insn on the newly added insns and rescan.
13680 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
13681 on the last operand.
13684 * tree.c (stabilize_reference_1): Return save_expr (e) for
13685 STATEMENT_LIST even if it doesn't have side-effects.
13687 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
13689 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
13691 2018-03-12 Renlin Li <renlin.li@arm.com>
13693 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
13694 aarch64_output_scalar_simd_mov_immediate.
13696 2018-03-12 Martin Sebor <msebor@redhat.com>
13698 PR tree-optimization/83456
13699 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
13700 for perfectly overlapping calls to memcpy.
13701 (gimple_fold_builtin_memory_chk): Same.
13702 (gimple_fold_builtin_strcpy): Handle no-warning.
13703 (gimple_fold_builtin_stxcpy_chk): Same.
13704 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
13706 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13708 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
13709 parameter. Use it for SFmode.
13710 (rs6000_function_arg_advance_1): Adjust.
13711 (rs6000_function_arg): Adjust.
13712 (rs6000_gimplify_va_arg): Pass false for that new parameter.
13714 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13716 PR rtl-optimization/84169
13717 PR rtl-optimization/84780
13718 * combine.c (can_combine_p): Check for a 2-insn combination whether
13719 the destination register is used between the two insns, too.
13721 2018-03-12 Richard Biener <rguenther@suse.de>
13723 PR tree-optimization/84803
13724 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
13725 for refs DR analysis didn't process.
13727 2018-03-12 Richard Biener <rguenther@suse.de>
13729 PR tree-optimization/84777
13730 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
13731 force-vectorize loops ignore whether we are optimizing for size.
13733 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
13735 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
13736 (TARGET_MD_ASM_ADJUST): Define.
13738 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
13739 Kito Cheng <kito.cheng@gmail.com>
13740 Chung-Ju Wu <jasonwucj@gmail.com>
13742 * config/nds32/nds32.c (nds32_compute_stack_frame,
13743 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
13744 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
13745 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
13746 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
13747 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
13748 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
13749 * config/nds32/nds32.md (prologue, epilogue): Use macro
13750 NDS32_V3PUSH_AVAILABLE_P to do checking.
13752 2018-03-11 Jakub Jelinek <jakub@redhat.com>
13755 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
13756 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
13757 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
13758 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
13759 addition of most attributes on !orig_type_die or the attribute not
13760 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
13762 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13763 Chung-Ju Wu <jasonwucj@gmail.com>
13765 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
13766 __NDS32_VH__ macro.
13767 * config/nds32/nds32.opt (mvh): New option.
13769 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13770 Chung-Ju Wu <jasonwucj@gmail.com>
13772 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
13774 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
13775 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
13778 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13779 Chung-Ju Wu <jasonwucj@gmail.com>
13781 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
13783 * config/nds32/nds32-multiple.md (strlensi): New pattern.
13784 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
13786 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
13787 Kito Cheng <kito.cheng@gmail.com>
13788 Chung-Ju Wu <jasonwucj@gmail.com>
13790 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
13791 UNSPEC_FFMISM and UNSPEC_FLMISM.
13792 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
13793 for ffb, ffmism and flmism.
13794 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
13795 (unspec_ffmism): Ditto.
13796 (unspec_flmism): Ditto.
13797 (nds32_expand_builtin_impl): Check if string extension is available.
13798 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
13799 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
13801 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
13804 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13807 * lra-assigns.c (assign_by_spills): Return a flag of reload
13808 assignment failure. Do not process the reload assignment
13809 failures. Do not spill other reload pseudos if they has the same
13811 (lra_assign): Add a return arg. Set up from the result of
13812 assign_by_spills call.
13813 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13814 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13815 usage_insns if it is not NULL.
13816 (spill_hard_reg_in_range): New function.
13817 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13818 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13819 function prototypes.
13820 (lra_assign): Change prototype.
13821 * lra.c (lra): Add code to deal with fails by splitting hard reg
13824 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
13827 * config/i386/i386.opt: Replace Enforcment with Enforcement.
13829 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
13832 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
13833 (dw_val_node): Add val_symbolic_view.
13834 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
13835 (symview_upper_bound): New.
13836 (new_line_info_table): Initialize symviews_since_reset.
13837 (dwarf2out_source_line): Count symviews_since_reset and set
13838 symview_upper_bound.
13839 (dw_val_equal_p): Handle symview.
13840 (add_AT_symview): New.
13841 (print_dw_val): Handle symview.
13842 (attr_checksum, attr_checksum_ordered): Likewise.
13843 (same_dw_val_p, size_of_die): Likewise.
13844 (value_format, output_die): Likewise.
13845 (add_high_low_attributes): Use add_AT_symview for entry_view.
13846 (dwarf2out_finish): Reset symview_upper_bound, clear
13849 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
13852 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
13853 Add strict argument and use it.
13854 (rs6000_split_multireg_move): Update for new strict argument.
13855 (mem_operand_gpr): Disallow all non-offsettable addresses.
13856 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
13858 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13861 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
13862 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
13863 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
13866 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
13867 decl, use remap_type if we want to use the type.
13869 2018-03-09 Martin Sebor <msebor@redhat.com>
13871 PR tree-optimization/84526
13872 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13874 (builtin_access::generic_overlap): Be prepared to handle non-array
13877 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
13879 PR rtl-optimization/84682
13880 * lra-constraints.c (process_address_1): Check is_address flag
13881 for address constraints.
13882 (process_alt_operands): Likewise.
13883 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
13884 preprocess_constraints.
13885 * recog.h (preprocess_constraints): Add oploc parameter.
13887 * recog.c (preprocess_constraints): Test address_operand for
13888 CT_ADDRESS constraints.
13890 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13893 * lra-assigns.c (assign_by_spills): Return a flag of reload
13894 assignment failure. Do not process the reload assignment
13895 failures. Do not spill other reload pseudos if they has the same
13897 (lra_assign): Add a return arg. Set up from the result of
13898 assign_by_spills call.
13899 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13900 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13901 usage_insns if it is not NULL.
13902 (spill_hard_reg_in_range): New function.
13903 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13904 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13905 function prototypes.
13906 (lra_assign): Change prototype.
13907 * lra.c (lra): Add code to deal with fails by splitting hard reg
13910 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13913 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
13914 Accept complain bool parameter. Only emit errors if it is true.
13915 (arm_parse_cpu_option_name): Likewise.
13916 (arm_target_thumb_only): Adjust callers of the above.
13917 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
13918 prototype to take a default true bool parameter.
13919 (arm_parse_arch_option_name): Likewise.
13921 2018-03-09 David Malcolm <dmalcolm@redhat.com>
13922 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
13926 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
13927 * configure: Regenerate.
13928 * configure.ac ("linker --version-script option"): New.
13929 ("linker soname option"): New.
13931 2018-03-09 Richard Biener <rguenther@suse.de>
13933 PR tree-optimization/84775
13934 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
13935 immediate uses of predicate stmts and mark them modified.
13938 PR tree-optimization/84178
13939 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13941 (version_loop_for_if_conversion): Delay update_ssa call.
13942 (tree_if_conversion): Delay update_ssa until after predicate
13945 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
13948 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
13949 when the function accesses prior frames.
13951 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13954 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
13955 gen_llsym, otherwise call maybe_gen_llsym.
13957 PR inline-asm/84742
13958 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
13959 has ',' character inside of it.
13961 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13964 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
13965 as clobbering CC_REGNUM.
13967 2018-03-08 Richard Biener <rguenther@suse.de>
13969 PR middle-end/84552
13970 * tree-scalar-evolution.c: Include tree-into-ssa.h.
13971 (follow_copies_to_constant): Do not follow SSA names registered
13974 2018-03-08 Richard Biener <rguenther@suse.de>
13976 PR tree-optimization/84178
13977 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13979 (version_loop_for_if_conversion): Delay update_ssa call.
13980 (tree_if_conversion): Delay update_ssa until after predicate
13983 2018-03-08 David Malcolm <dmalcolm@redhat.com>
13985 PR tree-optimization/84178
13986 * tree-if-conv.c (release_bb_predicate): Remove the
13987 the assertion that the stmts have NULL use_ops.
13988 Discard the statements, asserting that they haven't
13989 yet been added to a BB.
13991 2018-03-08 Richard Biener <rguenther@suse.de>
13993 PR tree-optimization/84746
13994 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
13995 (phi_translate): Pass in destination ANTIC_OUT set.
13996 (phi_translate_1): Likewise. For a simplified result lookup
13997 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
13998 (phi_translate_set): Adjust.
13999 (do_pre_regular_insertion): Likewise.
14000 (do_pre_partial_partial_insertion): Likewise.
14002 2018-03-08 Martin Liska <mliska@suse.cz>
14004 PR gcov-profile/84735
14005 * doc/gcov.texi: Document usage of profile files.
14006 * gcov-io.h: Document changes in the format.
14008 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
14012 * dwarf2out.c (struct dw_line_info_table): Update comments for
14014 (FORCE_RESET_NEXT_VIEW): New.
14015 (FORCE_RESETTING_VIEW_P): New.
14016 (RESETTING_VIEW_P): Check for -1 too.
14017 (ZERO_VIEW_P): Likewise.
14018 (new_line_info_table): Force-reset next view.
14019 (dwarf2out_begin_function): Likewise.
14020 (dwarf2out_source_line): Simplify zero_view_p initialization.
14021 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
14022 view directly. Omit view when omitting .loc at line 0.
14024 2018-03-08 Jakub Jelinek <jakub@redhat.com>
14026 PR tree-optimization/84740
14027 * tree-switch-conversion.c (process_switch): Call build_constructors
14028 only if info.phi_count is non-zero.
14030 PR tree-optimization/84739
14031 * tree-tailcall.c (find_tail_calls): Check call arguments against
14032 DECL_ARGUMENTS (current_function_decl) rather than
14033 DECL_ARGUMENTS (func) when checking for tail recursion.
14035 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14037 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
14038 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
14039 Volker Reichelt's entry and add entries for people that perform
14040 GCC fuzzy testing and report numerous bugs.
14042 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
14045 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
14046 readonly data in sdata, if that is disabled.
14047 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
14048 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
14049 -mreadonly-in-sdata option.
14051 2018-03-07 Martin Sebor <msebor@redhat.com>
14053 PR tree-optimization/84468
14054 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
14055 basic block when looking for nul assignment.
14057 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
14060 * except.h (output_function_exception_table): Adjust prototype.
14061 * except.c (output_function_exception_table): Remove FNNAME parameter
14062 and add SECTION parameter. Ouput one part of the table at a time.
14063 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
14064 the first part of the exception table and emit unwind directives.
14065 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
14066 (i386_pe_seh_cold_init): Likewise.
14067 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
14068 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
14069 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
14070 (ix86_output_call_insn): Emit a nop in one more case for SEH.
14071 * config/i386/winnt.c: Include except.h.
14072 (struct seh_frame_state): Add reg_offset, after_prologue and
14073 in_cold_section fields.
14074 (i386_pe_seh_end_prologue): Set seh->after_prologue.
14075 (i386_pe_seh_cold_init): New function.
14076 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
14077 to seh->in_cold_section.
14078 (seh_emit_push): Record the offset of the push.
14079 (seh_emit_save): Record the offet of the save.
14080 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
14081 Test seh->after_prologue to disregard the epilogue.
14082 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
14083 (i386_pe_end_cold_function): New function.
14085 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14088 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
14089 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
14092 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
14094 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
14095 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
14097 PR middle-end/84723
14098 * multiple_target.c: Include tree-inline.h and intl.h.
14099 (expand_target_clones): Diagnose and fail if node->definition and
14100 !tree_versionable_function_p (node->decl).
14102 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
14104 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
14106 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
14107 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
14108 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
14110 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14113 * combine.c (try_combine): Use reg_or_subregno instead of handling
14114 just paradoxical SUBREGs and REGs.
14116 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
14118 * config/arc/arc.c (arc_finalize_pic): Remove function.
14119 (arc_must_save_register): We use single base PIC register, remove
14120 checks to save/restore the PIC register.
14121 (arc_expand_prologue): Likewise.
14122 * config/arc/arc-protos.h (arc_set_default_type_attributes):
14124 (arc_verify_short): Likewise.
14125 (arc_attr_type): Likewise.
14126 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
14127 (walk_stores): Likewise.
14128 (arc_address_cost): Make it static.
14129 (arc_verify_short): Likewise.
14130 (branch_dest): Likewise.
14131 (arc_attr_type): Likewise.
14132 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
14133 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
14134 (arc_final_prescan_insn): Remove inserting the nops due to
14135 hardware hazards. It is done in reorg step.
14136 (insn_length_variant_t): Remove.
14137 (insn_length_parameters_t): Likewise.
14138 (arc_insn_length_parameters): Likewise.
14139 (arc_get_insn_variants): Likewise.
14140 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
14142 2018-03-06 Jakub Jelinek <jakub@redhat.com>
14144 PR inline-asm/84683
14145 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
14148 PR tree-optimization/84687
14149 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
14151 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
14153 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14155 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
14156 Rename to ppc_speculation_barrier.
14157 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
14158 __builtin_ppc_speculation_barrier.
14160 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14163 * combine.c (combine_simplify_rtx): Don't try to simplify if
14164 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
14167 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
14169 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
14170 to 32 bytes when compiling for POWER9.
14172 2018-03-05 Jakub Jelinek <jakub@redhat.com>
14175 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
14176 regparm >= 3 with no arg reg available also for calls with
14177 flag_force_indirect_call. Pass decl to ix86_function_regparm.
14180 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
14182 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
14184 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
14187 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
14189 2018-03-05 Richard Biener <rguenther@suse.de>
14191 PR tree-optimization/84486
14192 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
14193 When inserting a __builtin_assume_aligned call set the LHS
14194 SSA name alignment info accordingly.
14196 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
14198 PR tree-optimization/84114
14199 * config/aarch64/aarch64.c (aarch64_reassociation_width)
14200 Avoid reassociation of FLOAT_MODE addition.
14202 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
14204 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
14205 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
14206 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
14207 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
14208 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
14209 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
14210 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
14212 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
14213 __builtin_ia32_wbinvd): New builtins.
14214 (SPECIAL_ARGS2): New.
14215 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
14216 (SPECIAL_ARGS2): New.
14217 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
14218 (ix86_valid_target_attribute_inner_p): Ditto.
14219 (ix86_init_mmx_sse_builtins): Add special_args2.
14220 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
14221 TARGET_WBNOINVD_P): New.
14222 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
14223 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
14224 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
14225 * config/i386/immintrin.h (_wbinvd): New intrinsic.
14226 * config/i386/pconfigintrin.h: New file.
14227 * config/i386/wbnoinvdintrin.h: Ditto.
14228 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
14230 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
14232 2018-03-05 Richard Biener <rguenther@suse.de>
14234 PR tree-optimization/84670
14235 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14237 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14238 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14239 (compute_antic_aux): Only assert the number of values in ANTIC_IN
14240 doesn't grow if all successors (recursively) were visited at least
14243 2018-03-05 Richard Biener <rguenther@suse.de>
14245 PR tree-optimization/84650
14246 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
14247 if executed in the loop pipeline.
14249 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
14251 * doc/configfiles.texi (Configuration Files): Move info about
14252 conditionalizing $target-protos.h to...
14253 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
14254 differs from $target-protos.h.
14256 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
14257 Chung-Ju Wu <jasonwucj@gmail.com>
14259 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
14260 * config/nds32/nds32-multiple.md (setmemsi): Define.
14261 * config/nds32/nds32-memory-manipulation.c
14262 (nds32_gen_dup_4_byte_to_word_value): New.
14263 (emit_setmem_word_loop): New.
14264 (emit_setmem_byte_loop): New.
14265 (nds32_expand_setmem_loop): New.
14266 (nds32_expand_setmem_loop_v3m): New.
14267 (nds32_expand_setmem_unroll): New.
14268 (nds32_expand_setmem): New.
14270 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14271 Chung-Ju Wu <jasonwucj@gmail.com>
14273 * config/nds32/nds32-memory-manipulation.c
14274 (nds32_emit_load_store): New.
14275 (nds32_emit_post_inc_load_store): New.
14276 (nds32_emit_mem_move): New.
14277 (nds32_emit_mem_move_block): New.
14278 (nds32_expand_movmemsi_loop_unknown_size): New.
14279 (nds32_expand_movmemsi_loop_known_size): New.
14280 (nds32_expand_movmemsi_loop): New.
14281 (nds32_expand_movmemsi_unroll): New.
14282 (nds32_expand_movmemqi): Rename ...
14283 (nds32_expand_movmemsi): ... to this.
14284 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
14285 (movmemsi): ... to this.
14286 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
14287 (nds32_expand_movmemsi): ... to this.
14289 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14290 Monk Chiang <sh.chiang04@gmail.com>
14291 Chung-Ju Wu <jasonwucj@gmail.com>
14293 * config/nds32/nds32-protos.h
14294 (nds32_expand_load_multiple): New arguments.
14295 (nds32_expand_store_multiple): Ditto.
14296 (nds32_valid_multiple_load_store): Rename ...
14297 (nds32_valid_multiple_load_store_p): ... to this.
14298 * config/nds32/nds32-memory-manipulation.c
14299 (nds32_expand_load_multiple): Refine implementation.
14300 (nds32_expand_store_multiple): Ditto.
14301 * config/nds32/nds32-multiple.md
14302 (load_multiple): Update nds32_expand_load_multiple interface.
14303 (store_multiple): Update nds32_expand_store_multiple interface.
14304 * config/nds32/nds32-predicates.c
14305 (nds32_valid_multiple_load_store): Rename ...
14306 (nds32_valid_multiple_load_store_p): ... to this and refine
14308 * config/nds32/predicates.md
14309 (nds32_load_multiple_and_update_address_operation): New predicate.
14310 (nds32_store_multiple_and_update_address_operation): New predicate.
14312 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
14313 Chung-Ju Wu <jasonwucj@gmail.com>
14315 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
14316 (combo): New attribute.
14317 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
14319 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
14321 * config/nds32/nds32.opt: Change -mcmodel= default value.
14323 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
14324 Monk Chiang <sh.chiang04@gmail.com>
14325 Chung-Ju Wu <jasonwucj@gmail.com>
14327 * config/nds32/constants.md (unspec_element): New enum.
14328 * config/nds32/constraints.md (Umw): New constraint.
14329 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
14330 * config/nds32/nds32-intrinsic.md: Likewise.
14331 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
14332 (nds32_valid_smw_lwm_base_p): New.
14333 (nds32_output_smw_single_word): New.
14334 (nds32_output_lmw_single_word): New.
14335 (nds32_expand_unaligned_load): New.
14336 (nds32_expand_unaligned_store): New.
14337 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
14338 (nds32_output_smw_single_word): Declare.
14339 (nds32_output_lmw_single_word): Declare.
14340 (nds32_expand_unaligned_load): Declare.
14341 (nds32_expand_unaligned_store): Declare.
14342 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
14343 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
14344 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
14345 NDS32_BUILTIN_UASTORE_DW.
14346 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
14349 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
14350 Kito Cheng <kito.cheng@gmail.com>
14351 Chung-Ju Wu <jasonwucj@gmail.com>
14353 * config/nds32/nds32-intrinsic.c
14354 (nds32_expand_builtin_null_ftype_reg): Delete.
14355 (nds32_expand_builtin_reg_ftype_imm): Ditto.
14356 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
14357 (nds32_read_argument): New.
14358 (nds32_legitimize_target): Ditto.
14359 (nds32_legitimize_argument): Ditto.
14360 (nds32_check_constant_argument): Ditto.
14361 (nds32_expand_unop_builtin): Ditto.
14362 (nds32_expand_unopimm_builtin): Ditto.
14363 (nds32_expand_binop_builtin): Ditto.
14364 (nds32_builtin_decl_impl): Ditto.
14365 (builtin_description): Ditto.
14366 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
14367 (nds32_init_builtins_impl): Ditto.
14368 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
14369 (nds32_builtin_decl): New.
14370 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
14371 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
14373 2018-03-02 Jeff Law <law@redhat.com>
14375 * reorg.c (stop_search_p): Handle DEBUG_INSN.
14376 (redundant_insn, fill_simple_delay_slots): Likewise.
14377 (fill_slots_from_thread): Likewise.
14378 * resource.c (mark_referenced_resources): Likewise.
14379 (mark_set_resources, find_dead_or_set_registers): Likewise.
14381 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14383 * substring-locations.h (format_warning_va): Formatting fix for
14384 ATTRIBUTE_GCC_DIAG.
14385 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
14387 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
14388 * substring-locations.c: Include intl.h.
14389 (format_warning_va): Turned into small wrapper around
14390 format_warning_n_va, renamed to ...
14391 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
14392 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
14394 (format_warning_at_substring_n): New function.
14395 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
14396 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
14397 format_warning_at_substring with just a shorter name instead of
14398 const function pointer.
14399 (fmtwarn_n): New function.
14400 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
14401 appropriate, get rid of all the fmtstr temporaries, move conditionals
14402 with G_() wrapped string literals directly into fmtwarn arguments,
14403 cast dir.len to (int), formatting fixes.
14405 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
14407 * doc/invoke.texi: Remove "Cilk Plus" references.
14409 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14410 Richard Biener <rguenther@suse.de>
14413 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
14414 for error or warning attributes if CALL_FROM_THUNK_P is set.
14417 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14420 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
14421 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
14424 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
14425 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
14427 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
14428 instead of -1U in last predictors element's probability member.
14430 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14433 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14434 arguments if they are comparable.
14436 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14438 PR tree-optimization/84634
14439 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
14440 masks and masked_loop_p with a single loop_masks, making sure it's
14441 null for bb vectorization.
14443 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14445 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14446 (vect_analyze_data_ref_access): Use loop->safe_len rather than
14447 loop->force_vectorize to check whether there is no alias.
14449 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14452 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
14454 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
14456 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
14457 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
14458 instead of a loop around prev_real_insn.
14459 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
14462 PR inline-asm/84625
14463 * config/i386/i386.c (ix86_print_operand): Use conditional
14464 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
14467 2018-03-02 Richard Biener <rguenther@suse.de>
14469 PR tree-optimization/84427
14470 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
14471 (bitmap_set_subtract_values): Rewrite to handle multiple
14474 (prune_clobbered_mems): Likewise.
14475 (phi_translate): Take edge instead of pred/phiblock.
14476 (phi_translate_1): Likewise.
14477 (phi_translate_set): Likewise. Insert all translated
14478 exprs for a value into the set, keeping possibly multiple
14479 expressions per value.
14480 (compute_antic_aux): Adjust for phi_translate changes.
14481 When intersecting union the expressions and prune those
14482 not in the final value set, keeping possibly multiple
14483 expressions per value. Do not use value-insertion
14484 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
14485 all expressions. Add verification that the value-sets
14486 only shrink during iteration.
14487 (compute_partial_antic_aux): Adjust for the phi_translate changes.
14488 (do_pre_regular_insertion): Likewise.
14489 (do_pre_partial_partial_insertion): Likewise.
14491 2018-03-02 Richard Biener <rguenther@suse.de>
14494 * config/darwin.c (saved_debug_info_level): New static global.
14495 (darwin_asm_lto_start): Disable debug info generation for LTO out.
14496 (darwin_asm_lto_end): Restore debug info generation settings.
14498 2018-03-01 Martin Liska <mliska@suse.cz>
14501 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
14502 volatile arguments.
14504 2018-03-01 Richard Biener <rguenther@suse.de>
14507 * dwarf2out.c (gen_variable_die): Properly handle late VLA
14508 type annotation with LTO when debug was disabled at compile-time.
14510 2018-03-01 Matthew Fortune <mfortune@gmail.com>
14512 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
14514 (mips_final_postscan_insn): Likewise.
14516 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
14518 PR rtl-optimization/84528
14519 * alias.c (init_alias_target): Add commentary.
14520 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
14521 a unique base value if the frame pointer is not eliminated
14522 to the stack pointer.
14524 2018-03-01 Tom de Vries <tom@codesourcery.com>
14526 PR rtl-optimization/83327
14527 * lra-int.h (hard_regs_spilled_into): Declare.
14528 * lra.c (hard_regs_spilled_into): Define.
14529 (init_reg_info): Init hard_regs_spilled_into.
14530 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
14531 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
14532 (process_bb_lives): Handle hard_regs_spilled_into.
14533 (lra_create_live_ranges_1): Before doing liveness propagation, clear
14534 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
14536 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
14538 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
14539 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
14540 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
14541 * config/rs6000/aix72.h: New file.
14543 2018-02-28 Jakub Jelinek <jakub@redhat.com>
14545 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
14546 instead of warning_at with conditional singular and plural messages
14550 * stor-layout.c (update_alignment_for_field): For
14551 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
14552 && !DECL_PACKED (field), do the alignment update, just use
14553 only desired_align instead of MAX (type_align, desired_align)
14555 (place_field): Don't do known_align < desired_align handling
14556 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
14557 is non-NULL, instead do it after rli->prev_field handling and
14558 only if not within a bitfield word. For DECL_PACKED (field)
14559 use type_align of BITS_PER_UNIT.
14561 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
14563 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
14564 superfluous parentheses and trailing spaces.
14566 2018-02-28 Richard Biener <rguenther@suse.de>
14568 PR tree-optimization/84584
14569 * graphite-scop-detection.c (scop_detection::add_scop): Discard
14570 SCoPs with fake exit edge.
14572 2018-02-28 Martin Liska <mliska@suse.cz>
14575 * timevar.c (timer::print): Fix format to properly print 100%
14578 2018-02-28 Richard Biener <rguenther@suse.de>
14580 PR middle-end/84607
14581 * genmatch.c (capture_info::walk_match): Do not mark
14582 captured expressions without operands as expr_p given
14583 they act more like predicates and should be subject to
14584 "lost tail" side-effect preserving.
14586 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
14588 PR rtl-optimization/81611
14589 * auto-inc-dec.c (attempt_change): Move dead note from
14590 mem_insn if it's the next use of regno
14591 (find_address): Take address use of reg holding
14592 non-incremented value. Add parm to limit search to the named
14594 (merge_in_block): Attempt to use a mem insn that is the next
14595 use of the original regno.
14597 2018-02-27 Martin Sebor <msebor@redhat.com>
14600 * doc/invoke.texi (-Wmissing-attributes): New option.
14601 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
14603 2018-02-27 Martin Sebor <msebor@redhat.com>
14605 PR translation/84207
14606 * diagnostic-core.h (warning_n, error_n, inform_n): Change
14607 n argument to unsigned HOST_WIDE_INT.
14608 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
14609 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
14610 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
14611 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
14613 2018-02-27 Richard Biener <rguenther@suse.de>
14615 PR tree-optimization/84512
14616 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
14617 Do not use the estimate returned from record_stmt_cost for
14618 the scalar iteration cost but sum properly using add_stmt_cost.
14620 2018-02-27 Richard Biener <rguenther@suse.de>
14622 PR tree-optimization/84466
14623 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14624 Adjust last change to less strictly validate use operands.
14626 2018-02-27 Martin Liska <mliska@suse.cz>
14628 PR gcov-profile/84548
14629 * gcov.c (process_file): Allow partial overlap and consider it
14630 also as group functions.
14631 (output_lines): Properly calculate range of lines for a group.
14633 2018-02-27 Martin Liska <mliska@suse.cz>
14635 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
14636 'ggc' suffixes. Change first column width.
14637 (timer::print): Fix formatting of the column.
14639 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
14641 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
14642 preserve inline entry blocks for the sake of debug inline
14643 entry point markers alone.
14644 (remove_unused_locals): Suggest in comments a better place to
14645 force the preservation of inline entry blocks that are
14646 otherwise unused, but do not preserve them.
14648 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14650 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
14652 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14655 * config/i386/constraints.md (Bs): Replace
14656 ix86_indirect_branch_register with
14657 TARGET_INDIRECT_BRANCH_REGISTER.
14659 * config/i386/i386.md (indirect_jump): Likewise.
14660 (tablejump): Likewise.
14661 (*sibcall_memory): Likewise.
14662 (*sibcall_value_memory): Likewise.
14663 Peepholes of indirect call and jump via memory: Likewise.
14664 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
14665 (*sibcall_value_GOT_32): Likewise.
14666 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14667 (GOT_memory_operand): Likewise.
14668 (call_insn_operand): Likewise.
14669 (sibcall_insn_operand): Likewise.
14670 (GOT32_symbol_operand): Likewise.
14671 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
14673 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14675 PR rtl-optimization/83496
14676 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
14677 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
14678 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
14679 redundant insn, if any.
14680 (relax_delay_slots): Likewise.
14681 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
14683 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
14685 PR tree-optimization/83965
14686 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
14687 that grouped statements are part of a reduction chain. Return
14688 true if the statement is not marked as a reduction itself but
14689 is part of a group.
14690 (vect_recog_dot_prod_pattern): Don't check whether the statement
14691 is part of a group here.
14692 (vect_recog_sad_pattern): Likewise.
14693 (vect_recog_widen_sum_pattern): Likewise.
14695 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14698 * final.c (rest_of_clean_state): Also look for calls inside sequences.
14700 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14703 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
14705 (ix86_output_indirect_function_return): New prototype.
14706 (ix86_split_simple_return_pop_internal): Likewise.
14707 * config/i386/i386.c (indirect_return_via_cx): New.
14708 (indirect_return_via_cx_bnd): Likewise.
14709 (indirect_thunk_name): Handle return va CX_REG.
14710 (output_indirect_thunk_function): Create alias for
14711 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
14712 (ix86_output_indirect_jmp): Remove the bool argument.
14713 (ix86_output_indirect_function_return): New function.
14714 (ix86_split_simple_return_pop_internal): Likewise.
14715 * config/i386/i386.md (*indirect_jump): Don't pass false
14716 to ix86_output_indirect_jmp.
14717 (*tablejump_1): Likewise.
14718 (simple_return_pop_internal): Change it to define_insn_and_split.
14719 Call ix86_split_simple_return_pop_internal to split it for
14720 -mfunction-return=.
14721 (simple_return_indirect_internal): Call
14722 ix86_output_indirect_function_return instead of
14723 ix86_output_indirect_jmp.
14725 2018-02-26 Jakub Jelinek <jakub@redhat.com>
14728 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
14729 memset and value initialization afterwards.
14731 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
14733 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
14735 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14738 * common/config/aarch64/aarch64-common.c
14739 (aarch_option_optimization_table[]): Switch
14740 off fomit-frame-pointer
14742 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14743 Chung-Ju Wu <jasonwucj@gmail.com>
14745 * config/nds32/nds32-multiple.md (load_multiple): Disallow
14747 (store_multiple): Ditto.
14749 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14751 * config.gcc: Add --with-cpu support for nds32 target.
14752 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
14753 * config/nds32/nds32.opt: Add -mcpu= option.
14755 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
14757 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
14758 isel=yes): Warn for these deprecated options.
14760 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
14762 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
14763 ISA_2_5_MASKS_EMBEDDED.
14765 2018-02-23 Jakub Jelinek <jakub@redhat.com>
14767 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
14768 p->max as pointers rather than using iterative_hash_expr.
14770 2018-02-23 Carl Love <cel@us.ibm.com>
14772 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
14773 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
14775 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
14776 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
14777 P8V_BUILTIN_VEC_VUNSIGNED2.
14779 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
14782 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
14783 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
14785 * lra-constraints.c (curr_insn_transform): Set up
14786 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
14788 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
14789 * lra-eliminations.c (spill_pseudos): Ditto.
14790 (process_insn_for_elimination): Ditto.
14791 * lra-lives.c (reg_early_clobber_p): Use the new macros.
14792 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
14793 LRA_NON_CLOBBERED_ALT.
14795 2018-02-22 Martin Sebor <msebor@redhat.com>
14797 PR tree-optimization/84480
14798 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
14799 to maybe_diag_stxncpy_trunc. Call it.
14800 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
14801 from gimple_fold_builtin_strcpy. Print inlining stack.
14802 (handle_builtin_stxncpy): Print inlining stack.
14803 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
14805 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
14808 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
14809 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
14810 and -fcheck-pointer-bounds are used together.
14811 (indirect_thunk_prefix): New enum.
14812 (indirect_thunk_need_prefix): New function.
14813 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
14814 "_nt" instead of "_bnd" for NOTRACK prefix.
14815 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
14816 (output_indirect_thunk_function): Likewise.
14818 (ix86_code_end): Update output_indirect_thunk_function calls.
14819 (ix86_output_indirect_branch_via_reg): Replace
14820 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
14821 (ix86_output_indirect_branch_via_push): Likewise.
14822 (ix86_output_function_return): Likewise.
14823 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
14824 incompatible with -fcf-protection=branch and
14825 -fcheck-pointer-bounds.
14827 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14830 * config/aarch64/aarch64.c (aarch64_print_address_internal):
14831 Change gcc_assert call to output_operand_lossage.
14833 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14835 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
14837 2018-02-22 DJ Delorie <dj@redhat.com>
14838 Sebastian Perta <sebastian.perta@renesas.com>
14839 Oleg Endo <olegendo@gcc.gnu.org>
14841 * config/rx/rx.c (rx_rtx_costs): New function.
14842 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
14844 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
14846 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
14848 2018-02-22 Martin Liska <mliska@suse.cz>
14851 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
14852 Add "native" as a possible value.
14854 2018-02-22 Martin Liska <mliska@suse.cz>
14857 * config/i386/i386.c (ix86_option_override_internal):
14858 Add "native" as a possible value for -march and -mtune.
14860 2018-02-22 Jakub Jelinek <jakub@redhat.com>
14863 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
14864 to all type variants.
14866 PR tree-optimization/84503
14867 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
14868 width as info->bitpos + info->bitsize - start.
14869 (merged_store_group::merge_overlapping): Simplify width computation.
14870 (check_no_overlap): New function.
14871 (imm_store_chain_info::try_coalesce_bswap): Compute expected
14872 start + width and last_order of the group, fail if check_no_overlap
14874 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
14875 to group if check_no_overlap fails.
14877 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14879 * config/rs6000/altivec.md: Delete contraint arguments to
14880 define_expand, define_split, and define_peephole2, and in
14881 define_insn_and_split if always unused.
14882 * config/rs6000/darwin.md: Ditto.
14883 * config/rs6000/dfp.md: Ditto.
14884 * config/rs6000/rs6000.md: Ditto.
14885 * config/rs6000/sync.md: Ditto.
14886 * config/rs6000/vector.md: Ditto.
14887 * config/rs6000/vsx.md: Ditto.
14889 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14891 * config/rs6000/altivec.md: Write output control strings as braced
14892 blocks instead of double-quoted strings.
14893 * config/rs6000/darwin.md: Ditto.
14894 * config/rs6000/rs6000.md: Ditto.
14895 * config/rs6000/vector.md: Ditto.
14896 * config/rs6000/vsx.md: Ditto.
14898 2018-02-21 Jason Merrill <jason@redhat.com>
14900 PR c++/84314 - ICE with templates and fastcall attribute.
14901 * attribs.c (build_type_attribute_qual_variant): Remove assert.
14903 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14905 * ipa-cp.c (determine_versionability): Fix comment typos.
14907 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14910 * ipa-cp.c (determine_versionability): Do not version functions caling
14913 2018-02-21 Martin Liska <mliska@suse.cz>
14916 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
14917 Add "native" as a possible value.
14918 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
14919 the macro when native cpu detection is available.
14921 2018-02-21 Martin Liska <mliska@suse.cz>
14924 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
14925 Add "native" as a possible value.
14926 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
14927 when native cpu detection is available.
14929 2018-02-21 Jakub Jelinek <jakub@redhat.com>
14930 Martin Sebor <msebor@redhat.com>
14932 PR tree-optimization/84478
14933 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
14935 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
14936 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
14937 support which is conservatively correct, for 2 only stay conservative
14938 for maxlen. Formatting and comment capitalization fixes. Add STRICT
14939 argument to the 2 argument get_range_strlen, adjust 6 arg
14940 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
14942 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
14943 (gimple_fold_builtin_strlen): Pass true as last argument to
14946 2018-02-20 Martin Sebor <msebor@redhat.com>
14948 PR middle-end/84095
14949 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
14950 (builtin_memref::set_base_and_offset): Same. Handle inner references.
14951 (builtin_memref::builtin_memref): Factor out parts into
14952 set_base_and_offset and call it.
14954 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14956 PR middle-end/84406
14957 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
14958 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
14959 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
14960 search at the associated MODE_INT.
14962 2018-02-20 Jeff Law <law@redhat.com>
14964 PR middle-end/82123
14965 PR tree-optimization/81592
14966 PR middle-end/79257
14967 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
14968 for range data rather than using global data.
14969 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
14970 range data rather than using global data.
14971 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
14972 pass it to children as needed.
14973 (struct directive::fmtresult): Similarly.
14974 (struct directive::set_width): Similarly.
14975 (struct directive::set_precision): Similarly.
14976 (format_integer, format_directive, parse_directive): Similarly.
14977 (format_none): Accept unnamed vr_values parameter.
14978 (format_percent, format_floating, format_character): Similarly.
14979 (format_string, format_plain): Similarly.
14980 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
14981 the EVRP range analyzer for range data rather than using global data.
14982 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
14983 gimple-ssa-evrp-analyze.h
14984 (class sprintf_dom_walker): Add after_dom_children member function.
14985 Add evrp_range_analyzer member.
14986 (sprintf_dom_walker::before_dom_children): Call into the EVRP
14987 range analyzer as needed.
14988 (sprintf_dom_walker::after_dom_children): New member function.
14989 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
14991 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
14992 (evrp_range_analyzer::pop_to_marker): Likewise.
14994 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14996 PR tree-optimization/84419
14997 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
14998 with the required type if its current type is compatible but
15001 2018-02-20 Jakub Jelinek <jakub@redhat.com>
15003 PR middle-end/82004
15004 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
15005 after vectorization.
15007 2018-02-20 Martin Liska <mliska@suse.cz>
15010 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
15011 possible values if we don't have a hint.
15013 2018-02-20 Martin Liska <mliska@suse.cz>
15017 * final.c (shorten_branches): Build align_tab array with one
15019 * opts.c (finish_options): Add alignment option limit check.
15020 (MAX_CODE_ALIGN): Likewise.
15021 (MAX_CODE_ALIGN_VALUE): Likewise.
15022 * doc/invoke.texi: Document maximum allowed option value for
15023 all -falign-* options.
15025 2018-02-19 Jakub Jelinek <jakub@redhat.com>
15028 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
15029 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
15030 * var-tracking.c (emit_note_insn_var_location): Remove all references
15031 to NOTE_INSN_CALL_ARG_LOCATION.
15032 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
15033 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
15034 Use copy_rtx_if_shared.
15035 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
15036 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
15037 (dwarf2out_var_location): Remove handling of
15038 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
15040 * final.c (final_scan_insn): Remove all references to
15041 NOTE_INSN_CALL_ARG_LOCATION.
15042 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
15043 before dumping final insns.
15044 * except.c (emit_note_eh_region_end): Remove all references to
15045 NOTE_INSN_CALL_ARG_LOCATION.
15046 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
15047 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
15048 * config/arc/arc.c (hwloop_optimize): Likewise.
15049 * config/arm/arm.c (create_fix_barrier): Likewise.
15050 * config/s390/s390.c (s390_chunkify_start): Likewise.
15051 * config/sh/sh.c (find_barrier): Likewise.
15052 * config/i386/i386.c (rest_of_insert_endbranch,
15053 ix86_seh_fixup_eh_fallthru): Likewise.
15054 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
15055 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
15056 * config/frv/frv.c (frv_function_prologue): Likewise.
15057 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
15059 (note_outside_basic_block_p): Remove all references to
15060 NOTE_INSN_CALL_ARG_LOCATION.
15061 * gengtype.c (adjust_field_rtx_def): Likewise.
15062 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
15064 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
15065 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
15068 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
15071 PR tree-optimization/84452
15072 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
15073 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
15076 2018-02-19 Martin Liska <mliska@suse.cz>
15079 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
15081 2018-02-19 Martin Liska <mliska@suse.cz>
15082 Richard Sandiford <richard.sandiford@linaro.org>
15084 PR tree-optimization/82491
15085 * gimple-fold.c (get_base_constructor): Make earlier bail out
15088 2018-02-19 Carl Love <cel@us.ibm.com>
15090 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
15091 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
15093 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
15094 P8V_BUILTIN_VEC_NEG.
15096 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
15098 * config/rl78/rl78.md (movdf): New define expand.
15100 2018-02-19 Martin Liska <mliska@suse.cz>
15103 * doc/invoke.texi: Fix typo.
15104 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
15106 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
15108 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
15109 handle rs6000_single_float and rs6000_double_float specially for
15112 2018-02-16 Jeff Law <law@redhat.com>
15114 * config/rx/rx.c (add_pop_cfi_notes): New function.;
15115 (pop_regs): Use it.
15117 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15120 * ipa-inline.c (inline_small_functions): Fix a typo.
15122 2018-02-16 Nathan Sidwell <nathan@acm.org>
15124 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
15126 2018-02-16 Carl Love <cel@us.ibm.com>
15128 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
15129 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
15130 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
15131 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
15132 expansion to P8V_BUILTIN_VEC_FLOAT2.
15134 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
15136 PR rtl-optimization/70023
15137 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
15138 src_regno into account.
15140 2018-02-16 Carl Love <cel@us.ibm.com>
15142 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
15143 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
15144 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
15145 * config/rs6000/rs6000.c: Remove case statements for
15146 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
15147 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
15148 and P9V_BUILTIN_VEC_VINSERT4B.
15149 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
15150 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
15151 * config/rs6000/vsx.md:
15152 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
15155 2018-02-16 Carl Love <cel@us.ibm.com>
15157 * config/rs6000/altivec.h: Add builtin names vec_extract4b
15159 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
15161 * config/rs6000/rs6000-c.c: Add the definitions for
15162 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
15163 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
15164 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
15165 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
15166 definition for insert4b and define insn *insert3b_internal.
15167 * doc/extend.texi: Add documentation for vec_extract4b.
15169 2018-02-16 Nathan Sidwell <nathan@acm.org>
15171 * doc/extend.texi (Backwards Compatibility): Mention friend
15172 injection. Note for-scope is deprecated.
15173 * doc/invoke.texi (-ffriend-injection): Deprecate.
15175 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
15177 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
15178 that moved to I2, also allow destinations that are a paradoxical
15179 subreg (instead of a normal reg).
15181 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
15184 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
15187 2018-02-16 Richard Biener <rguenther@suse.de>
15189 PR tree-optimization/84037
15190 PR tree-optimization/84016
15192 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15193 Adjust vec_construct for the fact we need additional higher latency
15194 128bit inserts for AVX256 and AVX512 vector builds.
15195 (ix86_add_stmt_cost): Scale vector construction cost for
15198 2018-02-16 Richard Biener <rguenther@suse.de>
15200 PR tree-optimization/84417
15201 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
15202 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
15203 (non_rewritable_lvalue_p): Likewise, use poly-ints.
15205 2018-02-16 Martin Liska <mliska@suse.cz>
15208 * internal-fn.def (ASAN_CHECK): Set proper flags.
15209 (ASAN_MARK): Likewise.
15211 2018-02-16 Julia Koval <julia.koval@intel.com>
15213 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
15214 from PTA_CANNONLAKE.
15216 2018-02-16 Jakub Jelinek <jakub@redhat.com>
15219 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
15220 Use ++iter rather than iter++ for std::list iterators.
15221 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
15222 defer deleting them until all nodes in the forest are processed. Do
15223 free even leaf nodes. Change to_process into auto_vec.
15226 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
15227 * vec.h (vec_default_construct): Use memset instead of placement new
15228 if BROKEN_VALUE_INITIALIZATION is defined.
15229 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
15230 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
15233 PR rtl-optimization/83723
15234 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
15235 * lra.c (lra_substitute_pseudo): Likewise. If true, use
15236 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
15238 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
15240 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
15242 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
15244 PR rtl-optimization/81443
15245 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
15246 from inner REGs to paradoxical SUBREGs.
15248 2018-02-16 Richard Biener <rguenther@suse.de>
15250 PR tree-optimization/84399
15251 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15252 For operands we can analyze at their definition make sure we can
15253 analyze them at each use as well.
15255 2018-02-16 Richard Biener <rguenther@suse.de>
15257 PR tree-optimization/84190
15258 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
15259 volatile accesses if the decl isn't volatile.
15261 2018-02-15 Jason Merrill <jason@redhat.com>
15263 PR c++/84314 - ICE with templates and fastcall attribute.
15264 * attribs.c (build_type_attribute_qual_variant): Don't clobber
15265 TYPE_CANONICAL on an existing type.
15267 2018-02-15 Jakub Jelinek <jakub@redhat.com>
15269 PR tree-optimization/84383
15270 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
15271 dstoff nor call operand_equal_p if dstbase is NULL.
15273 PR tree-optimization/84334
15274 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
15275 also a CONSTANT_CLASS_P, punt.
15277 2018-02-14 Jim Wilson <jimw@sifive.com>
15279 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
15280 first SMALL_OPERAND check. New local min_second_step. Move assert
15281 to where locals are set. Add TARGET_RVC support.
15282 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
15284 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
15286 * doc/invoke.texi: Correct -Wformat-overflow code sample.
15288 2018-02-14 Martin Sebor <msebor@redhat.com>
15290 PR tree-optimization/83698
15291 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
15292 arrays constrain the offset range to their bounds.
15293 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
15294 (builtin_access::overlap): Avoid setting the size of overlap if it's
15296 (maybe_diag_overlap): Also consider arrays when deciding what values
15297 of offsets to include in diagnostics.
15299 2018-02-14 Martin Sebor <msebor@redhat.com>
15302 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
15303 that correspond to the kind of a declaration.
15305 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
15308 * config/pa/pa.md: Load address of PIC label using the linkage table
15309 if the label is nonlocal.
15311 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
15313 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
15314 warning message if user requests -maltivec=be.
15315 * doc/invoke.texi: Document deprecation of -maltivec=be.
15317 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
15320 * config/rs6000/rs6000-c.c: Update definitions for
15321 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
15322 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
15324 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15327 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
15328 add _get_ssp intrinsics. Remove argument from
15329 __builtin_ia32_rdssp[d|q].
15330 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
15331 * config/i386/i386-builtin.def: Remove argument from
15332 __builtin_ia32_rdssp[d|q].
15333 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
15334 ix86_expand_special_args_builtin for _rdssp[d|q].
15335 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
15336 Clear register before usage.
15337 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
15338 Add documentation for new _get_ssp and _inc_ssp intrinsics.
15340 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
15342 PR tree-optimization/84357
15343 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
15344 operand 1 of an ARRAY_REF too.
15346 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
15349 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
15350 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
15352 (set_of_reg): New struct.
15353 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
15354 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
15355 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
15357 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
15358 Split into bitclr, bitset, bitinvert patterns if appropriate.
15359 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
15360 use rx_fuse_in_memory_bitop.
15361 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
15362 to named insn, correct maximum insn length.
15364 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
15367 * machmode.def: Define a complex mode for PARTIAL_INT.
15368 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
15370 * doc/rtl.texi: Document CSPImode.
15371 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
15373 (msp430_hard_regno_nregs_with_padding): Likewise.
15375 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
15378 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
15380 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
15382 PR rtl-optimization/84169
15383 * combine.c (try_combine): New variable split_i2i3. Set it to true if
15384 we generated a parallel as new i3 and we split that to new i2 and i3
15385 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
15386 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
15387 those to i2, not i1. Partially rewrite this scan code.
15389 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15392 * stor-layout.c (place_field): For variable length fields, adjust
15393 offset_align afterwards not just based on the field's alignment,
15394 but also on the size.
15396 PR middle-end/84309
15397 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
15398 of exps and logs in the use_exp2 case.
15400 2018-02-13 Jeff Law <law@redhat.com>
15402 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
15403 entry for "vector".
15405 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
15408 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
15412 * common.opt (gas-loc-support, gas-locview-support): New.
15413 (ginline-points, ginternal-reset-location-views): New.
15414 * doc/invoke.texi: Document them. Use @itemx where intended.
15415 (gvariable-location-views): Adjust.
15416 * target.def (reset_location_view): New.
15417 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
15418 (TARGET_RESET_LOCATION_VIEW): New.
15419 * doc/tm.texi: Rebuilt.
15420 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
15421 (dwarf2out_default_as_locview_support): New.
15422 (output_asm_line_debug_info): Use option variables.
15423 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
15424 (output_loc_list): Likewise.
15425 (add_high_low_attributes): Check option variables.
15426 Don't output entry view attribute in strict mode.
15427 (gen_inlined_subroutine_die): Check option variables.
15428 (dwarf2out_inline_entry): Likewise.
15429 (init_sections_and_labels): Likewise.
15430 (dwarf2out_early_finish): Likewise.
15431 (maybe_reset_location_view): New, from...
15432 (dwarf2out_var_location): ... here. Call it.
15433 * debug.h (dwarf2out_default_as_loc_support): Declare.
15434 (dwarf2out_default_as_locview_support): Declare.
15435 * hooks.c (hook_int_rtx_insn_0): New.
15436 * hooks.h (hook_int_rtx_insn_0): Declare.
15437 * toplev.c (process_options): Take -gas-loc-support and
15438 -gas-locview-support from dwarf2out. Enable
15439 -gvariable-location-views by default only with locview
15440 assembler support. Enable -ginternal-reset-location-views by
15441 default only if the target defines the corresponding hook.
15442 Enable -ginline-points by default if location views are
15443 enabled; force it disabled if statement frontiers are
15445 * tree-inline.c (expand_call_inline): Check option variables.
15446 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
15448 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
15450 PR tree-optimization/84321
15451 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
15452 handling. Also check whether the anti-range contains any values
15453 that satisfy the mask; switch to a VR_RANGE if not.
15455 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
15458 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
15460 2018-02-13 Martin Jambor <mjambor@suse.cz>
15463 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
15464 of call statements, also set location of a load to a temporary.
15466 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
15468 * config/rl78/rl78.c (add_vector_labels): New function.
15469 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
15470 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
15471 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
15472 which checks that no arguments are passed.
15473 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
15474 * doc/extend.texi: Documentation for the new attribute.
15476 2018-02-13 Andreas Schwab <schwab@suse.de>
15478 * config/riscv/linux.h (CPP_SPEC): Define.
15480 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15483 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
15484 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
15485 OPTION_MASK_ISA_AES as first argument to def_builtin_const
15486 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
15487 instead of OPTION_MASK_ISA_PCLMUL as first argument to
15488 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
15489 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
15490 temporarily for AES and PCLMUL builtins.
15492 PR tree-optimization/84339
15493 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
15494 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
15497 PR middle-end/84309
15498 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
15499 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
15500 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
15502 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
15504 * omp-simd-clone.h: New file.
15505 * omp-simd-clone.c: Include omp-simd-clone.h.
15506 (expand_simd_clones): No longer static.
15507 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
15508 cgraph.h and omp-simd-clone.h.
15509 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
15510 (vect_recog_widen_shift_pattern): Formatting fix.
15511 (vect_pattern_recog_1): Don't check optab for calls.
15514 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
15515 operands[2] into a REG before using gen_lowpart on it.
15517 2018-02-12 Jeff Law <law@redhat.com>
15520 * config/sh/sh.c (find_barrier): Consider a sibling call
15523 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
15524 successfully back substituting a reg.
15526 2018-02-12 Richard Biener <rguenther@suse.de>
15528 PR tree-optimization/84037
15529 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
15530 parameter, move visited init to caller.
15531 (vect_slp_analyze_operations): Separate cost from validity
15532 check, initialize visited once for all instances.
15533 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
15535 * tree-vect-stmts.c (vect_model_simple_cost): Make early
15537 (vect_model_promotion_demotion_cost): Likewise.
15538 (vectorizable_bswap): Guard cost modeling with !slp_node
15539 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
15541 (vectorizable_call): Likewise.
15542 (vectorizable_conversion): Likewise.
15543 (vectorizable_assignment): Likewise.
15544 (vectorizable_shift): Likewise.
15545 (vectorizable_operation): Likewise.
15546 (vectorizable_store): Likewise.
15547 (vectorizable_load): Likewise.
15548 (vectorizable_condition): Likewise.
15549 (vectorizable_comparison): Likewise.
15551 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
15554 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
15555 (ASAN_MARK): Fix fnspec to account for return value, change pointer
15556 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
15558 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15560 PR middle-end/83665
15561 * params.def (inline-min-speedup): Increase from 8 to 15.
15562 (max-inline-insns-auto): Decrease from 40 to 30.
15563 * ipa-split.c (consider_split): Add some buffer for function to
15564 be considered inlining candidate.
15565 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
15568 2018-02-12 Richard Biener <rguenther@suse.de>
15570 PR tree-optimization/84037
15571 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
15572 matched stmts if we cannot swap the non-matched ones.
15574 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15576 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
15577 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
15578 _mm_maskz_scalef_round_ss): New intrinsics.
15579 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
15580 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
15581 __builtin_ia32_scalefss_round): Remove.
15582 (__builtin_ia32_scalefsd_mask_round,
15583 __builtin_ia32_scalefss_mask_round): New intrinsics.
15584 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
15585 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
15586 ((match_operand:VF_128 2 "<round_nimm_predicate>"
15587 "<round_constraint>")): Changed to ...
15588 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
15589 "<round_scalar_constraint>")): ... this.
15590 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
15591 %0, %1, %2<round_op3>}"): Changed to ...
15592 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
15593 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
15594 %2<round_scalar_mask_op3>}"): ... this.
15595 * config/i386/subst.md (round_scalar_nimm_predicate): New.
15597 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15599 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
15600 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
15601 (_mm_maskz_sqrt_round_ss): New intrinsics.
15602 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
15603 (__builtin_ia32_sqrtsd_mask_round)
15604 (__builtin_ia32_sqrtss_mask_round): New builtins.
15605 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
15606 (__builtin_ia32_sqrtss_round): Remove.
15607 (__builtin_ia32_sqrtsd_mask_round)
15608 (__builtin_ia32_sqrtss_mask_round): New builtins.
15609 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
15610 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
15611 ((match_operand:VF_128 1 "vector_operand"
15612 "xBm,<round_constraint>")): Changed to ...
15613 ((match_operand:VF_128 1 "vector_operand"
15614 "xBm,<round_scalar_constraint>")): ... this.
15615 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
15616 %0, %2, %<iptr>1<round_op3>}): Changed to ...
15617 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
15618 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
15619 %<iptr>1<round_scalar_mask_op3>}): ... this.
15620 ((set_attr "prefix" "<round_prefix>")): Changed to ...
15621 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
15623 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
15626 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
15627 Cast vec_cmpeq result to correct type.
15628 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
15629 Cast vec_cmpgt result to correct type.
15631 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
15633 * final.c (final_scan_insn_1): Renamed from...
15634 (final_scan_insn): ... this. New wrapper, to recover
15635 seen from the outermost call in recursive ones.
15636 * config/sparc/sparc.c (output_return): Drop seen from call.
15637 (output_sibcall): Likewise.
15638 * config/visium/visium.c (output_branch): Likewise.
15640 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
15642 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
15645 2018-02-10 Alan Modra <amodra@gmail.com>
15648 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
15649 Specify LR as an input.
15651 2018-02-10 Jakub Jelinek <jakub@redhat.com>
15654 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
15655 remove_member_access_dummy_vars): New functions.
15656 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
15657 lower_omp_1, execute_lower_omp): Use them.
15659 PR rtl-optimization/84308
15660 * shrink-wrap.c (spread_components): Release todo vector.
15662 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
15664 PR rtl-optimization/57193
15665 * ira-color.c (struct allocno_color_data): Add member
15666 conflict_allocno_hard_prefs.
15667 (update_conflict_allocno_hard_prefs): New.
15668 (bucket_allocno_compare_func): Add a preference based on
15669 conflict_allocno_hard_prefs.
15670 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
15671 (color_allocnos): Remove a dead code. Initiate
15672 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
15674 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15677 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
15678 constraint from =wa to wa. Avoid a subreg on the output operand,
15679 instead use a pseudo and subreg it in a move.
15680 (p9_xxbrd_<mode>): Changed to ...
15681 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
15682 (p9_xxbrd_v2df): New expander.
15683 (p9_xxbrw_<mode>): Changed to ...
15684 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
15685 (p9_xxbrw_v4sf): New expander.
15687 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15689 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
15691 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
15694 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
15695 multiply in 32-bit mode.
15696 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
15697 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
15700 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15702 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
15703 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
15704 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
15705 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
15707 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
15710 * dwarf2out.c (is_trivial_indirect_ref): New function.
15711 (dwarf2out_late_global_decl): Do not generate a location
15712 attribute for variables that have a non-trivial DECL_VALUE_EXPR
15713 and that are not defined in the current unit.
15715 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15717 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
15718 instead of a libcall for UNORDERED.
15720 2018-02-09 Tamar Christina <tamar.christina@arm.com>
15723 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
15724 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
15726 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15729 * config/s390/s390.c (s390_set_current_function): Invoke
15730 s390_indirect_branch_settings also if fndecl didn't change.
15732 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15734 * config/rs6000/rs6000.md (blockage): Set length to zero.
15736 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15738 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
15740 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15743 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
15744 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
15748 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
15749 PARALLEL incoming that failed vt_get_decl_and_offset check.
15751 PR middle-end/84237
15752 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
15753 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
15755 (get_variable_section): For decls in named .bss* sections pass true as
15756 second argument to bss_initializer_p.
15758 2018-02-09 Marek Polacek <polacek@redhat.com>
15759 Jakub Jelinek <jakub@redhat.com>
15762 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
15763 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
15764 Sync some changes from cxx_fold_indirect_ref.
15766 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15768 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
15770 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
15771 (BLOCK_INLINE_ENTRY_LABEL): New.
15772 (dwarf2out_var_location): Disregard inline entry markers.
15773 (inline_entry_data): New struct.
15774 (inline_entry_data_hasher): New hashtable type.
15775 (inline_entry_data_hasher::hash): New.
15776 (inline_entry_data_hasher::equal): New.
15777 (inline_entry_data_table): New variable.
15778 (add_high_low_attributes): Add DW_AT_entry_pc and
15779 DW_AT_GNU_entry_view attributes if a pending entry is found
15780 in inline_entry_data_table. Add old entry_pc attribute only
15781 if debug nonbinding markers are disabled.
15782 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
15783 markers are enabled.
15784 (block_within_block_p, dwarf2out_inline_entry): New.
15785 (dwarf2out_finish): Check that no entries remained in
15786 inline_entry_data_table.
15787 * final.c (reemit_insn_block_notes): Handle inline entry notes.
15788 (final_scan_insn, notice_source_line): Likewise.
15789 (rest_of_clean_state): Skip inline entry markers.
15790 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
15792 * gimple.c (gimple_build_debug_inline_entry): New.
15793 * gimple.h (enum gimple_debug_subcode): Add
15794 GIMPLE_DEBUG_INLINE_ENTRY.
15795 (gimple_build_debug_inline_entry): Declare.
15796 (gimple_debug_inline_entry_p): New.
15797 (gimple_debug_nonbind_marker_p): Adjust.
15798 * insn-notes.def (INLINE_ENTRY): New.
15799 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
15800 inline entry marker notes.
15801 (print_insn): Likewise.
15802 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
15803 (INSN_DEBUG_MARKER_KIND): Likewise.
15804 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
15805 * tree-inline.c (expand_call_inline): Build and insert
15806 debug_inline_entry stmt.
15807 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
15808 inline entry blocks early, if nonbind markers are enabled.
15809 (dump_scope_block): Dump fragment info.
15810 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
15811 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
15812 (gimple_build_debug_inline_entry): New.
15813 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
15814 Enable/disable inline entry points too.
15815 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
15816 (DEBUG_INSN): Describe inline entry markers.
15818 * common.opt (gvariable-location-views): New.
15819 (gvariable-location-views=incompat5): New.
15820 * config.in: Rebuilt.
15821 * configure: Rebuilt.
15822 * configure.ac: Test assembler for view support.
15823 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
15824 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
15825 * dwarf2out.c (var_loc_view): New typedef.
15826 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
15827 (dwarf2out_locviews_in_attribute): New.
15828 (dwarf2out_locviews_in_loclist): New.
15829 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
15830 (enum dw_line_info_opcode): Add LI_adv_address.
15831 (struct dw_line_info_table): Add view.
15832 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
15833 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
15834 (zero_view_p): New variable.
15835 (ZERO_VIEW_P): New macro.
15836 (output_asm_line_debug_info): New.
15837 (struct var_loc_node): Add view.
15838 (add_AT_view_list, AT_loc_list): New.
15839 (add_var_loc_to_decl): Add view param. Test it against last.
15840 (new_loc_list): Add view params. Record them.
15841 (AT_loc_list_ptr): Handle loc and view lists.
15842 (view_list_to_loc_list_val_node): New.
15843 (print_dw_val): Handle dw_val_class_view_list.
15844 (size_of_die): Likewise.
15845 (value_format): Likewise.
15846 (loc_list_has_views): New.
15847 (gen_llsym): Set vl_symbol too.
15848 (maybe_gen_llsym, skip_loc_list_entry): New.
15849 (dwarf2out_maybe_output_loclist_view_pair): New.
15850 (output_loc_list): Output view list or entries too.
15851 (output_view_list_offset): New.
15852 (output_die): Handle dw_val_class_view_list.
15853 (output_dwarf_version): New.
15854 (output_compilation_unit_header): Use it.
15855 (output_skeleton_debug_sections): Likewise.
15856 (output_rnglists, output_line_info): Likewise.
15857 (output_pubnames, output_aranges): Update version comments.
15858 (output_one_line_info_table): Output view numbers in asm comments.
15859 (dw_loc_list): Determine current endview, pass it to new_loc_list.
15860 Call maybe_gen_llsym.
15861 (loc_list_from_tree_1): Adjust.
15862 (add_AT_location_description): Create view list attribute if
15863 needed, check it's absent otherwise.
15864 (convert_cfa_to_fb_loc_list): Adjust.
15865 (maybe_emit_file): Call output_asm_line_debug_info for test.
15866 (dwarf2out_var_location): Reset views as needed. Precompute
15867 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
15868 attribute. Set view.
15869 (new_line_info_table): Reset next view.
15870 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
15871 (dwarf2out_source_line): Likewise. Output view resets and labels to
15872 the assembler, or select appropriate line info opcodes.
15873 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
15874 (optimize_string_length): Catch it. Adjust.
15875 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
15876 dw_val_class_view_list, and remove it if no longer needed.
15877 (hash_loc_list): Hash view numbers.
15878 (loc_list_hasher::equal): Compare them.
15879 (optimize_location_lists): Check whether a view list symbol is
15880 needed, and whether the locview attribute is present, and
15881 whether they match. Remove the locview attribute if no longer
15883 (index_location_lists): Call skip_loc_list_entry for test.
15884 (dwarf2out_finish): Call output_asm_line_debug_info for test.
15885 Use output_dwarf_version.
15886 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
15887 (struct dw_val_node): Add val_view_list.
15888 * final.c (SEEN_NEXT_VIEW): New.
15889 (set_next_view_needed): New.
15890 (clear_next_view_needed): New.
15891 (maybe_output_next_view): New.
15892 (final_start_function): Rename to...
15893 (final_start_function_1): ... this. Take pointer to FIRST,
15894 add SEEN parameter. Emit param bindings in the initial view.
15895 (final_start_function): Reintroduce SEEN-less interface.
15896 (final): Rename to...
15897 (final_1): ... this. Take SEEN parameter. Output final pending
15898 next view at the end.
15899 (final): Reintroduce seen-less interface.
15900 (final_scan_insn): Output pending next view before switching
15901 sections or ending a block. Mark the next view as needed when
15902 outputting variable locations. Notify debug backend of section
15903 changes, and of location view changes.
15904 (rest_of_handle_final): Adjust.
15905 * toplev.c (process_options): Autodetect value for debug variable
15906 location views option. Warn on incompat5 without -gdwarf-5.
15907 * doc/invoke.texi (gvariable-location-views): New.
15908 (gvariable-location-views=incompat5): New.
15909 (gno-variable-location-views): New.
15911 2018-02-08 David Malcolm <dmalcolm@redhat.com>
15913 PR tree-optimization/84136
15914 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
15915 that the result of find_edge is non-NULL.
15917 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15920 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
15921 storing integer register in SImode. Fix cost of 256 and 512
15922 byte aligned SSE register store.
15924 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15926 * config/i386/i386.c (ix86_multiplication_cost): Fix
15927 multiplication cost for TARGET_AVX512DQ.
15929 2018-02-08 Marek Polacek <polacek@redhat.com>
15931 PR tree-optimization/84238
15932 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
15935 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15937 PR tree-optimization/84265
15938 * tree-vect-stmts.c (vectorizable_store): Don't treat
15939 VMAT_CONTIGUOUS accesses as grouped.
15940 (vectorizable_load): Likewise.
15942 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15944 PR tree-optimization/81635
15945 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
15946 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
15947 (test_round_for_mask): New functions.
15948 (wide_int_cc_tests): Call test_round_for_mask.
15949 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
15950 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
15951 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
15952 range returned by get_range_info.
15954 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15957 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
15958 * symtab.c: Include builtins.h
15959 (symtab_node::output_to_lto_symbol_table_p): Move here
15960 from lto-streamer-out.c:output_symbol_p.
15961 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
15962 (output_symbol_p): Move all logic to symtab.c
15963 (produce_symtab): Update.
15965 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15967 * config/s390/s390-opts.h (enum indirect_branch): Define.
15968 * config/s390/s390-protos.h (s390_return_addr_from_memory)
15969 (s390_indirect_branch_via_thunk)
15970 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
15971 (enum s390_indirect_branch_type): Define.
15972 * config/s390/s390.c (struct s390_frame_layout, struct
15973 machine_function): Remove.
15974 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
15975 (indirect_branch_table_label_no, indirect_branch_table_name):
15977 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
15978 (enum s390_indirect_branch_option): Define.
15979 (s390_return_addr_from_memory): New function.
15980 (s390_handle_string_attribute): New function.
15981 (s390_attribute_table): Add new attribute handler.
15982 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
15983 (s390_indirect_branch_via_thunk): New function.
15984 (s390_indirect_branch_via_inline_thunk): New function.
15985 (s390_function_ok_for_sibcall): When jumping via thunk disallow
15986 sibling call optimization for non z10 compiles.
15987 (s390_emit_call): Force indirect branch target to be a single
15988 register. Add r1 clobber for non-z10 compiles.
15989 (s390_emit_epilogue): Emit return jump via return_use expander.
15990 (s390_reorg): Handle JUMP_INSNs as execute targets.
15991 (s390_option_override_internal): Perform validity checks for the
15992 new command line options.
15993 (s390_indirect_branch_attrvalue): New function.
15994 (s390_indirect_branch_settings): New function.
15995 (s390_set_current_function): Invoke s390_indirect_branch_settings.
15996 (s390_output_indirect_thunk_function): New function.
15997 (s390_code_end): Implement target hook.
15998 (s390_case_values_threshold): Implement target hook.
15999 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
16001 * config/s390/s390.h (struct s390_frame_layout)
16002 (struct machine_function): Move here from s390.c.
16003 (TARGET_INDIRECT_BRANCH_NOBP_RET)
16004 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
16005 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
16006 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
16007 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
16008 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
16009 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
16010 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
16011 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
16012 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
16013 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
16014 (mnemonic attribute): Add values which aren't recognized
16016 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
16017 pattern for branch conversion. Fix mnemonic attribute.
16018 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
16019 indirect branch via thunk if requested.
16020 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
16021 ("*indirect_jump"): Disable for branch conversion using out of
16023 ("indirect_jump_via_thunk<mode>_z10")
16024 ("indirect_jump_via_thunk<mode>")
16025 ("indirect_jump_via_inlinethunk<mode>_z10")
16026 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
16027 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
16028 ("casesi_jump_via_inlinethunk<mode>_z10")
16029 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
16030 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
16031 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
16032 ("*indirect2_jump"): Disable for branch conversion.
16033 ("casesi_jump"): Turn into expander and expand patterns for branch
16035 ("return_use"): New expander.
16036 ("*return"): Emit return via thunk and rename it to ...
16037 ("*return<mode>"): ... this one.
16038 * config/s390/s390.opt: Add new options and and enum for the
16041 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
16043 * lra-constraints.c (match_reload): Unconditionally use
16044 gen_lowpart_SUBREG, rather than selecting between that
16045 and equivalent gen_rtx_SUBREG code.
16047 2018-02-08 Richard Biener <rguenther@suse.de>
16049 PR tree-optimization/84233
16050 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
16051 changed flag instead of boguously re-using phi_inserted.
16053 2018-02-08 Martin Jambor <mjambor@suse.cz>
16055 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
16056 static local variables.
16058 2018-02-08 Richard Biener <rguenther@suse.de>
16060 PR tree-optimization/84278
16061 * tree-vect-stmts.c (vectorizable_store): When looking for
16062 smaller vector types to perform grouped strided loads/stores
16063 make sure the mode is supported by the target.
16064 (vectorizable_load): Likewise.
16066 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16068 * config/aarch64/aarch64.c (aarch64_components_for_bb):
16069 Increase LDP/STP opportunities by adding adjacent callee-saves.
16071 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
16073 PR rtl-optimization/84068
16074 PR rtl-optimization/83459
16075 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
16077 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
16079 PR tree-optimization/84224
16080 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
16081 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
16082 non-zero arguments.
16084 2018-02-07 Iain Sandoe <iain@codesourcery.com>
16087 * config/rs6000/altivec.md (*restore_world): Remove LR use.
16088 * config/rs6000/predicates.md (restore_world_operation): Adjust op
16089 count, remove one USE.
16091 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
16093 * doc/install.texi (Configuration): Document the
16094 --with-long-double-format={ibm,ieee} PowerPC configuration
16098 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
16099 Convert from define_expand to be define_insn_and_split. Rework
16100 float/double/_Float128 conversions to QI/HI/SImode to work with
16101 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
16102 conversions to QI/HImode types did a store and then a load to
16103 truncate the value. For conversions to VSX registers, don't split
16104 the insn, instead emit the code directly. Use the code iterator
16105 any_fix to combine signed and unsigned conversions.
16106 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
16107 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
16108 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16109 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
16110 (fix_<mode>di2_hw): Likewise.
16111 (fixuns_<mode>di2_hw): Likewise.
16112 (fix_<mode>si2_hw): Likewise.
16113 (fixuns_<mode>si2_hw): Likewise.
16114 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
16115 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
16116 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
16117 fix<uns>_trunc<SFDF:mode>si2_p8.
16118 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
16119 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
16120 (fix<uns>_<mode>_mem): Likewise.
16121 (fctiw<u>z_<mode>_mem): Likewise.
16122 (fix<uns>_<mode>_mem): Likewise.
16123 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
16124 the register allocator from doing a direct move to the GPRs to do
16125 a store, and instead use the ISA 3.0 store byte/half-word from
16126 vector register instruction. For IEEE 128-bit floating point,
16127 also optimize stores of 32-bit ints.
16128 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
16130 2018-02-07 Alan Hayward <alan.hayward@arm.com>
16132 * genextract.c (push_pathstr_operand): New function to support
16134 (walk_rtx): Call push_pathstr_operand.
16135 (print_path): Support [a-zA-Z].
16137 2018-02-07 Richard Biener <rguenther@suse.de>
16139 PR tree-optimization/84037
16140 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
16141 (cse_and_gimplify_to_preheader): Declare.
16142 (vect_get_place_in_interleaving_chain): Likewise.
16143 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16145 (_loop_vec_info::~_loop_vec_info): Delete it.
16146 (cse_and_gimplify_to_preheader): New function.
16147 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
16148 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
16149 (vectorizable_load): Likewise. For grouped stores always base
16150 the IV on the first element.
16151 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
16152 condition before gimplifying.
16154 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16156 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
16157 *DIV_EXPR and *MOD_EXPR.
16159 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
16162 * config/i386/i386.c (ix86_option_override_internal): Mask out
16163 the CF_SET bit when checking -fcf-protection.
16165 2018-02-07 Tom de Vries <tom@codesourcery.com>
16168 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
16171 2018-02-07 Richard Biener <rguenther@suse.de>
16173 PR tree-optimization/84204
16174 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
16177 PR tree-optimization/84205
16178 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
16179 special-case isl_ast_op_zdiv_r.
16181 PR tree-optimization/84223
16182 * graphite-scop-detection.c (gather_bbs::before_dom_children):
16183 Only add conditions from within the region.
16184 (gather_bbs::after_dom_children): Adjust.
16186 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
16189 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
16190 * config/avr/avr.md: Only post-reload split REG-REG moves if
16191 either register is GENERAL_REG_P.
16193 2018-02-07 Jakub Jelinek <jakub@redhat.com>
16195 PR tree-optimization/84235
16196 * tree-ssa-scopedtables.c
16197 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
16198 if the subtraction is performed in floating point type where NaNs are
16199 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
16200 build 1. Formatting fix.
16202 2018-02-06 Jakub Jelinek <jakub@redhat.com>
16205 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
16206 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
16207 and skip it regardless of bb boundaries. Use CALL_P macro,
16208 don't test INSN_P (insn) together with CALL_P or JUMP_P check
16209 unnecessarily, formatting fix.
16211 2018-02-06 Michael Collison <michael.collison@arm.com>
16213 * config/arm/thumb2.md:
16214 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
16215 (*thumb_mov_notscc): Ditto.
16217 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
16220 * config/rs6000/rs6000.md (su code attribute): Use "u" for
16221 unsigned_fix, not "s".
16223 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16225 * configure.ac (gcc_fn_eh_frame_ro): New function.
16226 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
16227 correct .eh_frame permissions.
16228 * configure: Regenerate.
16230 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
16232 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
16233 irrelevant options.
16235 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16237 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16238 Display warning message for -mno-speculate-indirect-jumps.
16240 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
16242 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
16244 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
16246 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
16248 PR tree-optimization/84225
16249 * tree-eh.c (find_trapping_overflow): Only call
16250 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
16252 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16255 * config/i386/i386.c: Reimplement the check of possible options
16256 -mibt/-mshstk conbination. Change error messages.
16257 * doc/invoke.texi: Fix a typo: remove extra '='.
16259 2018-02-06 Marek Polacek <polacek@redhat.com>
16261 PR tree-optimization/84228
16262 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
16264 2018-02-06 Tamar Christina <tamar.christina@arm.com>
16267 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
16268 emitted arch directives.
16269 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
16270 __ARM_FEATURE_COPROC before changing architectures.
16272 2018-02-06 Richard Biener <rguenther@suse.de>
16274 * config/i386/i386.c (print_reg): Fix typo.
16275 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
16277 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
16279 * configure: Regenerate.
16281 2018-02-05 Martin Sebor <msebor@redhat.com>
16283 PR tree-optimization/83369
16284 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
16287 2018-02-05 Martin Liska <mliska@suse.cz>
16289 * doc/invoke.texi: Cherry-pick upstream r323995.
16291 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
16293 * ira.c (ira_init_register_move_cost): Adjust comment.
16295 2018-02-05 Martin Liska <mliska@suse.cz>
16297 PR gcov-profile/84137
16298 * doc/gcov.texi: Fix typo in documentation.
16300 2018-02-05 Martin Liska <mliska@suse.cz>
16302 PR gcov-profile/83879
16303 * doc/gcov.texi: Document necessity of --dynamic-list-data when
16304 using dlopen functionality.
16306 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
16308 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
16309 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
16310 _mm_maskz_range_ss, _mm_mask_range_round_ss,
16311 _mm_maskz_range_round_ss): New intrinsics.
16312 (__builtin_ia32_rangesd128_round)
16313 (__builtin_ia32_rangess128_round): Remove.
16314 (__builtin_ia32_rangesd128_mask_round,
16315 __builtin_ia32_rangess128_mask_round): New builtins.
16316 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
16317 __builtin_ia32_rangess128_round): Remove.
16318 (__builtin_ia32_rangesd128_mask_round,
16319 __builtin_ia32_rangess128_mask_round): New builtins.
16320 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
16321 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
16322 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
16323 "<round_saeonly_constraint>")): Changed to ...
16324 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
16325 "<round_saeonly_scalar_constraint>")): ... this.
16326 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
16327 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
16328 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
16329 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
16330 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
16332 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
16334 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
16336 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
16337 Remove all values except native, 8540 and 8548.
16339 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
16341 * config/i386/i386.c (ix86_output_function_return): Pass
16342 INVALID_REGNUM, instead of -1, as invalid register number to
16343 indirect_thunk_name and output_indirect_thunk.
16345 2018-02-02 Julia Koval <julia.koval@intel.com>
16347 * config.gcc: Add -march=icelake.
16348 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
16349 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
16350 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
16351 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
16352 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
16353 (processor_target_table): Add icelake.
16354 (ix86_option_override_internal): Handle new PTAs.
16355 (get_builtin_code_for_version): Handle icelake.
16356 (M_INTEL_COREI7_ICELAKE): New.
16357 (fold_builtin_cpu): Handle icelake.
16358 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
16359 * doc/invoke.texi: Add -march=icelake.
16361 2018-02-02 Julia Koval <julia.koval@intel.com>
16363 * config/i386/i386.c (ix86_option_override_internal): Change flags type
16364 to wide_int_bitmask.
16365 * wide-int-bitmask.h: New.
16367 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16370 * config/i386/i386.md: Replace Pmode with word_mode in
16371 builtin_setjmp_setup and builtin_longjmp to support x32.
16373 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
16377 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
16379 (rs6000_supported_cpu_names): New static variable.
16380 (linux_cpu_translation_table): Likewise.
16381 (elf_platform) <cpu>: Define new static variable and use it.
16382 Translate kernel AT_PLATFORM name to canonical name if needed.
16383 Error if platform name is unknown.
16385 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
16388 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
16390 2018-02-01 Jeff Law <law@redhat.com>
16393 * config/i386/i386.c (release_scratch_register_on_entry): Add new
16394 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
16395 the scratch if RELEASE_VIA_POP is false.
16396 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
16397 If we have to save a temporary register, decrement SIZE appropriately.
16398 Pass new arguments to release_scratch_register_on_entry.
16399 (ix86_adjust_stack_and_probe): Likewise.
16400 (ix86_emit_probe_stack_range): Pass new arguments to
16401 release_scratch_register_on_entry.
16403 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
16405 PR rtl-optimization/84157
16406 * combine.c (change_zero_ext): Use REG_P predicate in
16407 front of HARD_REGISTER_P predicate.
16409 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
16411 * config/avr/avr.c (avr_option_override): Move disabling of
16412 -fdelete-null-pointer-checks to...
16413 * common/config/avr/avr-common.c (avr_option_optimization_table):
16416 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16418 PR tree-optimization/81635
16419 * tree-data-ref.c (split_constant_offset_1): For types that
16420 wrap on overflow, try to use range info to prove that wrapping
16423 2018-02-01 Renlin Li <renlin.li@arm.com>
16426 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
16427 TAILCALL_ADDR_REGS.
16428 (aarch64_register_move_cost): Likewise.
16429 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
16430 TAILCALL_ADDR_REGS.
16431 (REG_CLASS_NAMES): Likewise.
16432 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
16433 TAILCALL_ADDR_REGS. Remove IP registers.
16434 * config/aarch64/aarch64.md (Ucs): Update register constraint.
16436 2018-02-01 Richard Biener <rguenther@suse.de>
16438 * domwalk.h (dom_walker::dom_walker): Add additional constructor
16439 for specifying RPO order and allow NULL for that.
16440 * domwalk.c (dom_walker::dom_walker): Likewise.
16441 (dom_walker::walk): Handle NULL RPO order.
16442 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
16444 (rewrite_update_dom_walker): Likewise.
16445 (mark_def_dom_walker): Likewise.
16447 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16449 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
16450 (aarch64_maybe_expand_sve_subreg_move): Declare.
16451 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
16452 * config/aarch64/predicates.md (aarch64_any_register_operand): New
16454 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
16455 that are semantically a reverse operation.
16456 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
16457 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
16458 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
16460 (aarch64_can_change_mode_class): For big-endian, forbid changes
16461 between two SVE modes if they have different element sizes.
16463 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16465 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
16466 the TImode handling for big-endian targets.
16468 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16470 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
16471 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
16473 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
16474 Remove BSWAP handing for big-endian targets and use the form of
16475 LD1RQ appropariate for the mode.
16477 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16479 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
16480 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
16481 duplicated element.
16483 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16486 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
16487 check for operands that need to go through aarch64_sve_reload_be.
16489 2018-02-01 Jakub Jelinek <jakub@redhat.com>
16491 PR tree-optimization/81661
16492 PR tree-optimization/84117
16493 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
16494 * tree-eh.c: Include gimplify.h.
16495 (find_trapping_overflow, replace_trapping_overflow,
16496 rewrite_to_non_trapping_overflow): New functions.
16497 * tree-vect-loop.c: Include tree-eh.h.
16498 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
16499 * tree-data-ref.c: Include tree-eh.h.
16500 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
16502 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
16504 PR rtl-optimization/84123
16505 * combine.c (change_zero_ext): Check if hard register satisfies
16506 can_change_dest_mode before calling gen_lowpart_SUBREG.
16508 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
16511 * ira.c (ira_init_register_move_cost): Remove assert.
16513 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16515 PR rtl-optimization/84071
16516 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
16517 * doc/tm.texi: Regenerate.
16519 2018-01-31 Richard Biener <rguenther@suse.de>
16521 PR tree-optimization/84132
16522 * tree-data-ref.c (analyze_miv_subscript): Properly
16523 check whether evolution_function_is_affine_multivariate_p
16524 before calling gcd_of_steps_may_divide_p.
16526 2018-01-31 Julia Koval <julia.koval@intel.com>
16529 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
16530 * config/i386/i386.md (rdpid_rex64) New.
16531 (rdpid): Make 32bit only.
16533 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
16536 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
16537 an IDENTIFIER_NODE for FUNCTION_TYPE's.
16539 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16542 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16544 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16546 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16548 PR rtl-optimization/84071
16549 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
16550 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
16552 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16554 * config/arc/arc.c (arc_handle_aux_attribute): New function.
16555 (arc_attribute_table): Add 'aux' attribute.
16556 (arc_in_small_data_p): Consider aux like variables.
16557 (arc_is_aux_reg_p): New function.
16558 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
16559 (arc_get_aux_arg): New function.
16560 (prepare_move_operands): Handle aux-register access.
16561 (arc_handle_aux_attribute): New function.
16562 * doc/extend.texi (ARC Variable attributes): Add subsection.
16564 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16566 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
16567 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
16568 (arc_attribute_table): Add 'uncached' attribute.
16569 (arc_print_operand): Print '.di' flag for uncached memory
16571 (arc_in_small_data_p): Do not consider for small data the uncached
16573 (arc_is_uncached_mem_p): New function.
16574 * config/arc/predicates.md (compact_store_memory_operand): Check
16575 for uncached memory accesses.
16576 (nonvol_nonimm_operand): Likewise.
16577 * doc/extend.texi (ARC Type Attribute): New subsection.
16579 2018-01-31 Jakub Jelinek <jakub@redhat.com>
16582 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
16583 falign-loops=): Add Optimization flag.
16585 2018-01-30 Jeff Law <law@redhat.com>
16588 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
16589 INT_REGISTERS_SAVED. Check it prior to calling
16590 get_scratch_register_on_entry.
16591 (ix86_adjust_stack_and_probe): Similarly.
16592 (ix86_emit_probe_stack_range): Similarly.
16593 (ix86_expand_prologue): Corresponding changes.
16595 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16598 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
16599 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
16601 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
16604 * lra-constraints.c (curr_insn_transform): Process AND in the
16607 2018-01-30 Jakub Jelinek <jakub@redhat.com>
16609 PR rtl-optimization/83986
16610 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
16611 dependence against last_pending_memory_flush in addition to
16612 pending_jump_insns.
16614 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
16616 PR tree-optimization/81611
16617 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
16620 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16623 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
16626 2018-01-30 Richard Biener <rguenther@suse.de>
16627 Jakub Jelinek <jakub@redhat.com>
16629 PR tree-optimization/84111
16630 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
16631 inner loops added during recursion, as they don't have up-to-date
16634 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16637 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
16638 (can_inline_edge_by_limits_p): ... here.
16639 (can_early_inline_edge_p, check_callers,
16640 update_caller_keys, update_callee_keys, recursive_inlining,
16641 add_new_edges_to_heap, speculation_useful_p,
16642 inline_small_functions,
16643 inline_small_functions, flatten_function,
16644 inline_to_all_callers_1): Update.
16646 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16648 * profile-count.c (profile_count::combine_with_ipa_count): Handle
16651 2018-01-30 Richard Biener <rguenther@suse.de>
16653 PR tree-optimization/83008
16654 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
16655 invariant and constant vector uses in stmts when they need
16656 more than one stmt.
16658 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16661 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
16662 * configure: Regenerate.
16664 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16666 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
16668 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
16669 Use gen_rtx_REG rather than gen_lowpart.
16671 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16673 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
16674 rather than 0 when creating partial subregs.
16676 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16678 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
16681 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
16684 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
16685 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
16686 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
16687 flags. This restores the settings used before the 2017-07-24.
16688 Turning off pre increment/decrement/modify allows IVOPTS to
16689 optimize DF/SF loops where the index is an int.
16691 2018-01-29 Richard Biener <rguenther@suse.de>
16692 Kelvin Nilsen <kelvin@gcc.gnu.org>
16695 * tree-vect-stmts.c (vectorizable_call): Don't call
16696 targetm.vectorize_builtin_md_vectorized_function if callee is
16699 2018-01-22 Carl Love <cel@us.ibm.com>
16701 * doc/extend.tex: Fix typo in second arg in
16702 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
16704 2018-01-29 Richard Biener <rguenther@suse.de>
16706 PR tree-optimization/84086
16707 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
16708 (flush_ssaname_freelist): When SSA names were released reset
16709 the SCEV hash table.
16711 2018-01-29 Richard Biener <rguenther@suse.de>
16713 PR tree-optimization/84057
16714 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
16715 removed paths when removing edges.
16717 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
16719 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
16720 -mfunction-return=@var{choice}.
16722 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
16724 PR diagnostic/84034
16725 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
16726 Handle CR like TAB.
16727 (layout::print_source_line): Likewise.
16728 (test_get_line_width_without_trailing_whitespace): Add test cases.
16730 2018-01-27 Jakub Jelinek <jakub@redhat.com>
16732 PR middle-end/84040
16733 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
16736 2018-01-26 Jim Wilson <jimw@sifive.com>
16738 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
16740 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
16743 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16745 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
16746 and CMP + SUB-immediate -> SUBS.
16748 2018-01-26 Martin Sebor <msebor@redhat.com>
16750 PR tree-optimization/83896
16751 * tree-ssa-strlen.c (get_string_len): Rename...
16752 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
16753 Avoid assuming length is constant.
16754 (handle_char_store): Use HOST_WIDE_INT for string length.
16756 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
16759 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
16760 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
16762 2018-01-26 Richard Biener <rguenther@suse.de>
16764 PR rtl-optimization/84003
16765 * dse.c (record_store): Only record redundant stores when
16766 the earlier store aliases at least all accesses the later one does.
16768 2018-01-26 Jakub Jelinek <jakub@redhat.com>
16770 PR rtl-optimization/83985
16771 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
16772 REG_CFA_RESTORE insns.
16773 (delete_unmarked_insns): Don't ignore separate shrink wrapping
16774 REG_CFA_RESTORE insns here.
16777 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
16778 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
16780 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16782 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
16783 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
16784 (arc_init): Likewise.
16785 (arc_override_options): Likewise.
16786 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
16788 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
16790 * config/arc/arc.h (TARGET_DBNZ): Define.
16791 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
16792 properly set the tune attribute.
16793 (dbnz): Use TARGET_DBNZ guard.
16794 * config/arc/arc.opt (mtune): Add core3 option.
16796 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16798 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
16799 recognize new pic like addresses.
16800 (arc_delegitimize_address): Clean up.
16802 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16804 * config/arc/arc-arches.def: Option mrf16 valid for all
16806 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
16807 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
16808 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
16809 * config/arc/arc-tables.opt: Regenerate.
16810 * config/arc/arc.c (arc_conditional_register_usage): Handle
16811 reduced register file case.
16812 (arc_file_start): Set must have build attributes.
16813 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
16814 mrf16 option value.
16815 * config/arc/arc.opt (mrf16): Add new option.
16816 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
16817 * config/arc/genmultilib.awk: Handle new mrf16 option.
16818 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
16819 * config/arc/t-multilib: Regenerate.
16820 * doc/invoke.texi (ARC Options): Document mrf16 option.
16822 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16824 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
16825 * config/arc/arc.c (arc_handle_secure_attribute): New function.
16826 (arc_attribute_table): Add 'secure_call' attribute.
16827 (arc_print_operand): Print secure call operand.
16828 (arc_function_ok_for_sibcall): Don't optimize tail calls when
16830 (arc_is_secure_call_p): New function. * config/arc/arc.md
16831 (call_i): Add support for sjli instruction.
16832 (call_value_i): Likewise.
16833 * config/arc/constraints.md (Csc): New constraint.
16835 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16836 John Eric Martin <John.Martin@emmicro-us.com>
16838 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
16839 * config/arc/arc.c (_arc_jli_section): New struct.
16840 (arc_jli_section): New type.
16841 (rc_jli_sections): New static variable.
16842 (arc_handle_jli_attribute): New function.
16843 (arc_attribute_table): Add jli_always and jli_fixed attribute.
16844 (arc_file_end): New function.
16845 (TARGET_ASM_FILE_END): Define.
16846 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
16847 (arc_add_jli_section): New function.
16848 (jli_call_scan): Likewise.
16849 (arc_reorg): Call jli_call_scan.
16850 (arc_output_addsi): Remove 'S' from printing asm operand.
16851 (arc_is_jli_call_p): New function.
16852 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
16854 (movhi_insn): Likewise.
16855 (movsi_insn): Likewise.
16856 (movsi_set_cc_insn): Likewise.
16857 (loadqi_update): Likewise.
16858 (load_zeroextendqisi_update): Likewise.
16859 (load_signextendqisi_update): Likewise.
16860 (loadhi_update): Likewise.
16861 (load_zeroextendhisi_update): Likewise.
16862 (load_signextendhisi_update): Likewise.
16863 (loadsi_update): Likewise.
16864 (loadsf_update): Likewise.
16865 (movsicc_insn): Likewise.
16866 (bset_insn): Likewise.
16867 (bxor_insn): Likewise.
16868 (bclr_insn): Likewise.
16869 (bmsk_insn): Likewise.
16870 (bicsi3_insn): Likewise.
16871 (cmpsi_cc_c_insn): Likewise.
16872 (movsi_ne): Likewise.
16873 (movsi_cond_exec): Likewise.
16874 (clrsbsi2): Likewise.
16875 (norm_f): Likewise.
16885 (call_i): Remove 'S' asm letter, add jli instruction.
16886 (call_value_i): Likewise.
16887 * config/arc/arc.op (mjli-always): New option.
16888 * config/arc/constraints.md (Cji): New constraint.
16889 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
16891 (subsf3_fpx): Likewise.
16892 (mulsf3_fpx): Likewise.
16893 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
16895 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
16896 function attrbutes.
16897 * doc/invoke.texi (ARC): Document mjli-always option.
16899 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
16901 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
16902 avoid addition with 0 and use incw and decw where possible.
16904 2018-01-26 Richard Biener <rguenther@suse.de>
16906 PR tree-optimization/81082
16907 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
16908 association if it requires casting to unsigned.
16909 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
16910 from fold_plusminus_mult_expr to catch important cases late when
16911 range info is available.
16913 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16915 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
16916 * configure.ac (hidden_linkonce): New test.
16917 * configure: Regenerate.
16918 * config.in: Regenerate.
16920 2018-01-26 Julia Koval <julia.koval@intel.com>
16922 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
16923 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
16924 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
16925 _mm_mask_bitshuffle_epi64_mask): Fix type.
16926 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
16927 USI_FTYPE_V4DI_V4DI_USI): Remove.
16928 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
16929 __builtin_ia32_vpshufbitqmb256_mask,
16930 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
16931 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
16932 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
16934 2018-01-26 Alan Modra <amodra@gmail.com>
16937 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
16938 UNSPEC_VBPERMQ. Sort other unspecs.
16940 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
16942 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
16944 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
16946 PR middle-end/83055
16947 * predict.c (drop_profile): Do not push/pop cfun; update also
16949 (handle_missing_profiles): Fix logic looking for zero profiles.
16951 2018-01-25 Jakub Jelinek <jakub@redhat.com>
16953 PR middle-end/83977
16954 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
16955 on functions with #pragma omp declare simd or functions with simd
16957 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
16958 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
16959 Remove trailing \n from warning_at calls.
16961 2018-01-25 Tom de Vries <tom@codesourcery.com>
16964 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16965 for neutered workers.
16967 2018-01-24 Joseph Myers <joseph@codesourcery.com>
16970 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
16971 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
16973 2018-01-24 Jeff Law <law@redhat.com>
16976 * i386.c (get_probe_interval): Move to earlier point.
16977 (ix86_compute_frame_layout): If -fstack-clash-protection and
16978 the frame is larger than the probe interval, then use pushes
16979 to save registers rather than reg->mem moves.
16980 (ix86_expand_prologue): Remove conditional for int_registers_saved
16983 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
16986 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
16987 min/max for never referenced object.
16989 2018-01-24 Jakub Jelinek <jakub@redhat.com>
16991 PR middle-end/83977
16992 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
16994 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
16995 attributes from DECL_ATTRIBUTES (decl) without affecting
16996 DECL_ATTRIBUTES (current_function_decl).
16997 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
16998 functions with non-NULL DECL_ABSTRACT_ORIGIN.
17000 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
17002 PR tree-optimization/83979
17003 * fold-const.c (fold_comparison): Use constant_boolean_node
17004 instead of boolean_{true,false}_node.
17006 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
17008 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
17011 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17013 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17014 Simplify the clause that sets the length attribute.
17015 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17016 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
17017 clause that sets the length attribute.
17018 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17020 2018-01-24 Tom de Vries <tom@codesourcery.com>
17023 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
17024 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
17025 Add strict parameter.
17026 (prevent_branch_around_nothing): Insert dummy insn between branch to
17027 label and label with no ptx insn inbetween.
17028 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
17030 2018-01-24 Tom de Vries <tom@codesourcery.com>
17033 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
17034 for neutered threads in warp.
17035 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
17037 2018-01-24 Richard Biener <rguenther@suse.de>
17039 PR tree-optimization/83176
17040 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
17043 2018-01-24 Richard Biener <rguenther@suse.de>
17045 PR tree-optimization/82819
17046 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
17047 code generating pluses that are no-ops in the target precision.
17049 2018-01-24 Richard Biener <rguenther@suse.de>
17051 PR middle-end/84000
17052 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
17054 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17056 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
17057 to merge probabilities.
17058 * predict.c (probably_never_executed): Also mark as cold functions
17059 with global 0 profile and guessed local profile.
17060 * profile-count.c (profile_probability::combine_with_count): New
17062 * profile-count.h (profile_probability::operator*,
17063 profile_probability::operator*=, profile_probability::operator/,
17064 profile_probability::operator/=): Reduce precision to adjusted
17065 and set value to guessed on contradictory divisions.
17066 (profile_probability::combine_with_freq): Remove.
17067 (profile_probability::combine_wiht_count): Declare.
17068 (profile_count::force_nonzero):: Set to adjusted.
17069 (profile_count::probability_in):: Set quality to adjusted.
17070 * tree-ssa-tail-merge.c (replace_block_by): Use
17071 combine_with_count.
17073 2018-01-23 Andrew Waterman <andrew@sifive.com>
17074 Jim Wilson <jimw@sifive.com>
17076 * config/riscv/riscv.c (riscv_stack_boundary): New.
17077 (riscv_option_override): Set riscv_stack_boundary. Handle
17078 riscv_preferred_stack_boundary_arg.
17079 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
17080 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
17081 (STACK_BOUNDARY): Set to riscv_stack_boundary.
17082 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
17083 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
17084 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
17086 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
17089 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
17090 of struct ix86_frame.
17091 (ix86_expand_epilogue): Likewise. Add a local variable for
17092 the reg_save_offset field in struct ix86_frame.
17094 2018-01-23 Bin Cheng <bin.cheng@arm.com>
17096 PR tree-optimization/82604
17097 * tree-loop-distribution.c (enum partition_kind): New enum item
17098 PKIND_PARTIAL_MEMSET.
17099 (partition_builtin_p): Support above new enum item.
17100 (generate_code_for_partition): Ditto.
17101 (compute_access_range): Differentiate cases that equality can be
17102 proven at all loops, the innermost loops or no loops.
17103 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
17104 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
17105 (finalize_partitions, distribute_loop): Don't fuse partition of
17106 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
17107 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
17108 parloop is enabled.
17110 2018-01-23 Martin Liska <mliska@suse.cz>
17112 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
17113 order to ignore the predictor.
17114 (PRED_POLYMORPHIC_CALL): Likewise.
17115 (PRED_RECURSIVE_CALL): Likewise.
17117 2018-01-23 Martin Liska <mliska@suse.cz>
17119 * tree-profile.c (tree_profiling): Print function header to
17120 aware reader which function we are working on.
17121 * value-prof.c (gimple_find_values_to_profile): Do not print
17122 not interesting value histograms.
17124 2018-01-23 Martin Liska <mliska@suse.cz>
17126 * profile-count.h (enum profile_quality): Add
17127 profile_uninitialized as the first value. Do not number values
17128 as they are zero based.
17129 (profile_count::verify): Update sanity check.
17130 (profile_probability::verify): Likewise.
17132 2018-01-23 Nathan Sidwell <nathan@acm.org>
17134 * doc/invoke.texi (ffor-scope): Deprecate.
17136 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17138 PR tree-optimization/83510
17139 * domwalk.c (set_all_edges_as_executable): New function.
17140 (dom_walker::dom_walker): Convert bool param
17141 "skip_unreachable_blocks" to enum reachability. Move setup of
17142 edge flags to set_all_edges_as_executable and only do it when
17143 reachability is REACHABLE_BLOCKS.
17144 * domwalk.h (enum dom_walker::reachability): New enum.
17145 (dom_walker::dom_walker): Convert bool param
17146 "skip_unreachable_blocks" to enum reachability.
17147 (set_all_edges_as_executable): New decl.
17148 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
17149 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
17151 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
17152 but converting true to REACHABLE_BLOCKS.
17153 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
17155 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
17156 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
17157 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
17159 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
17160 if check_all_array_refs will be called.
17162 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17164 * tree.c (selftest::test_location_wrappers): Add more test
17167 2018-01-23 David Malcolm <dmalcolm@redhat.com>
17169 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
17170 (selftest::test_bit_in_range): Likewise.
17172 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17175 * doc/sourcebuild.texi (vect_float): Say that the selector
17176 only describes the situation when -funsafe-math-optimizations is on.
17177 (vect_float_strict): Document.
17179 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
17181 PR tree-optimization/83965
17182 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17183 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
17184 instead of checking only for a reduction.
17185 (vect_recog_widen_sum_pattern): Likewise.
17187 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17189 * predict.c (probably_never_executed): Only use precise profile info.
17190 (compute_function_frequency): Skip after inlining hack since we now
17191 have quality checking.
17193 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
17195 * profile-count.h (profile_probability::very_unlikely,
17196 profile_probability::unlikely, profile_probability::even): Set
17197 precision to guessed.
17199 2018-01-23 Richard Biener <rguenther@suse.de>
17201 PR tree-optimization/83963
17202 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
17203 Properly terminate dominator walk when crossing the exit edge not
17204 when visiting its source block.
17206 2018-01-23 Jakub Jelinek <jakub@redhat.com>
17209 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
17210 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
17212 2018-01-22 Jakub Jelinek <jakub@redhat.com>
17214 PR tree-optimization/83957
17215 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
17216 semicolon after for body surrounded by braces.
17218 PR tree-optimization/83081
17219 * profile-count.h (profile_probability::split): New method.
17220 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
17221 Use profile_probability::split.
17222 (do_compare_rtx_and_jump): Fix adjustment of probabilities
17223 when splitting a single conditional jump into 2.
17225 2018-01-22 David Malcolm <dmalcolm@redhat.com>
17227 PR tree-optimization/69452
17228 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
17231 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17233 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
17234 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
17235 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
17237 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17239 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
17241 * config/rl78/rl78.md (movdi): New define_expand.
17242 * config/rl78/rl78.c (rl78_split_movdi): New function.
17244 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
17247 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
17249 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
17250 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
17251 128-bit to produce an UNSPEC move to get the double word with the
17252 signbit and then a shift directly to do signbit.
17253 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
17254 implementation with a new version that just does either a direct
17255 move or a regular move. Move memory interface to separate insns.
17256 Move insns so they are next to the expander.
17257 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
17258 with signbit move. Split big and little endian case.
17259 (signbit<mode>2_dm_mem_le): Likewise.
17260 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
17261 (signbit<mode>2_dm2): Likewise.
17263 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17265 * config/rl78/rl78.md (anddi3): New define_expand.
17267 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17269 * config/rl78/rl78.md (umindi3): New define_expand.
17271 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17273 * config/rl78/rl78.md (smindi3): New define_expand.
17275 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17277 * config/rl78/rl78.md (smaxdi3): New define_expand.
17279 2018-01-22 Carl Love <cel@us.ibm.com>
17281 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
17282 LVX_V1TI): Add macro expansion.
17283 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
17284 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
17285 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
17286 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
17287 Change check to determine if the instruction is a byte reversing
17288 entry. Fix typo in comment.
17289 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
17290 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
17291 Add def_builtin calls for new builtins.
17292 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
17293 Add define_insn expansion.
17295 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17297 * config/rl78/rl78.md (umaxdi3): New define_expand.
17299 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
17301 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
17302 for non-QImode registers.
17304 2018-01-22 Richard Biener <rguenther@suse.de>
17306 PR tree-optimization/83963
17307 * graphite-scop-detection.c (scop_detection::get_sese): Delay
17308 including the loop exit block.
17309 (scop_detection::merge_sese): Likewise.
17310 (scop_detection::add_scop): Do it here instead.
17312 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17314 * doc/sourcebuild.texi (arm_softfloat): Document.
17316 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
17319 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
17320 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
17321 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
17323 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17324 David Edelsohn <dje.gcc@gmail.com>
17327 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17328 Change "crset eq" to "crset 2".
17329 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17330 (*call_indirect_aix<mode>_nospec): Likewise.
17331 (*call_value_indirect_aix<mode>_nospec): Likewise.
17332 (*call_indirect_elfv2<mode>_nospec): Likewise.
17333 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
17334 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
17335 change assembly output from . to $.
17336 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17337 (indirect_jump<mode>_nospec): Change assembly output from . to $.
17338 (*tablejump<mode>_internal1_nospec): Likewise.
17340 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
17343 * config/sh/sh_optimize_sett_clrt.cc:
17344 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
17346 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
17348 PR tree-optimization/83940
17349 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
17350 offset_dt to vect_constant_def rather than vect_unknown_def_type.
17351 (vect_check_load_store_mask): Add a mask_dt_out parameter and
17352 use it to pass back the definition type.
17353 (vect_check_store_rhs): Likewise rhs_dt_out.
17354 (vect_build_gather_load_calls): Add a mask_dt argument and use
17355 it instead of a call to vect_is_simple_use.
17356 (vectorizable_store): Update calls to vect_check_load_store_mask
17357 and vect_check_store_rhs. Use the dt returned by the latter instead
17358 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
17359 instead of calls to vect_is_simple_use. Pass the scalar rather
17360 than the vector operand to vect_is_simple_use when handling
17361 second and subsequent copies of an rhs value.
17362 (vectorizable_load): Update calls to vect_check_load_store_mask
17363 and vect_build_gather_load_calls. Use the cached mask_dt and
17364 gs_info.offset_dt instead of calls to vect_is_simple_use.
17366 2018-01-20 Jakub Jelinek <jakub@redhat.com>
17368 PR middle-end/83945
17369 * tree-emutls.c: Include gimplify.h.
17370 (lower_emutls_2): New function.
17371 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
17372 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
17373 it before further processing.
17376 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
17377 UINTVAL (trueop1) instead of INTVAL (op1).
17379 2018-01-19 Jakub Jelinek <jakub@redhat.com>
17383 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
17384 INCOMING_FRAME_SP_OFFSET if not defined.
17385 (scan_trace): Add ENTRY argument. If true and
17386 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
17387 emit a note to adjust the CFA offset.
17388 (create_cfi_notes): Adjust scan_trace callers.
17389 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
17390 INCOMING_FRAME_SP_OFFSET in the CIE.
17391 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
17392 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
17394 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
17395 * doc/tm.texi: Regenerated.
17397 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17399 PR rtl-optimization/83147
17400 * lra-constraints.c (remove_inheritance_pseudos): Use
17401 lra_substitute_pseudo_within_insn.
17403 2018-01-19 Tom de Vries <tom@codesourcery.com>
17404 Cesar Philippidis <cesar@codesourcery.com>
17407 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
17409 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
17412 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
17413 spaces for function labels.
17415 2018-01-19 Martin Liska <mliska@suse.cz>
17417 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
17418 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
17419 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
17420 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
17421 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
17422 (PRED_CONST_RETURN): Change from 69 to 65.
17423 (PRED_NULL_RETURN): Change from 91 to 71.
17424 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
17425 (PRED_LOOP_GUARD): Change from 66 to 73.
17427 2018-01-19 Martin Liska <mliska@suse.cz>
17429 * predict.c (predict_insn_def): Add new assert.
17430 (struct branch_predictor): Change type to signed integer.
17431 (test_prediction_value_range): Amend test to cover
17432 PROB_UNINITIALIZED.
17433 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
17434 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
17435 (PRED_LOOP_ITERATIONS_MAX): Likewise.
17436 (PRED_LOOP_IV_COMPARE): Likewise.
17437 * predict.h (PROB_UNINITIALIZED): Define new constant.
17439 2018-01-19 Martin Liska <mliska@suse.cz>
17441 * predict.c (dump_prediction): Add new format for
17442 analyze_brprob.py script which is enabled with -details
17444 * profile-count.h (precise_p): New function.
17446 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17448 PR tree-optimization/83922
17449 * tree-vect-loop.c (vect_verify_full_masking): Return false if
17450 there are no statements that need masking.
17451 (vect_active_double_reduction_p): New function.
17452 (vect_analyze_loop_operations): Use it when handling phis that
17453 are not in the loop header.
17455 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17457 PR tree-optimization/83914
17458 * tree-vect-loop.c (vectorizable_induction): Don't convert
17459 init_expr or apply the peeling adjustment for inductions
17460 that are nested within the vectorized loop.
17462 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17464 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
17467 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17471 * function.h (gimplify_parameters): Add gimple_seq * argument.
17472 * function.c: Include gimple.h and options.h.
17473 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
17474 for the added local temporaries if needed.
17475 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
17476 if there are any parameter cleanups, wrap whole body into a
17477 try/finally with the cleanups.
17479 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
17482 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
17483 Use GET_MODE_CLASS for scalar floating point.
17485 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17489 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
17490 Fix call of call_cgraph_insertion_hooks.
17492 2018-01-18 Martin Sebor <msebor@redhat.com>
17494 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
17496 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17499 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
17502 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
17505 * common.opt: (-ffile-prefix-map): New option.
17506 * opts.c (common_handle_option): Defer it.
17507 * opts-global.c (handle_common_deferred_options): Handle it.
17508 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
17509 * file-prefix-map.h: New file.
17510 (remap_debug_filename, add_debug_prefix_map): ...here.
17511 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
17512 * final.c (debug_prefix_map, add_debug_prefix_map
17513 remap_debug_filename): Move to...
17514 * file-prefix-map.c: New file.
17515 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
17516 generalize, get rid of alloca(), use strrchr() instead of strchr().
17517 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
17518 Implement in terms of add_prefix_map().
17519 (remap_macro_filename, remap_debug_filename): Implement in term of
17521 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
17522 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
17523 * dbxout.c: Include file-prefix-map.h.
17524 * varasm.c: Likewise.
17525 * vmsdbgout.c: Likewise.
17526 * xcoffout.c: Likewise.
17527 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
17528 * doc/cppopts.texi (-fmacro-prefix-map): Document.
17529 * doc/invoke.texi (-ffile-prefix-map): Document.
17530 (-fdebug-prefix-map): Update description.
17532 2018-01-18 Martin Liska <mliska@suse.cz>
17534 * config/i386/i386.c (indirect_thunk_name): Document that also
17536 (output_indirect_thunk): Document why both instructions
17537 (pause and lfence) are generated.
17539 2018-01-18 Richard Biener <rguenther@suse.de>
17541 PR tree-optimization/83887
17542 * graphite-scop-detection.c
17543 (scop_detection::get_nearest_dom_with_single_entry): Remove.
17544 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
17545 (scop_detection::merge_sese): Re-implement with a flood-fill
17546 algorithm that properly finds a SESE region if it exists.
17548 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17551 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
17552 pointer_diff optimizations use view_convert instead of convert.
17554 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17556 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17557 Generate different code for -mno-speculate-indirect-jumps.
17558 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17559 (*call_indirect_aix<mode>): Disable for
17560 -mno-speculate-indirect-jumps.
17561 (*call_indirect_aix<mode>_nospec): New define_insn.
17562 (*call_value_indirect_aix<mode>): Disable for
17563 -mno-speculate-indirect-jumps.
17564 (*call_value_indirect_aix<mode>_nospec): New define_insn.
17565 (*sibcall_nonlocal_sysv<mode>): Generate different code for
17566 -mno-speculate-indirect-jumps.
17567 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17569 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17571 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
17572 long double type, set the flags for noting the default long double
17573 type, even if we don't pass or return a long double type.
17575 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
17578 * ipa-inline.c (flatten_function): Do not overwrite final inlining
17581 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
17583 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
17584 support for merge[hl].
17585 (fold_mergehl_helper): New helper function.
17586 (tree-vector-builder.h): New #include for tree_vector_builder usage.
17587 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
17588 (altivec_vmrglw_direct): Add xxmrglw insn.
17590 2018-01-17 Andrew Waterman <andrew@sifive.com>
17592 * config/riscv/riscv.c (riscv_conditional_register_usage): If
17593 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
17595 2018-01-17 David Malcolm <dmalcolm@redhat.com>
17598 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
17599 call the lto_location_cache before reading the
17600 DECL_SOURCE_LOCATION of the types.
17602 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
17603 Richard Sandiford <richard.sandiford@linaro.org>
17605 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
17606 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
17607 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
17608 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
17609 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
17611 * config/aarch64/constraints.md (aarch64_movti_operand):
17613 * config/aarch64/predicates.md (Uti): Add new constraint.
17615 2018-01-17 Carl Love <cel@us.ibm.com>
17617 * config/rs6000/vsx.md (define_expand xl_len_r,
17618 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
17619 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
17621 (define_expand, define_insn): Move the shift left from the
17622 define_insn to the define_expand for lxvl and stxvl instructions.
17623 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
17624 and XL_LEN_R definitions to PURE.
17626 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
17628 * config/i386/i386.c (indirect_thunk_name): Declare regno
17629 as unsigned int. Compare regno with INVALID_REGNUM.
17630 (output_indirect_thunk): Ditto.
17631 (output_indirect_thunk_function): Ditto.
17632 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
17633 in the call to output_indirect_thunk_function.
17635 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
17637 PR middle-end/83884
17638 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
17639 rather than the size of inner_type to determine the stack slot size
17640 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
17642 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
17645 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
17648 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
17650 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
17651 endian Linux systems to optionally enable multilibs for selecting
17652 the long double type if the user configured an explicit type.
17653 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
17654 have no long double multilibs if not defined.
17655 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
17656 warn if the user used -mabi={ieee,ibm}longdouble and we built
17657 multilibs for long double.
17658 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
17659 appropriate multilib option.
17660 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
17662 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
17663 for building long double multilibs.
17664 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
17666 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
17668 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
17671 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
17673 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
17676 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
17679 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
17682 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17684 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
17685 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
17687 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
17689 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
17690 different rtl trees depending on TARGET_64BIT.
17691 (rs6000_gen_lvx): Likewise.
17693 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17695 * config/visium/visium.md (nop): Tweak comment.
17696 (hazard_nop): Likewise.
17698 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17700 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
17701 -mspeculate-indirect-jumps.
17702 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
17703 for -mno-speculate-indirect-jumps.
17704 (*call_indirect_elfv2<mode>_nospec): New define_insn.
17705 (*call_value_indirect_elfv2<mode>): Disable for
17706 -mno-speculate-indirect-jumps.
17707 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
17708 (indirect_jump): Emit different RTL for
17709 -mno-speculate-indirect-jumps.
17710 (*indirect_jump<mode>): Disable for
17711 -mno-speculate-indirect-jumps.
17712 (*indirect_jump<mode>_nospec): New define_insn.
17713 (tablejump): Emit different RTL for
17714 -mno-speculate-indirect-jumps.
17715 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
17716 (tablejumpsi_nospec): New define_expand.
17717 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
17718 (tablejumpdi_nospec): New define_expand.
17719 (*tablejump<mode>_internal1): Disable for
17720 -mno-speculate-indirect-jumps.
17721 (*tablejump<mode>_internal1_nospec): New define_insn.
17722 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
17725 2018-01-16 Artyom Skrobov tyomitch@gmail.com
17727 * caller-save.c (insert_save): Drop unnecessary parameter. All
17730 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17731 Richard Biener <rguenth@suse.de>
17734 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
17735 return early, inline manually is_gimple_sizepos. Make sure if we
17736 call gimplify_expr we don't end up with a gimple constant.
17737 * tree.c (variably_modified_type_p): Don't return true for
17738 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
17739 * gimplify.h (is_gimple_sizepos): Remove.
17741 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17743 PR tree-optimization/83857
17744 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
17745 vectorizable_live_operation for pure SLP statements.
17746 (vectorizable_live_operation): Handle PHIs.
17748 2018-01-16 Richard Biener <rguenther@suse.de>
17750 PR tree-optimization/83867
17751 * tree-vect-stmts.c (vect_transform_stmt): Precompute
17752 nested_in_vect_loop_p since the scalar stmt may get invalidated.
17754 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17757 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
17758 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
17759 If off is not INTEGER_CST, issue a may not be aligned warning
17760 rather than isn't aligned. Use isn%'t rather than isn't.
17761 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
17763 <case MULT_EXPR>: Improve the case when bottom and one of the
17764 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
17765 operand, in that case check if the other operand is multiple of
17766 bottom divided by the INTEGER_CST operand.
17768 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17771 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
17772 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
17773 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
17774 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
17775 * config/pa/pa.c (pa_function_arg_advance): Likewise.
17776 (pa_function_arg, pa_arg_partial_bytes): Likewise.
17777 (pa_function_arg_size): New function.
17779 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17781 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
17782 in a separate statement.
17784 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17786 PR tree-optimization/83847
17787 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
17788 group gathers and scatters.
17790 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17792 PR rtl-optimization/86620
17793 * params.def (max-sched-ready-insns): Bump minimum value to 1.
17795 PR rtl-optimization/83213
17796 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
17797 to last if both are JUMP_INSNs.
17799 PR tree-optimization/83843
17800 * gimple-ssa-store-merging.c
17801 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
17802 store_immediate_info for bswap/nop orig_stores.
17804 2018-01-15 Andrew Waterman <andrew@sifive.com>
17806 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
17808 <UDIV>: Increase cost if !TARGET_DIV.
17810 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
17812 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
17813 (define_attr "cr_logical_3op"): New.
17814 (cceq_ior_compare): Adjust.
17815 (cceq_ior_compare_complement): Adjust.
17816 (*cceq_rev_compare): Adjust.
17817 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
17818 (is_cracked_insn): Adjust.
17819 (insn_must_be_first_in_group): Adjust.
17820 * config/rs6000/40x.md: Adjust.
17821 * config/rs6000/440.md: Adjust.
17822 * config/rs6000/476.md: Adjust.
17823 * config/rs6000/601.md: Adjust.
17824 * config/rs6000/603.md: Adjust.
17825 * config/rs6000/6xx.md: Adjust.
17826 * config/rs6000/7450.md: Adjust.
17827 * config/rs6000/7xx.md: Adjust.
17828 * config/rs6000/8540.md: Adjust.
17829 * config/rs6000/cell.md: Adjust.
17830 * config/rs6000/e300c2c3.md: Adjust.
17831 * config/rs6000/e500mc.md: Adjust.
17832 * config/rs6000/e500mc64.md: Adjust.
17833 * config/rs6000/e5500.md: Adjust.
17834 * config/rs6000/e6500.md: Adjust.
17835 * config/rs6000/mpc.md: Adjust.
17836 * config/rs6000/power4.md: Adjust.
17837 * config/rs6000/power5.md: Adjust.
17838 * config/rs6000/power6.md: Adjust.
17839 * config/rs6000/power7.md: Adjust.
17840 * config/rs6000/power8.md: Adjust.
17841 * config/rs6000/power9.md: Adjust.
17842 * config/rs6000/rs64.md: Adjust.
17843 * config/rs6000/titan.md: Adjust.
17845 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17847 * config/i386/predicates.md (indirect_branch_operand): Rewrite
17848 ix86_indirect_branch_register logic.
17850 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17852 * config/i386/constraints.md (Bs): Update
17853 ix86_indirect_branch_register check. Don't check
17854 ix86_indirect_branch_register with GOT_memory_operand.
17856 * config/i386/predicates.md (GOT_memory_operand): Don't check
17857 ix86_indirect_branch_register here.
17858 (GOT32_symbol_operand): Likewise.
17860 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17862 * config/i386/predicates.md (constant_call_address_operand):
17863 Rewrite ix86_indirect_branch_register logic.
17864 (sibcall_insn_operand): Likewise.
17866 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17868 * config/i386/constraints.md (Bs): Replace
17869 ix86_indirect_branch_thunk_register with
17870 ix86_indirect_branch_register.
17872 * config/i386/i386.md (indirect_jump): Likewise.
17873 (tablejump): Likewise.
17874 (*sibcall_memory): Likewise.
17875 (*sibcall_value_memory): Likewise.
17876 Peepholes of indirect call and jump via memory: Likewise.
17877 * config/i386/i386.opt: Likewise.
17878 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17879 (GOT_memory_operand): Likewise.
17880 (call_insn_operand): Likewise.
17881 (sibcall_insn_operand): Likewise.
17882 (GOT32_symbol_operand): Likewise.
17884 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17886 PR middle-end/83837
17887 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
17888 type rather than type addr's type points to.
17889 (expand_omp_atomic_mutex): Likewise.
17890 (expand_omp_atomic): Likewise.
17892 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17895 * config/i386/i386.c (output_indirect_thunk_function): Use
17896 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
17897 for __x86_return_thunk.
17899 2018-01-15 Richard Biener <rguenther@suse.de>
17901 PR middle-end/83850
17902 * expmed.c (extract_bit_field_1): Fix typo.
17904 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17907 * config/arm/iterators.md (VF): New mode iterator.
17908 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
17909 Remove integer-related logic from pattern.
17910 (neon_vabd<mode>_3): Likewise.
17912 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17914 PR middle-end/82694
17915 * common.opt (fstrict-overflow): No longer an alias.
17916 (fwrapv-pointer): New option.
17917 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
17918 also for pointer types based on flag_wrapv_pointer.
17919 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
17920 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
17921 opts->x_flag_wrapv got set.
17922 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
17923 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
17924 POINTER_TYPE_OVERFLOW_UNDEFINED.
17925 * match.pd: Likewise in address comparison pattern.
17926 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
17928 2018-01-15 Richard Biener <rguenther@suse.de>
17931 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
17932 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
17933 Reset type names to their identifier if their TYPE_DECL doesn't
17934 have linkage (and thus is used for ODR and devirt).
17935 (save_debug_info_for_decl): Remove.
17936 (save_debug_info_for_type): Likewise.
17937 (add_tree_to_fld_list): Adjust.
17938 * tree-pretty-print.c (dump_generic_node): Make dumping of
17939 type names more robust.
17941 2018-01-15 Richard Biener <rguenther@suse.de>
17943 * BASE-VER: Bump to 8.0.1.
17945 2018-01-14 Martin Sebor <msebor@redhat.com>
17948 * builtins.c (check_access): Avoid warning when the no-warning bit
17951 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
17953 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
17954 * ira-color (allocno_hard_regs_compare): Likewise.
17956 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
17959 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
17960 Use .pushsection/.popsection.
17962 2018-01-14 Martin Sebor <msebor@redhat.com>
17965 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
17967 2018-01-14 Jakub Jelinek <jakub@redhat.com>
17969 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
17970 entry from extra_headers.
17971 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
17972 extra_headers, make the list bitwise identical to the i?86-*-* one.
17974 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17976 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
17977 -mcmodel=large with -mindirect-branch=thunk,
17978 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
17979 -mfunction-return=thunk-extern.
17980 * doc/invoke.texi: Document -mcmodel=large is incompatible with
17981 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
17982 -mfunction-return=thunk and -mfunction-return=thunk-extern.
17984 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17986 * config/i386/i386.c (print_reg): Print the name of the full
17987 integer register without '%'.
17988 (ix86_print_operand): Handle 'V'.
17989 * doc/extend.texi: Document 'V' modifier.
17991 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17993 * config/i386/constraints.md (Bs): Disallow memory operand for
17994 -mindirect-branch-register.
17996 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17997 (GOT_memory_operand): Likewise.
17998 (call_insn_operand): Likewise.
17999 (sibcall_insn_operand): Likewise.
18000 (GOT32_symbol_operand): Likewise.
18001 * config/i386/i386.md (indirect_jump): Call convert_memory_address
18002 for -mindirect-branch-register.
18003 (tablejump): Likewise.
18004 (*sibcall_memory): Likewise.
18005 (*sibcall_value_memory): Likewise.
18006 Disallow peepholes of indirect call and jump via memory for
18007 -mindirect-branch-register.
18008 (*call_pop): Replace m with Bw.
18009 (*call_value_pop): Likewise.
18010 (*sibcall_pop_memory): Replace m with Bs.
18011 * config/i386/i386.opt (mindirect-branch-register): New option.
18012 * doc/invoke.texi: Document -mindirect-branch-register option.
18014 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18016 * config/i386/i386-protos.h (ix86_output_function_return): New.
18017 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
18018 set function_return_type.
18019 (indirect_thunk_name): Add ret_p to indicate thunk for function
18021 (output_indirect_thunk_function): Pass false to
18022 indirect_thunk_name.
18023 (ix86_output_indirect_branch_via_reg): Likewise.
18024 (ix86_output_indirect_branch_via_push): Likewise.
18025 (output_indirect_thunk_function): Create alias for function
18026 return thunk if regno < 0.
18027 (ix86_output_function_return): New function.
18028 (ix86_handle_fndecl_attribute): Handle function_return.
18029 (ix86_attribute_table): Add function_return.
18030 * config/i386/i386.h (machine_function): Add
18031 function_return_type.
18032 * config/i386/i386.md (simple_return_internal): Use
18033 ix86_output_function_return.
18034 (simple_return_internal_long): Likewise.
18035 * config/i386/i386.opt (mfunction-return=): New option.
18036 (indirect_branch): Mention -mfunction-return=.
18037 * doc/extend.texi: Document function_return function attribute.
18038 * doc/invoke.texi: Document -mfunction-return= option.
18040 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
18042 * config/i386/i386-opts.h (indirect_branch): New.
18043 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
18044 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
18045 with local indirect jump when converting indirect call and jump.
18046 (ix86_set_indirect_branch_type): New.
18047 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
18048 (indirectlabelno): New.
18049 (indirect_thunk_needed): Likewise.
18050 (indirect_thunk_bnd_needed): Likewise.
18051 (indirect_thunks_used): Likewise.
18052 (indirect_thunks_bnd_used): Likewise.
18053 (INDIRECT_LABEL): Likewise.
18054 (indirect_thunk_name): Likewise.
18055 (output_indirect_thunk): Likewise.
18056 (output_indirect_thunk_function): Likewise.
18057 (ix86_output_indirect_branch_via_reg): Likewise.
18058 (ix86_output_indirect_branch_via_push): Likewise.
18059 (ix86_output_indirect_branch): Likewise.
18060 (ix86_output_indirect_jmp): Likewise.
18061 (ix86_code_end): Call output_indirect_thunk_function if needed.
18062 (ix86_output_call_insn): Call ix86_output_indirect_branch if
18064 (ix86_handle_fndecl_attribute): Handle indirect_branch.
18065 (ix86_attribute_table): Add indirect_branch.
18066 * config/i386/i386.h (machine_function): Add indirect_branch_type
18067 and has_local_indirect_jump.
18068 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
18070 (tablejump): Likewise.
18071 (*indirect_jump): Use ix86_output_indirect_jmp.
18072 (*tablejump_1): Likewise.
18073 (simple_return_indirect_internal): Likewise.
18074 * config/i386/i386.opt (mindirect-branch=): New option.
18075 (indirect_branch): New.
18078 (thunk-inline): Likewise.
18079 (thunk-extern): Likewise.
18080 * doc/extend.texi: Document indirect_branch function attribute.
18081 * doc/invoke.texi: Document -mindirect-branch= option.
18083 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
18086 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
18088 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
18090 * ipa-inline.c (want_inline_small_function_p): Return false if
18091 inlining has already failed with CIF_FINAL_ERROR.
18092 (update_caller_keys): Call want_inline_small_function_p before
18094 (update_callee_keys): Likewise.
18096 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18098 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
18100 (rs6000_quadword_masked_address_p): Likewise.
18101 (quad_aligned_load_p): Likewise.
18102 (quad_aligned_store_p): Likewise.
18103 (const_load_sequence_p): Add comment to describe the outer-most loop.
18104 (mimic_memory_attributes_and_flags): New function.
18105 (rs6000_gen_stvx): Likewise.
18106 (replace_swapped_aligned_store): Likewise.
18107 (rs6000_gen_lvx): Likewise.
18108 (replace_swapped_aligned_load): Likewise.
18109 (replace_swapped_load_constant): Capitalize argument name in
18110 comment describing this function.
18111 (rs6000_analyze_swaps): Add a third pass to search for vector loads
18112 and stores that access quad-word aligned addresses and replace
18113 with stvx or lvx instructions when appropriate.
18114 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
18115 New function prototype.
18116 (rs6000_quadword_masked_address_p): Likewise.
18117 (rs6000_gen_lvx): Likewise.
18118 (rs6000_gen_stvx): Likewise.
18119 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
18120 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
18121 when memory address is aligned.
18122 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
18123 this split to select lvx instruction when memory address is aligned.
18124 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
18125 instruction when memory address is aligned.
18126 (*vsx_le_perm_load_v16qi): Likewise.
18127 (four unnamed splitters): Modify to select the stvx instruction
18128 when memory is aligned.
18130 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
18132 * predict.c (determine_unlikely_bbs): Handle correctly BBs
18133 which appears in the queue multiple times.
18135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18136 Alan Hayward <alan.hayward@arm.com>
18137 David Sherwood <david.sherwood@arm.com>
18139 * tree-vectorizer.h (vec_lower_bound): New structure.
18140 (_loop_vec_info): Add check_nonzero and lower_bounds.
18141 (LOOP_VINFO_CHECK_NONZERO): New macro.
18142 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
18143 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
18144 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
18145 fields. Make seg_len the distance travelled, not including the
18147 (dr_direction_indicator): Declare.
18148 (dr_zero_step_indicator): Likewise.
18149 (dr_known_forward_stride_p): Likewise.
18150 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
18152 (runtime_alias_check_p): Allow runtime alias checks with
18154 (operator ==): Compare access_size and align.
18155 (prune_runtime_alias_test_list): Rework for new distinction between
18156 the access_size and seg_len.
18157 (create_intersect_range_checks_index): Likewise. Cope with polynomial
18159 (get_segment_min_max): New function.
18160 (create_intersect_range_checks): Use it.
18161 (dr_step_indicator): New function.
18162 (dr_direction_indicator): Likewise.
18163 (dr_zero_step_indicator): Likewise.
18164 (dr_known_forward_stride_p): Likewise.
18165 * tree-loop-distribution.c (data_ref_segment_size): Return
18166 DR_STEP * (niters - 1).
18167 (compute_alias_check_pairs): Update call to the dr_with_seg_len
18169 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
18170 (vect_preserves_scalar_order_p): New function, split out from...
18171 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
18172 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
18173 (vect_vfa_access_size): New function.
18174 (vect_vfa_align): Likewise.
18175 (vect_compile_time_alias): Take access_size_a and access_b arguments.
18176 (dump_lower_bound): New function.
18177 (vect_check_lower_bound): Likewise.
18178 (vect_small_gap_p): Likewise.
18179 (vectorizable_with_step_bound_p): Likewise.
18180 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
18181 depencies if the vectorization factor is 1. Convert the checks
18182 for nonzero steps into checks on the bounds of DR_STEP. Try using
18183 a bunds check for variable steps if the minimum required step is
18184 relatively small. Update calls to the dr_with_seg_len
18185 constructor and to vect_compile_time_alias.
18186 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
18188 (vect_loop_versioning): Call it.
18189 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
18191 (vect_estimate_min_profitable_iters): Account for any bounds checks.
18193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18194 Alan Hayward <alan.hayward@arm.com>
18195 David Sherwood <david.sherwood@arm.com>
18197 * doc/sourcebuild.texi (vect_scatter_store): Document.
18198 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
18200 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
18202 * genopinit.c (main): Add supports_vec_scatter_store and
18203 supports_vec_scatter_store_cached to target_optabs.
18204 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
18205 IFN_MASK_SCATTER_STORE.
18206 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
18208 * internal-fn.h (internal_store_fn_p): Declare.
18209 (internal_fn_stored_value_index): Likewise.
18210 * internal-fn.c (scatter_store_direct): New macro.
18211 (expand_scatter_store_optab_fn): New function.
18212 (direct_scatter_store_optab_supported_p): New macro.
18213 (internal_store_fn_p): New function.
18214 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
18215 IFN_MASK_SCATTER_STORE.
18216 (internal_fn_mask_index): Likewise.
18217 (internal_fn_stored_value_index): New function.
18218 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
18219 for scatter stores.
18220 * optabs-query.h (supports_vec_scatter_store_p): Declare.
18221 * optabs-query.c (supports_vec_scatter_store_p): New function.
18222 * tree-vectorizer.h (vect_get_store_rhs): Declare.
18223 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
18224 true for scatter stores.
18225 (vect_gather_scatter_fn_p): Handle scatter stores too.
18226 (vect_check_gather_scatter): Consider using scatter stores if
18227 supports_vec_scatter_store_p.
18228 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
18229 scatter stores too.
18230 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
18231 internal_fn_stored_value_index.
18232 (check_load_store_masking): Handle scatter stores too.
18233 (vect_get_store_rhs): Make public.
18234 (vectorizable_call): Use internal_store_fn_p.
18235 (vectorizable_store): Handle scatter store internal functions.
18236 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
18237 when deciding whether the end of the group has been reached.
18238 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
18239 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
18240 (mask_scatter_store<mode>): New insns.
18242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18243 Alan Hayward <alan.hayward@arm.com>
18244 David Sherwood <david.sherwood@arm.com>
18246 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
18247 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
18248 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
18250 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
18251 Use vect_truncate_gather_scatter_offset if we can't treat the
18252 operation as a normal gather load or scatter store.
18253 (get_group_load_store_type): Take the gather_scatter_info
18254 as argument. Try using a gather load or scatter store for
18255 single-element groups.
18256 (get_load_store_type): Update calls to get_group_load_store_type
18257 and vect_use_strided_gather_scatters_p.
18259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18260 Alan Hayward <alan.hayward@arm.com>
18261 David Sherwood <david.sherwood@arm.com>
18263 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
18264 optional tree argument.
18265 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
18267 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
18268 but continue to use the current value as a fallback.
18269 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
18270 to compare the updates.
18271 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
18272 (get_load_store_type): Use it when handling a strided access.
18273 (vect_get_strided_load_store_ops): New function.
18274 (vect_get_data_ptr_increment): Likewise.
18275 (vectorizable_load): Handle strided gather loads. Always pass
18276 a step to vect_create_data_ref_ptr and bump_vector_ptr.
18278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18279 Alan Hayward <alan.hayward@arm.com>
18280 David Sherwood <david.sherwood@arm.com>
18282 * doc/md.texi (gather_load@var{m}): Document.
18283 (mask_gather_load@var{m}): Likewise.
18284 * genopinit.c (main): Add supports_vec_gather_load and
18285 supports_vec_gather_load_cached to target_optabs.
18286 * optabs-tree.c (init_tree_optimization_optabs): Use
18287 ggc_cleared_alloc to allocate target_optabs.
18288 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
18289 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
18291 * internal-fn.h (internal_load_fn_p): Declare.
18292 (internal_gather_scatter_fn_p): Likewise.
18293 (internal_fn_mask_index): Likewise.
18294 (internal_gather_scatter_fn_supported_p): Likewise.
18295 * internal-fn.c (gather_load_direct): New macro.
18296 (expand_gather_load_optab_fn): New function.
18297 (direct_gather_load_optab_supported_p): New macro.
18298 (direct_internal_fn_optab): New function.
18299 (internal_load_fn_p): Likewise.
18300 (internal_gather_scatter_fn_p): Likewise.
18301 (internal_fn_mask_index): Likewise.
18302 (internal_gather_scatter_fn_supported_p): Likewise.
18303 * optabs-query.c (supports_at_least_one_mode_p): New function.
18304 (supports_vec_gather_load_p): Likewise.
18305 * optabs-query.h (supports_vec_gather_load_p): Declare.
18306 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
18307 and memory_type field.
18308 (NUM_PATTERNS): Bump to 15.
18309 * tree-vect-data-refs.c: Include internal-fn.h.
18310 (vect_gather_scatter_fn_p): New function.
18311 (vect_describe_gather_scatter_call): Likewise.
18312 (vect_check_gather_scatter): Try using internal functions for
18313 gather loads. Recognize existing calls to a gather load function.
18314 (vect_analyze_data_refs): Consider using gather loads if
18315 supports_vec_gather_load_p.
18316 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
18317 (vect_get_gather_scatter_offset_type): Likewise.
18318 (vect_convert_mask_for_vectype): Likewise.
18319 (vect_add_conversion_to_patterm): Likewise.
18320 (vect_try_gather_scatter_pattern): Likewise.
18321 (vect_recog_gather_scatter_pattern): New pattern recognizer.
18322 (vect_vect_recog_func_ptrs): Add it.
18323 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
18324 internal_fn_mask_index and internal_gather_scatter_fn_p.
18325 (check_load_store_masking): Take the gather_scatter_info as an
18326 argument and handle gather loads.
18327 (vect_get_gather_scatter_ops): New function.
18328 (vectorizable_call): Check internal_load_fn_p.
18329 (vectorizable_load): Likewise. Handle gather load internal
18331 (vectorizable_store): Update call to check_load_store_masking.
18332 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
18333 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
18334 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
18335 (aarch64_gather_scale_operand_d): New predicates.
18336 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
18337 (mask_gather_load<mode>): New insns.
18339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18340 Alan Hayward <alan.hayward@arm.com>
18341 David Sherwood <david.sherwood@arm.com>
18343 * optabs.def (fold_left_plus_optab): New optab.
18344 * doc/md.texi (fold_left_plus_@var{m}): Document.
18345 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
18346 * internal-fn.c (fold_left_direct): Define.
18347 (expand_fold_left_optab_fn): Likewise.
18348 (direct_fold_left_optab_supported_p): Likewise.
18349 * fold-const-call.c (fold_const_fold_left): New function.
18350 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
18351 * tree-parloops.c (valid_reduction_p): New function.
18352 (gather_scalar_reductions): Use it.
18353 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
18354 (vect_finish_replace_stmt): Declare.
18355 * tree-vect-loop.c (fold_left_reduction_fn): New function.
18356 (needs_fold_left_reduction_p): New function, split out from...
18357 (vect_is_simple_reduction): ...here. Accept reductions that
18358 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
18359 (vect_force_simple_reduction): Also store the reduction type in
18360 the assignment's STMT_VINFO_REDUC_TYPE.
18361 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
18362 (merge_with_identity): New function.
18363 (vect_expand_fold_left): Likewise.
18364 (vectorize_fold_left_reduction): Likewise.
18365 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
18366 scalar phi in place for it. Check for target support and reject
18367 cases that would reassociate the operation. Defer the transform
18368 phase to vectorize_fold_left_reduction.
18369 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
18370 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
18371 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
18373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18375 * tree-if-conv.c (predicate_mem_writes): Remove redundant
18376 call to ifc_temp_var.
18378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18379 Alan Hayward <alan.hayward@arm.com>
18380 David Sherwood <david.sherwood@arm.com>
18382 * target.def (legitimize_address_displacement): Take the original
18383 offset as a poly_int.
18384 * targhooks.h (default_legitimize_address_displacement): Update
18386 * targhooks.c (default_legitimize_address_displacement): Likewise.
18387 * doc/tm.texi: Regenerate.
18388 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
18389 as an argument, moving assert of ad->disp == ad->disp_term to...
18390 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
18391 Try calling targetm.legitimize_address_displacement before expanding
18392 the address rather than afterwards, and adjust for the new interface.
18393 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
18394 Match the new hook interface. Handle SVE addresses.
18395 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
18396 new hook interface.
18398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18400 * Makefile.in (OBJS): Add early-remat.o.
18401 * target.def (select_early_remat_modes): New hook.
18402 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
18403 * doc/tm.texi: Regenerate.
18404 * targhooks.h (default_select_early_remat_modes): Declare.
18405 * targhooks.c (default_select_early_remat_modes): New function.
18406 * timevar.def (TV_EARLY_REMAT): New timevar.
18407 * passes.def (pass_early_remat): New pass.
18408 * tree-pass.h (make_pass_early_remat): Declare.
18409 * early-remat.c: New file.
18410 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
18412 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
18414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18415 Alan Hayward <alan.hayward@arm.com>
18416 David Sherwood <david.sherwood@arm.com>
18418 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
18419 vfm1 with a bound_epilog parameter.
18420 (vect_do_peeling): Update calls accordingly, and move the prologue
18421 call earlier in the function. Treat the base bound_epilog as 0 for
18422 fully-masked loops and retain vf - 1 for other loops. Add 1 to
18423 this base when peeling for gaps.
18424 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
18425 with fully-masked loops.
18426 (vect_estimate_min_profitable_iters): Handle the single peeled
18427 iteration in that case.
18429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18430 Alan Hayward <alan.hayward@arm.com>
18431 David Sherwood <david.sherwood@arm.com>
18433 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
18434 single-element interleaving even if the size is not a power of 2.
18435 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
18436 accesses for single-element interleaving if the group size is
18439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18440 Alan Hayward <alan.hayward@arm.com>
18441 David Sherwood <david.sherwood@arm.com>
18443 * doc/md.texi (fold_extract_last_@var{m}): Document.
18444 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
18445 * optabs.def (fold_extract_last_optab): New optab.
18446 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
18447 * internal-fn.c (fold_extract_direct): New macro.
18448 (expand_fold_extract_optab_fn): Likewise.
18449 (direct_fold_extract_optab_supported_p): Likewise.
18450 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
18451 * tree-vect-loop.c (vect_model_reduction_cost): Handle
18452 EXTRACT_LAST_REDUCTION.
18453 (get_initial_def_for_reduction): Do not create an initial vector
18454 for EXTRACT_LAST_REDUCTION reductions.
18455 (vectorizable_reduction): Leave the scalar phi in place for
18456 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
18457 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
18458 epilogue code for EXTRACT_LAST_REDUCTION and defer the
18459 transform phase to vectorizable_condition.
18460 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
18462 (vect_finish_stmt_generation): ...here.
18463 (vect_finish_replace_stmt): New function.
18464 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
18465 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
18467 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
18469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18470 Alan Hayward <alan.hayward@arm.com>
18471 David Sherwood <david.sherwood@arm.com>
18473 * doc/md.texi (extract_last_@var{m}): Document.
18474 * optabs.def (extract_last_optab): New optab.
18475 * internal-fn.def (EXTRACT_LAST): New internal function.
18476 * internal-fn.c (cond_unary_direct): New macro.
18477 (expand_cond_unary_optab_fn): Likewise.
18478 (direct_cond_unary_optab_supported_p): Likewise.
18479 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
18480 loops using EXTRACT_LAST.
18481 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
18482 (extract_last_<mode>): ...this optab.
18483 (vec_extract<mode><Vel>): Update accordingly.
18485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18486 Alan Hayward <alan.hayward@arm.com>
18487 David Sherwood <david.sherwood@arm.com>
18489 * target.def (empty_mask_is_expensive): New hook.
18490 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
18491 * doc/tm.texi: Regenerate.
18492 * targhooks.h (default_empty_mask_is_expensive): Declare.
18493 * targhooks.c (default_empty_mask_is_expensive): New function.
18494 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
18495 if the target says that empty masks are expensive.
18496 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
18498 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
18500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18501 Alan Hayward <alan.hayward@arm.com>
18502 David Sherwood <david.sherwood@arm.com>
18504 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
18505 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
18506 (vect_use_loop_mask_for_alignment_p): New function.
18507 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
18508 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
18509 niters_skip argument. Make sure that the first niters_skip elements
18510 of the first iteration are inactive.
18511 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
18512 Update call to vect_set_loop_masks_directly.
18513 (get_misalign_in_elems): New function, split out from...
18514 (vect_gen_prolog_loop_niters): ...here.
18515 (vect_update_init_of_dr): Take a code argument that specifies whether
18516 the adjustment should be added or subtracted.
18517 (vect_update_init_of_drs): Likewise.
18518 (vect_prepare_for_masked_peels): New function.
18519 (vect_do_peeling): Skip prologue peeling if we're using a mask
18520 instead. Update call to vect_update_inits_of_drs.
18521 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18523 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
18524 alignment. Do not include the number of peeled iterations in
18525 the minimum threshold in that case.
18526 (vectorizable_induction): Adjust the start value down by
18527 LOOP_VINFO_MASK_SKIP_NITERS iterations.
18528 (vect_transform_loop): Call vect_prepare_for_masked_peels.
18529 Take the number of skipped iterations into account when calculating
18531 * tree-vect-stmts.c (vect_gen_while_not): New function.
18533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18534 Alan Hayward <alan.hayward@arm.com>
18535 David Sherwood <david.sherwood@arm.com>
18537 * doc/sourcebuild.texi (vect_fully_masked): Document.
18538 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
18539 default value to 0.
18540 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
18542 (vect_analyze_loop_2): ...here. Don't check the vectorization
18543 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
18551 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
18552 (dump_groups): Update accordingly.
18553 (iv_use::mem_type): New member variable.
18554 (address_p): New function.
18555 (record_use): Add a mem_type argument and initialize the new
18557 (record_group_use): Add a mem_type argument. Use address_p.
18558 Remove obsolete null checks of base_object. Update call to record_use.
18559 (find_interesting_uses_op): Update call to record_group_use.
18560 (find_interesting_uses_cond): Likewise.
18561 (find_interesting_uses_address): Likewise.
18562 (get_mem_type_for_internal_fn): New function.
18563 (find_address_like_use): Likewise.
18564 (find_interesting_uses_stmt): Try find_address_like_use before
18565 calling find_interesting_uses_op.
18566 (addr_offset_valid_p): Use the iv mem_type field as the type
18567 of the addressed memory.
18568 (add_autoinc_candidates): Likewise.
18569 (get_address_cost): Likewise.
18570 (split_small_address_groups_p): Use address_p.
18571 (split_address_groups): Likewise.
18572 (add_iv_candidate_for_use): Likewise.
18573 (autoinc_possible_for_pair): Likewise.
18574 (rewrite_groups): Likewise.
18575 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
18576 (determine_group_iv_cost): Update after split of USE_ADDRESS.
18577 (get_alias_ptr_type_for_ptr_address): New function.
18578 (rewrite_use_address): Rewrite address uses in calls that were
18579 identified by find_address_like_use.
18581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18582 Alan Hayward <alan.hayward@arm.com>
18583 David Sherwood <david.sherwood@arm.com>
18585 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
18587 * gimple-expr.h (is_gimple_addressable: Likewise.
18588 * gimple-expr.c (is_gimple_address): Likewise.
18589 * internal-fn.c (expand_call_mem_ref): New function.
18590 (expand_mask_load_optab_fn): Use it.
18591 (expand_mask_store_optab_fn): Likewise.
18593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18594 Alan Hayward <alan.hayward@arm.com>
18595 David Sherwood <david.sherwood@arm.com>
18597 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
18598 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
18599 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
18600 (cond_umax@var{mode}): Document.
18601 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
18602 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
18603 (cond_umin_optab, cond_umax_optab): New optabs.
18604 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
18605 (COND_IOR, COND_XOR): New internal functions.
18606 * internal-fn.h (get_conditional_internal_fn): Declare.
18607 * internal-fn.c (cond_binary_direct): New macro.
18608 (expand_cond_binary_optab_fn): Likewise.
18609 (direct_cond_binary_optab_supported_p): Likewise.
18610 (get_conditional_internal_fn): New function.
18611 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
18612 Cope with reduction statements that are vectorized as calls rather
18614 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
18615 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
18616 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
18617 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18618 (UNSPEC_COND_EOR): New unspecs.
18619 (optab): Add mappings for them.
18620 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
18621 (sve_int_op, sve_fp_op): New int attributes.
18623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18624 Alan Hayward <alan.hayward@arm.com>
18625 David Sherwood <david.sherwood@arm.com>
18627 * optabs.def (while_ult_optab): New optab.
18628 * doc/md.texi (while_ult@var{m}@var{n}): Document.
18629 * internal-fn.def (WHILE_ULT): New internal function.
18630 * internal-fn.h (direct_internal_fn_supported_p): New override
18631 that takes two types as argument.
18632 * internal-fn.c (while_direct): New macro.
18633 (expand_while_optab_fn): New function.
18634 (convert_optab_supported_p): Likewise.
18635 (direct_while_optab_supported_p): New macro.
18636 * wide-int.h (wi::udiv_ceil): New function.
18637 * tree-vectorizer.h (rgroup_masks): New structure.
18638 (vec_loop_masks): New typedef.
18639 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
18640 and fully_masked_p.
18641 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
18642 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
18643 (vect_max_vf): New function.
18644 (slpeel_make_loop_iterate_ntimes): Delete.
18645 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
18646 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
18647 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
18648 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
18649 internal-fn.h, stor-layout.h and optabs-query.h.
18650 (vect_set_loop_mask): New function.
18651 (add_preheader_seq): Likewise.
18652 (add_header_seq): Likewise.
18653 (interleave_supported_p): Likewise.
18654 (vect_maybe_permute_loop_masks): Likewise.
18655 (vect_set_loop_masks_directly): Likewise.
18656 (vect_set_loop_condition_masked): Likewise.
18657 (vect_set_loop_condition_unmasked): New function, split out from
18658 slpeel_make_loop_iterate_ntimes.
18659 (slpeel_make_loop_iterate_ntimes): Rename to..
18660 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
18661 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
18662 (vect_do_peeling): Update call accordingly.
18663 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
18665 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18666 mask_compare_type, can_fully_mask_p and fully_masked_p.
18667 (release_vec_loop_masks): New function.
18668 (_loop_vec_info): Use it to free the loop masks.
18669 (can_produce_all_loop_masks_p): New function.
18670 (vect_get_max_nscalars_per_iter): Likewise.
18671 (vect_verify_full_masking): Likewise.
18672 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
18673 retries, and free the mask rgroups before retrying. Check loop-wide
18674 reasons for disallowing fully-masked loops. Make the final decision
18675 about whether use a fully-masked loop or not.
18676 (vect_estimate_min_profitable_iters): Do not assume that peeling
18677 for the number of iterations will be needed for fully-masked loops.
18678 (vectorizable_reduction): Disable fully-masked loops.
18679 (vectorizable_live_operation): Likewise.
18680 (vect_halve_mask_nunits): New function.
18681 (vect_double_mask_nunits): Likewise.
18682 (vect_record_loop_mask): Likewise.
18683 (vect_get_loop_mask): Likewise.
18684 (vect_transform_loop): Handle the case in which the final loop
18685 iteration might handle a partial vector. Call vect_set_loop_condition
18686 instead of slpeel_make_loop_iterate_ntimes.
18687 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
18688 (check_load_store_masking): New function.
18689 (prepare_load_store_mask): Likewise.
18690 (vectorizable_store): Handle fully-masked loops.
18691 (vectorizable_load): Likewise.
18692 (supportable_widening_operation): Use vect_halve_mask_nunits for
18694 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
18695 (vect_gen_while): New function.
18696 * config/aarch64/aarch64.md (umax<mode>3): New expander.
18697 (aarch64_uqdec<mode>): New insn.
18699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18700 Alan Hayward <alan.hayward@arm.com>
18701 David Sherwood <david.sherwood@arm.com>
18703 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
18704 (reduc_xor_scal_optab): New optabs.
18705 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
18706 (reduc_xor_scal_@var{m}): Document.
18707 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
18708 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
18709 internal functions.
18710 * fold-const-call.c (fold_const_call): Handle them.
18711 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
18712 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
18713 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
18714 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
18715 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
18716 (UNSPEC_XORV): New unspecs.
18717 (optab): Add entries for them.
18718 (BITWISEV): New int iterator.
18719 (bit_reduc_op): New int attributes.
18721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18722 Alan Hayward <alan.hayward@arm.com>
18723 David Sherwood <david.sherwood@arm.com>
18725 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
18726 * internal-fn.def (VEC_SHL_INSERT): New internal function.
18727 * optabs.def (vec_shl_insert_optab): New optab.
18728 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
18729 (duplicate_and_interleave): Likewise.
18730 * tree-vect-loop.c: Include internal-fn.h.
18731 (neutral_op_for_slp_reduction): New function, split out from
18732 get_initial_defs_for_reduction.
18733 (get_initial_def_for_reduction): Handle option 2 for variable-length
18734 vectors by loading the neutral value into a vector and then shifting
18735 the initial value into element 0.
18736 (get_initial_defs_for_reduction): Replace the code argument with
18737 the neutral value calculated by neutral_op_for_slp_reduction.
18738 Use gimple_build_vector for constant-length vectors.
18739 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
18740 but the first group_size elements have a neutral value.
18741 Use duplicate_and_interleave otherwise.
18742 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
18743 Update call to get_initial_defs_for_reduction. Handle SLP
18744 reductions for variable-length vectors by creating one vector
18745 result for each scalar result, with the elements associated
18746 with other scalar results stubbed out with the neutral value.
18747 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
18748 Require IFN_VEC_SHL_INSERT for double reductions on
18749 variable-length vectors, or SLP reductions that have
18750 a neutral value. Require can_duplicate_and_interleave_p
18751 support for variable-length unchained SLP reductions if there
18752 is no neutral value, such as for MIN/MAX reductions. Also require
18753 the number of vector elements to be a multiple of the number of
18754 SLP statements when doing variable-length unchained SLP reductions.
18755 Update call to vect_create_epilog_for_reduction.
18756 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
18757 and remove initial values.
18758 (duplicate_and_interleave): Make public.
18759 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
18760 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
18762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18763 Alan Hayward <alan.hayward@arm.com>
18764 David Sherwood <david.sherwood@arm.com>
18766 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
18767 (can_duplicate_and_interleave_p): New function.
18768 (vect_get_and_check_slp_defs): Take the vector of statements
18769 rather than just the current one. Remove excess parentheses.
18770 Restriction rejectinon of vect_constant_def and vect_external_def
18771 for variable-length vectors to boolean types, or types for which
18772 can_duplicate_and_interleave_p is false.
18773 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
18774 (duplicate_and_interleave): New function.
18775 (vect_get_constant_vectors): Use gimple_build_vector for
18776 constant-length vectors and suitable variable-length constant
18777 vectors. Use duplicate_and_interleave for other variable-length
18778 vectors. Don't defer the update when inserting new statements.
18780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18781 Alan Hayward <alan.hayward@arm.com>
18782 David Sherwood <david.sherwood@arm.com>
18784 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
18785 min_profitable_iters doesn't go negative.
18787 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18788 Alan Hayward <alan.hayward@arm.com>
18789 David Sherwood <david.sherwood@arm.com>
18791 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
18792 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
18793 * optabs.def (vec_mask_load_lanes_optab): New optab.
18794 (vec_mask_store_lanes_optab): Likewise.
18795 * internal-fn.def (MASK_LOAD_LANES): New internal function.
18796 (MASK_STORE_LANES): Likewise.
18797 * internal-fn.c (mask_load_lanes_direct): New macro.
18798 (mask_store_lanes_direct): Likewise.
18799 (expand_mask_load_optab_fn): Handle masked operations.
18800 (expand_mask_load_lanes_optab_fn): New macro.
18801 (expand_mask_store_optab_fn): Handle masked operations.
18802 (expand_mask_store_lanes_optab_fn): New macro.
18803 (direct_mask_load_lanes_optab_supported_p): Likewise.
18804 (direct_mask_store_lanes_optab_supported_p): Likewise.
18805 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
18807 (vect_load_lanes_supported): Likewise.
18808 * tree-vect-data-refs.c (strip_conversion): New function.
18809 (can_group_stmts_p): Likewise.
18810 (vect_analyze_data_ref_accesses): Use it instead of checking
18811 for a pair of assignments.
18812 (vect_store_lanes_supported): Take a masked_p parameter.
18813 (vect_load_lanes_supported): Likewise.
18814 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
18815 vect_store_lanes_supported and vect_load_lanes_supported.
18816 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
18817 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
18818 parameter. Don't allow gaps for masked accesses.
18819 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
18820 and vect_load_lanes_supported.
18821 (get_load_store_type): Take a masked_p parameter and update
18822 call to get_group_load_store_type.
18823 (vectorizable_store): Update call to get_load_store_type.
18824 Handle IFN_MASK_STORE_LANES.
18825 (vectorizable_load): Update call to get_load_store_type.
18826 Handle IFN_MASK_LOAD_LANES.
18828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18829 Alan Hayward <alan.hayward@arm.com>
18830 David Sherwood <david.sherwood@arm.com>
18832 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
18834 * config/aarch64/aarch64-protos.h
18835 (aarch64_sve_struct_memory_operand_p): Declare.
18836 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
18837 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
18838 (VPRED, vpred): Handle SVE structure modes.
18839 * config/aarch64/constraints.md (Utx): New constraint.
18840 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
18841 (aarch64_sve_struct_nonimmediate_operand): New predicates.
18842 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
18843 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
18844 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
18845 structure modes. Split into pieces after RA.
18846 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
18847 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
18849 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
18850 SVE structure modes.
18851 (aarch64_classify_address): Likewise.
18852 (sizetochar): Move earlier in file.
18853 (aarch64_print_operand): Handle SVE register lists.
18854 (aarch64_array_mode): New function.
18855 (aarch64_sve_struct_memory_operand_p): Likewise.
18856 (TARGET_ARRAY_MODE): Redefine.
18858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18859 Alan Hayward <alan.hayward@arm.com>
18860 David Sherwood <david.sherwood@arm.com>
18862 * target.def (array_mode): New target hook.
18863 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
18864 * doc/tm.texi: Regenerate.
18865 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
18866 * hooks.c (hook_optmode_mode_uhwi_none): New function.
18867 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
18868 targetm.array_mode.
18869 * stor-layout.c (mode_for_array): Likewise. Support polynomial
18872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18873 Alan Hayward <alan.hayward@arm.com>
18874 David Sherwood <david.sherwood@arm.com>
18876 * fold-const.c (fold_binary_loc): Check the argument types
18877 rather than the result type when testing for a vector operation.
18879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18881 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
18882 * doc/tm.texi: Regenerate.
18884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18885 Alan Hayward <alan.hayward@arm.com>
18886 David Sherwood <david.sherwood@arm.com>
18888 * doc/invoke.texi (-msve-vector-bits=): Document new option.
18889 (sve): Document new AArch64 extension.
18890 * doc/md.texi (w): Extend the description of the AArch64
18891 constraint to include SVE vectors.
18892 (Upl, Upa): Document new AArch64 predicate constraints.
18893 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
18895 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
18896 (msve-vector-bits=): New option.
18897 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
18898 SVE when these are disabled.
18899 (sve): New extension.
18900 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
18901 modes. Adjust their number of units based on aarch64_sve_vg.
18902 (MAX_BITSIZE_MODE_ANY_MODE): Define.
18903 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
18904 aarch64_addr_query_type.
18905 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
18906 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
18907 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
18908 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
18909 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
18910 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
18911 (aarch64_simd_imm_zero_p): Delete.
18912 (aarch64_check_zero_based_sve_index_immediate): Declare.
18913 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18914 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18915 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18916 (aarch64_sve_float_mul_immediate_p): Likewise.
18917 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18918 rather than an rtx.
18919 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
18920 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
18921 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
18922 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
18923 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
18924 (aarch64_regmode_natural_size): Likewise.
18925 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
18926 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
18928 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
18929 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
18930 for VG and the SVE predicate registers.
18931 (V_ALIASES): Add a "z"-prefixed alias.
18932 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
18933 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
18934 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
18935 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
18936 (REG_CLASS_NAMES): Add entries for them.
18937 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
18938 and the predicate registers.
18939 (aarch64_sve_vg): Declare.
18940 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
18941 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
18942 (REGMODE_NATURAL_SIZE): Define.
18943 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
18945 * config/aarch64/aarch64.c: Include cfgrtl.h.
18946 (simd_immediate_info): Add a constructor for series vectors,
18947 and an associated step field.
18948 (aarch64_sve_vg): New variable.
18949 (aarch64_dbx_register_number): Handle VG and the predicate registers.
18950 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
18951 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
18952 (VEC_ANY_DATA, VEC_STRUCT): New constants.
18953 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
18954 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
18955 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
18956 (aarch64_get_mask_mode): New functions.
18957 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
18958 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
18959 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
18960 predicate modes and predicate registers. Explicitly restrict
18961 GPRs to modes of 16 bytes or smaller. Only allow FP registers
18962 to store a vector mode if it is recognized by
18963 aarch64_classify_vector_mode.
18964 (aarch64_regmode_natural_size): New function.
18965 (aarch64_hard_regno_caller_save_mode): Return the original mode
18967 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
18968 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
18969 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
18970 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
18972 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
18973 does not overlap dest if the function is frame-related. Handle
18975 (aarch64_split_add_offset): New function.
18976 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
18977 them aarch64_add_offset.
18978 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
18979 and update call to aarch64_sub_sp.
18980 (aarch64_add_cfa_expression): New function.
18981 (aarch64_expand_prologue): Pass extra temporary registers to the
18982 functions above. Handle the case in which we need to emit new
18983 DW_CFA_expressions for registers that were originally saved
18984 relative to the stack pointer, but now have to be expressed
18985 relative to the frame pointer.
18986 (aarch64_output_mi_thunk): Pass extra temporary registers to the
18988 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
18989 IP0 and IP1 values for SVE frames.
18990 (aarch64_expand_vec_series): New function.
18991 (aarch64_expand_sve_widened_duplicate): Likewise.
18992 (aarch64_expand_sve_const_vector): Likewise.
18993 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
18994 Handle SVE constants. Use emit_move_insn to move a force_const_mem
18995 into the register, rather than emitting a SET directly.
18996 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
18997 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
18998 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
18999 (offset_9bit_signed_scaled_p): New functions.
19000 (aarch64_replicate_bitmask_imm): New function.
19001 (aarch64_bitmask_imm): Use it.
19002 (aarch64_cannot_force_const_mem): Reject expressions involving
19003 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
19004 (aarch64_classify_index): Handle SVE indices, by requiring
19005 a plain register index with a scale that matches the element size.
19006 (aarch64_classify_address): Handle SVE addresses. Assert that
19007 the mode of the address is VOIDmode or an integer mode.
19008 Update call to aarch64_classify_symbol.
19009 (aarch64_classify_symbolic_expression): Update call to
19010 aarch64_classify_symbol.
19011 (aarch64_const_vec_all_in_range_p): New function.
19012 (aarch64_print_vector_float_operand): Likewise.
19013 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
19014 "vN" for FP registers with SVE modes. Handle (const ...) vectors
19015 and the FP immediates 1.0 and 0.5.
19016 (aarch64_print_address_internal): Handle SVE addresses.
19017 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
19018 (aarch64_regno_regclass): Handle predicate registers.
19019 (aarch64_secondary_reload): Handle big-endian reloads of SVE
19021 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
19022 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
19023 (aarch64_convert_sve_vector_bits): New function.
19024 (aarch64_override_options): Use it to handle -msve-vector-bits=.
19025 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
19026 rather than an rtx.
19027 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
19028 Handle SVE vector and predicate modes. Accept VL-based constants
19029 that need only one temporary register, and VL offsets that require
19030 no temporary registers.
19031 (aarch64_conditional_register_usage): Mark the predicate registers
19032 as fixed if SVE isn't available.
19033 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
19034 Return true for SVE vector and predicate modes.
19035 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
19036 rather than an unsigned int. Handle SVE modes.
19037 (aarch64_preferred_simd_mode): Update call accordingly. Handle
19039 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
19041 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
19042 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
19043 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
19044 (aarch64_sve_float_mul_immediate_p): New functions.
19045 (aarch64_sve_valid_immediate): New function.
19046 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
19047 Explicitly reject structure modes. Check for INDEX constants.
19048 Handle PTRUE and PFALSE constants.
19049 (aarch64_check_zero_based_sve_index_immediate): New function.
19050 (aarch64_simd_imm_zero_p): Delete.
19051 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
19052 vector modes. Accept constants in the range of CNT[BHWD].
19053 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
19054 ask for an Advanced SIMD mode.
19055 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
19056 (aarch64_simd_vector_alignment): Handle SVE predicates.
19057 (aarch64_vectorize_preferred_vector_alignment): New function.
19058 (aarch64_simd_vector_alignment_reachable): Use it instead of
19060 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
19061 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
19063 (MAX_VECT_LEN): Delete.
19064 (expand_vec_perm_d): Add a vec_flags field.
19065 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
19066 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
19067 (aarch64_evpc_ext): Don't apply a big-endian lane correction
19069 (aarch64_evpc_rev): Rename to...
19070 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
19071 (aarch64_evpc_rev_global): New function.
19072 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
19073 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
19075 (aarch64_evpc_sve_tbl): New function.
19076 (aarch64_expand_vec_perm_const_1): Update after rename of
19077 aarch64_evpc_rev. Handle SVE permutes too, trying
19078 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
19079 than aarch64_evpc_tbl.
19080 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
19081 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
19082 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
19083 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
19084 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
19085 (aarch64_expand_sve_vcond): New functions.
19086 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
19087 of aarch64_vector_mode_p.
19088 (aarch64_dwarf_poly_indeterminate_value): New function.
19089 (aarch64_compute_pressure_classes): Likewise.
19090 (aarch64_can_change_mode_class): Likewise.
19091 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
19092 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
19093 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
19094 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
19095 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
19096 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
19097 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
19098 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
19100 (Dn, Dl, Dr): Accept const as well as const_vector.
19101 (Dz): Likewise. Compare against CONST0_RTX.
19102 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
19103 of "vector" where appropriate.
19104 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
19105 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
19106 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
19107 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
19108 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
19109 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
19110 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
19111 (v_int_equiv): Extend to SVE modes.
19112 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
19114 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
19115 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
19116 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
19117 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
19118 (SVE_COND_FP_CMP): New int iterators.
19119 (perm_hilo): Handle the new unpack unspecs.
19120 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
19122 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
19123 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
19124 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
19125 (aarch64_equality_operator, aarch64_constant_vector_operand)
19126 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
19127 (aarch64_sve_nonimmediate_operand): Likewise.
19128 (aarch64_sve_general_operand): Likewise.
19129 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
19130 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
19131 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
19132 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
19133 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
19134 (aarch64_sve_float_arith_immediate): Likewise.
19135 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
19136 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
19137 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
19138 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
19139 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
19140 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
19141 (aarch64_sve_float_arith_operand): Likewise.
19142 (aarch64_sve_float_arith_with_sub_operand): Likewise.
19143 (aarch64_sve_float_mul_operand): Likewise.
19144 (aarch64_sve_vec_perm_operand): Likewise.
19145 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
19146 (aarch64_mov_operand): Accept const_poly_int and const_vector.
19147 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
19148 as well as const_vector.
19149 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
19150 in file. Use CONST0_RTX and CONSTM1_RTX.
19151 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
19152 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
19153 Use aarch64_simd_imm_zero.
19154 * config/aarch64/aarch64-sve.md: New file.
19155 * config/aarch64/aarch64.md: Include it.
19156 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
19157 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
19158 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
19159 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
19160 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
19161 (sve): New attribute.
19162 (enabled): Disable instructions with the sve attribute unless
19164 (movqi, movhi): Pass CONST_POLY_INT operaneds through
19165 aarch64_expand_mov_immediate.
19166 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
19167 CNT[BHSD] immediates.
19168 (movti): Split CONST_POLY_INT moves into two halves.
19169 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
19170 Split additions that need a temporary here if the destination
19171 is the stack pointer.
19172 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
19173 (*add<mode>3_poly_1): New instruction.
19174 (set_clobber_cc): New expander.
19176 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19178 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
19179 parameter and use it instead of GET_MODE_SIZE (innermode). Use
19180 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
19181 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
19182 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
19183 Change innermode from fixed_mode_size to machine_mode.
19184 (simplify_subreg): Update call accordingly. Handle a constant-sized
19185 subreg of a variable-length CONST_VECTOR.
19187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19188 Alan Hayward <alan.hayward@arm.com>
19189 David Sherwood <david.sherwood@arm.com>
19191 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
19192 (add_offset_to_base): New function, split out from...
19193 (create_mem_ref): ...here. When handling a scale other than 1,
19194 check first whether the address is valid without the offset.
19195 Add it into the base if so, leaving the index and scale as-is.
19197 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19200 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
19201 fold_for_warn before checking if arg2 is INTEGER_CST.
19203 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
19205 * config/rs6000/predicates.md (load_multiple_operation): Delete.
19206 (store_multiple_operation): Delete.
19207 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
19208 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
19209 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
19210 guarded by TARGET_STRING.
19211 (rs6000_output_load_multiple): Delete.
19212 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
19213 OPTION_MASK_STRING / TARGET_STRING handling.
19214 (print_operand) <'N', 'O'>: Add comment that these are unused now.
19215 (const rs6000_opt_masks) <"string">: Change mask to 0.
19216 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
19217 (MASK_STRING): Delete.
19218 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
19220 (load_multiple): Delete.
19227 (store_multiple): Delete.
19234 (movmemsi_8reg): Delete.
19235 (corresponding unnamed define_insn): Delete.
19236 (movmemsi_6reg): Delete.
19237 (corresponding unnamed define_insn): Delete.
19238 (movmemsi_4reg): Delete.
19239 (corresponding unnamed define_insn): Delete.
19240 (movmemsi_2reg): Delete.
19241 (corresponding unnamed define_insn): Delete.
19242 (movmemsi_1reg): Delete.
19243 (corresponding unnamed define_insn): Delete.
19244 * config/rs6000/rs6000.opt (mno-string): New.
19245 (mstring): Replace by deprecation warning stub.
19246 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
19248 2018-01-12 Jakub Jelinek <jakub@redhat.com>
19250 * regrename.c (regrename_do_replace): If replacing the same
19251 reg multiple times, try to reuse last created gen_raw_REG.
19254 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
19255 main to workaround a bug in GDB.
19257 2018-01-12 Tom de Vries <tom@codesourcery.com>
19260 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
19262 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
19264 PR rtl-optimization/80481
19265 * ira-color.c (get_cap_member): New function.
19266 (allocnos_conflict_by_live_ranges_p): Use it.
19267 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
19268 (setup_slot_coalesced_allocno_live_ranges): Ditto.
19270 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
19273 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
19274 (*saddl_se_1): Ditto.
19275 (*ssubsi_1): Ditto.
19276 (*ssubl_se_1): Ditto.
19278 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19280 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
19281 rather than wi::to_widest for DR_INITs.
19282 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
19283 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
19284 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
19286 (vect_analyze_group_access_1): Note that here.
19288 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19290 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
19291 polynomial type sizes.
19293 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
19295 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
19296 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
19297 (gimple_add_tmp_var): Likewise.
19299 2018-01-12 Martin Liska <mliska@suse.cz>
19301 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
19302 (gimple_alloc_sizes): Likewise.
19303 (dump_gimple_statistics): Use PRIu64 in printf format.
19304 * gimple.h: Change uint64_t to int.
19306 2018-01-12 Martin Liska <mliska@suse.cz>
19308 * tree-core.h: Use uint64_t instead of int.
19309 * tree.c (tree_node_counts): Likewise.
19310 (tree_node_sizes): Likewise.
19311 (dump_tree_statistics): Use PRIu64 in printf format.
19313 2018-01-12 Martin Liska <mliska@suse.cz>
19315 * Makefile.in: As qsort_chk is implemented in vec.c, add
19316 vec.o to linkage of gencfn-macros.
19317 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
19318 passing the info to record_node_allocation_statistics.
19319 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
19321 * ggc-common.c (struct ggc_usage): Add operator== and use
19322 it in operator< and compare function.
19323 * mem-stats.h (struct mem_usage): Likewise.
19324 * vec.c (struct vec_usage): Remove operator< and compare
19325 function. Can be simply inherited.
19327 2018-01-12 Martin Jambor <mjambor@suse.cz>
19330 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
19331 * tree-ssa-math-opts.c: Include domwalk.h.
19332 (convert_mult_to_fma_1): New function.
19333 (fma_transformation_info): New type.
19334 (fma_deferring_state): Likewise.
19335 (cancel_fma_deferring): New function.
19336 (result_of_phi): Likewise.
19337 (last_fma_candidate_feeds_initial_phi): Likewise.
19338 (convert_mult_to_fma): Added deferring logic, split actual
19339 transformation to convert_mult_to_fma_1.
19340 (math_opts_dom_walker): New type.
19341 (math_opts_dom_walker::after_dom_children): New method, body moved
19342 here from pass_optimize_widening_mul::execute, added deferring logic
19344 (pass_optimize_widening_mul::execute): Moved most of code to
19345 math_opts_dom_walker::after_dom_children.
19346 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
19347 * config/i386/i386.c (ix86_option_override_internal): Added
19348 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
19350 2018-01-12 Richard Biener <rguenther@suse.de>
19353 * dwarf2out.c (gen_variable_die): Do not reset old_die for
19354 inline instance vars.
19356 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
19359 * config/rx/rx.c (rx_is_restricted_memory_address):
19360 Handle SUBREG case.
19362 2018-01-12 Richard Biener <rguenther@suse.de>
19364 PR tree-optimization/80846
19365 * target.def (split_reduction): New target hook.
19366 * targhooks.c (default_split_reduction): New function.
19367 * targhooks.h (default_split_reduction): Declare.
19368 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
19369 target requests first reduce vectors by combining low and high
19371 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
19372 (get_vectype_for_scalar_type_and_size): Export.
19373 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
19374 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
19375 * doc/tm.texi: Regenerate.
19376 * config/i386/i386.c (ix86_split_reduction): Implement
19377 TARGET_VECTORIZE_SPLIT_REDUCTION.
19379 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19382 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
19383 in PIC mode except for TARGET_VXWORKS_RTP.
19384 * config/sparc/sparc.c: Include cfgrtl.h.
19385 (TARGET_INIT_PIC_REG): Define.
19386 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
19387 (sparc_pic_register_p): New predicate.
19388 (sparc_legitimate_address_p): Use it.
19389 (sparc_legitimize_pic_address): Likewise.
19390 (sparc_delegitimize_address): Likewise.
19391 (sparc_mode_dependent_address_p): Likewise.
19392 (gen_load_pcrel_sym): Remove 4th parameter.
19393 (load_got_register): Adjust call to above. Remove obsolete stuff.
19394 (sparc_expand_prologue): Do not call load_got_register here.
19395 (sparc_flat_expand_prologue): Likewise.
19396 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
19397 (sparc_use_pseudo_pic_reg): New function.
19398 (sparc_init_pic_reg): Likewise.
19399 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19400 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
19402 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
19404 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
19405 Add item for branch_cost.
19407 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19409 PR rtl-optimization/83565
19410 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
19411 not extend the result to a larger mode for rotate operations.
19412 (num_sign_bit_copies1): Likewise.
19414 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19417 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
19419 Use values-Xc.o for -pedantic.
19420 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
19422 2018-01-12 Martin Liska <mliska@suse.cz>
19425 * ipa-devirt.c (final_warning_record::grow_type_warnings):
19427 (possible_polymorphic_call_targets): Use it.
19428 (ipa_devirt): Likewise.
19430 2018-01-12 Martin Liska <mliska@suse.cz>
19432 * profile-count.h (enum profile_quality): Use 0 as invalid
19433 enum value of profile_quality.
19435 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
19437 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
19438 -mext-string options.
19440 2018-01-12 Richard Biener <rguenther@suse.de>
19442 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
19443 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
19444 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
19446 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
19448 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
19450 * configure.ac (--with-long-double-format): Add support for the
19451 configuration option to change the default long double format on
19453 * config.gcc (powerpc*-linux*-*): Likewise.
19454 * configure: Regenerate.
19455 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
19456 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
19457 used without modification.
19459 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19461 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
19462 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
19463 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
19464 MISC_BUILTIN_SPEC_BARRIER.
19465 (rs6000_init_builtins): Likewise.
19466 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
19468 (speculation_barrier): New define_insn.
19469 * doc/extend.texi: Document __builtin_speculation_barrier.
19471 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19474 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
19475 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
19476 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
19478 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
19479 integral modes instead of "ss" and "sd".
19480 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
19481 vectors with 32-bit and 64-bit elements.
19482 (vecdupssescalarmodesuffix): New mode attribute.
19483 (vec_dup<mode>): Use it.
19485 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
19488 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
19489 frame if argument is passed on stack.
19491 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19494 * ree.c (combine_reaching_defs): Optimize also
19495 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
19496 reg2=any_extend(exp); reg1=reg2;, formatting fix.
19498 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19500 PR middle-end/83189
19501 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
19503 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19505 PR middle-end/83718
19506 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
19507 after they are computed.
19509 2018-01-11 Bin Cheng <bin.cheng@arm.com>
19511 PR tree-optimization/83695
19512 * gimple-loop-linterchange.cc
19513 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
19514 reset cached scev information after interchange.
19515 (pass_linterchange::execute): Remove call to scev_reset_htab.
19517 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19519 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
19520 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
19521 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
19522 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
19523 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
19524 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
19525 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
19526 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
19527 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
19528 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
19529 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
19530 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
19531 (V_lane_reg): Likewise.
19532 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
19534 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
19535 (vfmal_lane_low<mode>_intrinsic,
19536 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
19537 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
19538 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
19539 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
19540 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
19541 vfmsl_lane_high<mode>_intrinsic): New define_insns.
19543 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19545 * config/arm/arm-cpus.in (fp16fml): New feature.
19546 (ALL_SIMD): Add fp16fml.
19547 (armv8.2-a): Add fp16fml as an option.
19548 (armv8.3-a): Likewise.
19549 (armv8.4-a): Add fp16fml as part of fp16.
19550 * config/arm/arm.h (TARGET_FP16FML): Define.
19551 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
19553 * config/arm/arm-modes.def (V2HF): Define.
19554 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
19555 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
19556 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
19557 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
19558 vfmsl_low, vfmsl_high): New set of builtins.
19559 * config/arm/iterators.md (PLUSMINUS): New code iterator.
19560 (vfml_op): New code attribute.
19561 (VFMLHALVES): New int iterator.
19562 (VFML, VFMLSEL): New mode attributes.
19563 (V_reg): Define mapping for V2HF.
19564 (V_hi, V_lo): New mode attributes.
19565 (VF_constraint): Likewise.
19566 (vfml_half, vfml_half_selector): New int attributes.
19567 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
19569 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
19570 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
19572 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
19573 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
19574 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
19575 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
19577 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
19578 Document new effective target and option set.
19580 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19582 * config/arm/arm-cpus.in (armv8_4): New feature.
19583 (ARMv8_4a): New fgroup.
19584 (armv8.4-a): New arch.
19585 * config/arm/arm-tables.opt: Regenerate.
19586 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
19587 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
19588 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
19589 Add matching rules for -march=armv8.4-a and extensions.
19590 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
19592 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
19595 * config/rx/rx.md (BW): New mode attribute.
19596 (sync_lock_test_and_setsi): Add mode suffix to insn output.
19598 2018-01-11 Richard Biener <rguenther@suse.de>
19600 PR tree-optimization/83435
19601 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
19602 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
19603 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
19605 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19606 Alan Hayward <alan.hayward@arm.com>
19607 David Sherwood <david.sherwood@arm.com>
19609 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
19611 (aarch64_classify_address): Initialize it. Track polynomial offsets.
19612 (aarch64_print_address_internal): Use it to check for a zero offset.
19614 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19615 Alan Hayward <alan.hayward@arm.com>
19616 David Sherwood <david.sherwood@arm.com>
19618 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
19619 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
19620 Return a poly_int64 rather than a HOST_WIDE_INT.
19621 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
19622 rather than a HOST_WIDE_INT.
19623 * config/aarch64/aarch64.h (aarch64_frame): Protect with
19624 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
19625 hard_fp_offset, frame_size, initial_adjust, callee_offset and
19626 final_offset from HOST_WIDE_INT to poly_int64.
19627 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
19628 to_constant when getting the number of units in an Advanced SIMD
19630 (aarch64_builtin_vectorized_function): Check for a constant number
19632 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
19634 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
19635 attribute instead of GET_MODE_NUNITS.
19636 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
19637 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
19638 GET_MODE_SIZE for fixed-size registers.
19639 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
19640 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
19641 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
19642 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
19643 (aarch64_print_operand, aarch64_print_address_internal)
19644 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
19645 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
19646 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
19647 Handle polynomial GET_MODE_SIZE.
19648 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
19649 wider than SImode without modification.
19650 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
19651 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
19652 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
19653 passing and returning SVE modes.
19654 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
19655 rather than GEN_INT.
19656 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
19657 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
19658 (aarch64_allocate_and_probe_stack_space): Likewise.
19659 (aarch64_layout_frame): Cope with polynomial offsets.
19660 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
19661 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
19662 polynomial offsets.
19663 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
19664 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
19665 poly_int64 rather than a HOST_WIDE_INT.
19666 (aarch64_get_separate_components, aarch64_process_components)
19667 (aarch64_expand_prologue, aarch64_expand_epilogue)
19668 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
19669 (aarch64_anchor_offset): New function, split out from...
19670 (aarch64_legitimize_address): ...here.
19671 (aarch64_builtin_vectorization_cost): Handle polynomial
19672 TYPE_VECTOR_SUBPARTS.
19673 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
19675 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
19676 number of elements from the PARALLEL rather than the mode.
19677 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
19678 rather than GET_MODE_BITSIZE.
19679 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
19680 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
19681 (aarch64_expand_vec_perm_const_1): Handle polynomial
19682 d->perm.length () and d->perm elements.
19683 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
19684 Apply to_constant to d->perm elements.
19685 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
19686 polynomial CONST_VECTOR_NUNITS.
19687 (aarch64_move_pointer): Take amount as a poly_int64 rather
19689 (aarch64_progress_pointer): Avoid temporary variable.
19690 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
19691 the mode attribute instead of GET_MODE.
19693 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19694 Alan Hayward <alan.hayward@arm.com>
19695 David Sherwood <david.sherwood@arm.com>
19697 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
19698 x exists before using it.
19699 (aarch64_add_constant_internal): Rename to...
19700 (aarch64_add_offset_1): ...this. Replace regnum with separate
19701 src and dest rtxes. Handle the case in which they're different,
19702 including when the offset is zero. Replace scratchreg with an rtx.
19703 Use 2 additions if there is no spare register into which we can
19704 move a 16-bit constant.
19705 (aarch64_add_constant): Delete.
19706 (aarch64_add_offset): Replace reg with separate src and dest
19707 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
19708 Use aarch64_add_offset_1.
19709 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
19710 an rtx rather than an int. Take the delta as a poly_int64
19711 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
19712 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
19713 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
19714 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
19715 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
19716 and aarch64_add_sp.
19717 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
19718 aarch64_add_constant.
19720 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19722 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
19723 Use scalar_float_mode.
19725 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19727 * config/aarch64/aarch64-simd.md
19728 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
19729 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
19730 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
19731 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
19732 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
19733 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
19734 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
19735 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
19736 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
19737 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
19739 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19742 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
19743 targ_options->x_arm_arch_string is non NULL.
19745 2018-01-11 Tamar Christina <tamar.christina@arm.com>
19747 * config/aarch64/aarch64.h
19748 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
19750 2018-01-11 Sudakshina Das <sudi.das@arm.com>
19753 * expmed.c (emit_store_flag_force): Swap if const op0
19754 and change VOIDmode to mode of op0.
19756 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19758 PR rtl-optimization/83761
19759 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
19760 than bytes to mode_for_size.
19762 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19764 PR middle-end/83189
19765 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
19766 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
19769 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19771 PR middle-end/83575
19772 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
19773 when in layout mode.
19774 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
19775 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
19778 2018-01-10 Michael Collison <michael.collison@arm.com>
19780 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
19781 * config/aarch64/aarch64-option-extension.def: Add
19782 AARCH64_OPT_EXTENSION of 'fp16fml'.
19783 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19784 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
19785 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
19786 * config/aarch64/constraints.md (Ui7): New constraint.
19787 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
19788 (VFMLA_SEL_W): Ditto.
19791 (VFMLA16_LOW): New int iterator.
19792 (VFMLA16_HIGH): Ditto.
19793 (UNSPEC_FMLAL): New unspec.
19794 (UNSPEC_FMLSL): Ditto.
19795 (UNSPEC_FMLAL2): Ditto.
19796 (UNSPEC_FMLSL2): Ditto.
19797 (f16mac): New code attribute.
19798 * config/aarch64/aarch64-simd-builtins.def
19799 (aarch64_fmlal_lowv2sf): Ditto.
19800 (aarch64_fmlsl_lowv2sf): Ditto.
19801 (aarch64_fmlalq_lowv4sf): Ditto.
19802 (aarch64_fmlslq_lowv4sf): Ditto.
19803 (aarch64_fmlal_highv2sf): Ditto.
19804 (aarch64_fmlsl_highv2sf): Ditto.
19805 (aarch64_fmlalq_highv4sf): Ditto.
19806 (aarch64_fmlslq_highv4sf): Ditto.
19807 (aarch64_fmlal_lane_lowv2sf): Ditto.
19808 (aarch64_fmlsl_lane_lowv2sf): Ditto.
19809 (aarch64_fmlal_laneq_lowv2sf): Ditto.
19810 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
19811 (aarch64_fmlalq_lane_lowv4sf): Ditto.
19812 (aarch64_fmlsl_lane_lowv4sf): Ditto.
19813 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
19814 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
19815 (aarch64_fmlal_lane_highv2sf): Ditto.
19816 (aarch64_fmlsl_lane_highv2sf): Ditto.
19817 (aarch64_fmlal_laneq_highv2sf): Ditto.
19818 (aarch64_fmlsl_laneq_highv2sf): Ditto.
19819 (aarch64_fmlalq_lane_highv4sf): Ditto.
19820 (aarch64_fmlsl_lane_highv4sf): Ditto.
19821 (aarch64_fmlalq_laneq_highv4sf): Ditto.
19822 (aarch64_fmlsl_laneq_highv4sf): Ditto.
19823 * config/aarch64/aarch64-simd.md:
19824 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
19825 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19826 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
19827 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19828 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
19829 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
19830 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
19831 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
19832 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
19833 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
19834 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
19835 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
19836 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
19837 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
19838 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
19839 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
19840 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
19841 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
19842 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
19843 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
19844 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
19845 (vfmlsl_low_u32): Ditto.
19846 (vfmlalq_low_u32): Ditto.
19847 (vfmlslq_low_u32): Ditto.
19848 (vfmlal_high_u32): Ditto.
19849 (vfmlsl_high_u32): Ditto.
19850 (vfmlalq_high_u32): Ditto.
19851 (vfmlslq_high_u32): Ditto.
19852 (vfmlal_lane_low_u32): Ditto.
19853 (vfmlsl_lane_low_u32): Ditto.
19854 (vfmlal_laneq_low_u32): Ditto.
19855 (vfmlsl_laneq_low_u32): Ditto.
19856 (vfmlalq_lane_low_u32): Ditto.
19857 (vfmlslq_lane_low_u32): Ditto.
19858 (vfmlalq_laneq_low_u32): Ditto.
19859 (vfmlslq_laneq_low_u32): Ditto.
19860 (vfmlal_lane_high_u32): Ditto.
19861 (vfmlsl_lane_high_u32): Ditto.
19862 (vfmlal_laneq_high_u32): Ditto.
19863 (vfmlsl_laneq_high_u32): Ditto.
19864 (vfmlalq_lane_high_u32): Ditto.
19865 (vfmlslq_lane_high_u32): Ditto.
19866 (vfmlalq_laneq_high_u32): Ditto.
19867 (vfmlslq_laneq_high_u32): Ditto.
19868 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
19869 (AARCH64_FL_FOR_ARCH8_4): New.
19870 (AARCH64_ISA_F16FML): New ISA flag.
19871 (TARGET_F16FML): New feature flag for fp16fml.
19872 (doc/invoke.texi): Document new fp16fml option.
19874 2018-01-10 Michael Collison <michael.collison@arm.com>
19876 * config/aarch64/aarch64-builtins.c:
19877 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
19878 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19879 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
19880 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
19881 (AARCH64_ISA_SHA3): New ISA flag.
19882 (TARGET_SHA3): New feature flag for sha3.
19883 * config/aarch64/iterators.md (sha512_op): New int attribute.
19884 (CRYPTO_SHA512): New int iterator.
19885 (UNSPEC_SHA512H): New unspec.
19886 (UNSPEC_SHA512H2): Ditto.
19887 (UNSPEC_SHA512SU0): Ditto.
19888 (UNSPEC_SHA512SU1): Ditto.
19889 * config/aarch64/aarch64-simd-builtins.def
19890 (aarch64_crypto_sha512hqv2di): New builtin.
19891 (aarch64_crypto_sha512h2qv2di): Ditto.
19892 (aarch64_crypto_sha512su0qv2di): Ditto.
19893 (aarch64_crypto_sha512su1qv2di): Ditto.
19894 (aarch64_eor3qv8hi): Ditto.
19895 (aarch64_rax1qv2di): Ditto.
19896 (aarch64_xarqv2di): Ditto.
19897 (aarch64_bcaxqv8hi): Ditto.
19898 * config/aarch64/aarch64-simd.md:
19899 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
19900 (aarch64_crypto_sha512su0qv2di): Ditto.
19901 (aarch64_crypto_sha512su1qv2di): Ditto.
19902 (aarch64_eor3qv8hi): Ditto.
19903 (aarch64_rax1qv2di): Ditto.
19904 (aarch64_xarqv2di): Ditto.
19905 (aarch64_bcaxqv8hi): Ditto.
19906 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
19907 (vsha512h2q_u64): Ditto.
19908 (vsha512su0q_u64): Ditto.
19909 (vsha512su1q_u64): Ditto.
19910 (veor3q_u16): Ditto.
19911 (vrax1q_u64): Ditto.
19912 (vxarq_u64): Ditto.
19913 (vbcaxq_u16): Ditto.
19914 * config/arm/types.md (crypto_sha512): New type attribute.
19915 (crypto_sha3): Ditto.
19916 (doc/invoke.texi): Document new sha3 option.
19918 2018-01-10 Michael Collison <michael.collison@arm.com>
19920 * config/aarch64/aarch64-builtins.c:
19921 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
19922 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19923 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
19924 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
19925 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
19926 (AARCH64_ISA_SM4): New ISA flag.
19927 (TARGET_SM4): New feature flag for sm4.
19928 * config/aarch64/aarch64-simd-builtins.def
19929 (aarch64_sm3ss1qv4si): Ditto.
19930 (aarch64_sm3tt1aq4si): Ditto.
19931 (aarch64_sm3tt1bq4si): Ditto.
19932 (aarch64_sm3tt2aq4si): Ditto.
19933 (aarch64_sm3tt2bq4si): Ditto.
19934 (aarch64_sm3partw1qv4si): Ditto.
19935 (aarch64_sm3partw2qv4si): Ditto.
19936 (aarch64_sm4eqv4si): Ditto.
19937 (aarch64_sm4ekeyqv4si): Ditto.
19938 * config/aarch64/aarch64-simd.md:
19939 (aarch64_sm3ss1qv4si): Ditto.
19940 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
19941 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
19942 (aarch64_sm4eqv4si): Ditto.
19943 (aarch64_sm4ekeyqv4si): Ditto.
19944 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
19945 (sm3part_op): Ditto.
19946 (CRYPTO_SM3TT): Ditto.
19947 (CRYPTO_SM3PART): Ditto.
19948 (UNSPEC_SM3SS1): New unspec.
19949 (UNSPEC_SM3TT1A): Ditto.
19950 (UNSPEC_SM3TT1B): Ditto.
19951 (UNSPEC_SM3TT2A): Ditto.
19952 (UNSPEC_SM3TT2B): Ditto.
19953 (UNSPEC_SM3PARTW1): Ditto.
19954 (UNSPEC_SM3PARTW2): Ditto.
19955 (UNSPEC_SM4E): Ditto.
19956 (UNSPEC_SM4EKEY): Ditto.
19957 * config/aarch64/constraints.md (Ui2): New constraint.
19958 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
19959 * config/arm/types.md (crypto_sm3): New type attribute.
19960 (crypto_sm4): Ditto.
19961 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
19962 (vsm3tt1aq_u32): Ditto.
19963 (vsm3tt1bq_u32): Ditto.
19964 (vsm3tt2aq_u32): Ditto.
19965 (vsm3tt2bq_u32): Ditto.
19966 (vsm3partw1q_u32): Ditto.
19967 (vsm3partw2q_u32): Ditto.
19968 (vsm4eq_u32): Ditto.
19969 (vsm4ekeyq_u32): Ditto.
19970 (doc/invoke.texi): Document new sm4 option.
19972 2018-01-10 Michael Collison <michael.collison@arm.com>
19974 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
19975 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
19976 (AARCH64_FL_FOR_ARCH8_4): New.
19977 (AARCH64_FL_V8_4): New flag.
19978 (doc/invoke.texi): Document new armv8.4-a option.
19980 2018-01-10 Michael Collison <michael.collison@arm.com>
19982 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19983 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
19984 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
19985 * config/aarch64/aarch64-option-extension.def: Add
19986 AARCH64_OPT_EXTENSION of 'sha2'.
19987 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
19988 (crypto): Disable sha2 and aes if crypto disabled.
19989 (crypto): Enable aes and sha2 if enabled.
19990 (simd): Disable sha2 and aes if simd disabled.
19991 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
19993 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
19994 (TARGET_SHA2): New feature flag for sha2.
19995 (TARGET_AES): New feature flag for aes.
19996 * config/aarch64/aarch64-simd.md:
19997 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
19998 conditional on TARGET_AES.
19999 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
20000 (aarch64_crypto_sha1hsi): Make pattern conditional
20002 (aarch64_crypto_sha1hv4si): Ditto.
20003 (aarch64_be_crypto_sha1hv4si): Ditto.
20004 (aarch64_crypto_sha1su1v4si): Ditto.
20005 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
20006 (aarch64_crypto_sha1su0v4si): Ditto.
20007 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
20008 (aarch64_crypto_sha256su0v4si): Ditto.
20009 (aarch64_crypto_sha256su1v4si): Ditto.
20010 (doc/invoke.texi): Document new aes and sha2 options.
20012 2018-01-10 Martin Sebor <msebor@redhat.com>
20014 PR tree-optimization/83781
20015 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
20018 2018-01-11 Martin Sebor <msebor@gmail.com>
20019 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20021 PR tree-optimization/83501
20022 PR tree-optimization/81703
20024 * tree-ssa-strlen.c (get_string_cst): Rename...
20025 (get_string_len): ...to this. Handle global constants.
20026 (handle_char_store): Adjust.
20028 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
20029 Jim Wilson <jimw@sifive.com>
20031 * config/riscv/riscv-protos.h (riscv_output_return): New.
20032 * config/riscv/riscv.c (struct machine_function): New naked_p field.
20033 (riscv_attribute_table, riscv_output_return),
20034 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
20035 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
20036 (riscv_compute_frame_info): Only compute frame->mask if not a naked
20038 (riscv_expand_prologue): Add early return for naked function.
20039 (riscv_expand_epilogue): Likewise.
20040 (riscv_function_ok_for_sibcall): Return false for naked function.
20041 (riscv_set_current_function): New.
20042 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
20043 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
20044 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
20045 * doc/extend.texi (RISC-V Function Attributes): New.
20047 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
20049 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
20050 check for 128-bit long double before checking TCmode.
20051 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
20052 128-bit long doubles before checking TFmode or TCmode.
20053 (FLOAT128_IBM_P): Likewise.
20055 2018-01-10 Martin Sebor <msebor@redhat.com>
20057 PR tree-optimization/83671
20058 * builtins.c (c_strlen): Unconditionally return zero for the empty
20060 Use -Warray-bounds for warnings.
20061 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
20062 for non-constant array indices with COMPONENT_REF, arrays of
20063 arrays, and pointers to arrays.
20064 (gimple_fold_builtin_strlen): Determine and set length range for
20065 non-constant character arrays.
20067 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
20069 PR middle-end/81897
20070 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
20073 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
20075 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
20077 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20080 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
20081 VECTOR_MEM_ALTIVEC_OR_VSX_P.
20082 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
20083 indexed_or_indirect_operand predicate.
20084 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
20085 (*vsx_le_perm_load_v8hi): Likewise.
20086 (*vsx_le_perm_load_v16qi): Likewise.
20087 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
20088 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
20089 (*vsx_le_perm_store_v8hi): Likewise.
20090 (*vsx_le_perm_store_v16qi): Likewise.
20091 (eight unnamed splitters): Likewise.
20093 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
20095 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
20096 * config/rs6000/emmintrin.h: Likewise.
20097 * config/rs6000/mmintrin.h: Likewise.
20098 * config/rs6000/xmmintrin.h: Likewise.
20100 2018-01-10 David Malcolm <dmalcolm@redhat.com>
20103 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
20105 * tree.c (tree_nop_conversion): Return true for location wrapper
20107 (maybe_wrap_with_location): New function.
20108 (selftest::check_strip_nops): New function.
20109 (selftest::test_location_wrappers): New function.
20110 (selftest::tree_c_tests): Call it.
20111 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
20112 (maybe_wrap_with_location): New decl.
20113 (EXPR_LOCATION_WRAPPER_P): New macro.
20114 (location_wrapper_p): New inline function.
20115 (tree_strip_any_location_wrapper): New inline function.
20117 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
20120 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
20121 stack_realign_offset for the largest alignment of stack slot
20123 (ix86_find_max_used_stack_alignment): New function.
20124 (ix86_finalize_stack_frame_flags): Use it. Set
20125 max_used_stack_alignment if we don't realign stack.
20126 * config/i386/i386.h (machine_function): Add
20127 max_used_stack_alignment.
20129 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
20131 * config/arm/arm.opt (-mbranch-cost): New option.
20132 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
20135 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
20138 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
20139 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
20141 2018-01-10 Richard Biener <rguenther@suse.de>
20144 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
20145 early out so it also covers the case where we have a non-NULL
20148 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20150 PR tree-optimization/83753
20151 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
20152 for non-strided grouped accesses if the number of elements is 1.
20154 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
20157 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
20158 * i386.h (TARGET_USE_GATHER): Define.
20159 * x86-tune.def (X86_TUNE_USE_GATHER): New.
20161 2018-01-10 Martin Liska <mliska@suse.cz>
20164 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
20165 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
20167 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
20168 CLEANUP_NO_PARTITIONING is not set.
20170 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
20172 * doc/rtl.texi: Remove documentation of (const ...) wrappers
20173 for vectors, as a partial revert of r254296.
20174 * rtl.h (const_vec_p): Delete.
20175 (const_vec_duplicate_p): Don't test for vector CONSTs.
20176 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
20177 * expmed.c (make_tree): Likewise.
20180 * common.md (E, F): Use CONSTANT_P instead of checking for
20182 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
20183 checking for CONST_VECTOR.
20185 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20187 PR middle-end/83575
20188 * predict.c (force_edge_cold): Handle in more sane way edges
20189 with no prediction.
20191 2018-01-09 Carl Love <cel@us.ibm.com>
20193 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
20195 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
20196 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
20197 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
20198 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
20199 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
20200 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
20201 * config/rs6000/rs6000-protos.h: Add extern defition for
20202 rs6000_generate_float2_double_code.
20203 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
20205 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
20206 (float2_v2df): Add define_expand.
20208 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
20211 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
20212 op_mode in the force_to_mode call.
20214 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20216 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
20217 instead of checking each element individually.
20218 (aarch64_evpc_uzp): Likewise.
20219 (aarch64_evpc_zip): Likewise.
20220 (aarch64_evpc_ext): Likewise.
20221 (aarch64_evpc_rev): Likewise.
20222 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
20223 instead of checking each element individually. Return true without
20225 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
20226 whether all selected elements come from the same input, instead of
20227 checking each element individually. Remove calls to gen_rtx_REG,
20228 start_sequence and end_sequence and instead assert that no rtl is
20231 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20233 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
20234 order of HIGH and CONST checks.
20236 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
20238 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
20239 if the destination isn't an SSA_NAME.
20241 2018-01-09 Richard Biener <rguenther@suse.de>
20243 PR tree-optimization/83668
20244 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
20246 (canonicalize_loop_form): ... here, renamed from ...
20247 (canonicalize_loop_closed_ssa_form): ... this and amended to
20248 swap successor edges for loop exit blocks to make us use
20249 the RPO order we need for initial schedule generation.
20251 2018-01-09 Joseph Myers <joseph@codesourcery.com>
20253 PR tree-optimization/64811
20254 * match.pd: When optimizing comparisons with Inf, avoid
20255 introducing or losing exceptions from comparisons with NaN.
20257 2018-01-09 Martin Liska <mliska@suse.cz>
20260 * asan.c (shadow_mem_size): Add gcc_assert.
20262 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
20264 Don't save registers in main().
20267 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
20268 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
20269 * config/avr/avr.c (avr_set_current_function): Don't error if
20270 naked, OS_task or OS_main are specified at the same time.
20271 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
20273 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
20275 * common/config/avr/avr-common.c (avr_option_optimization_table):
20276 Switch on -mmain-is-OS_task for optimizing compilations.
20278 2018-01-09 Richard Biener <rguenther@suse.de>
20280 PR tree-optimization/83572
20281 * graphite.c: Include cfganal.h.
20282 (graphite_transform_loops): Connect infinite loops to exit
20283 and remove fake edges at the end.
20285 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20287 * ipa-inline.c (edge_badness): Revert accidental checkin.
20289 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
20292 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
20293 symbols; not inline clones.
20295 2018-01-09 Jakub Jelinek <jakub@redhat.com>
20298 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
20299 hard registers. Formatting fixes.
20301 PR preprocessor/83722
20302 * gcc.c (try_generate_repro): Pass
20303 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
20304 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
20307 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
20308 Kito Cheng <kito.cheng@gmail.com>
20310 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
20311 (riscv_leaf_function_p): Delete.
20312 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
20314 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20316 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
20318 (do_ifelse): New function.
20319 (do_isel): New function.
20320 (do_sub3): New function.
20321 (do_add3): New function.
20322 (do_load_mask_compare): New function.
20323 (do_overlap_load_compare): New function.
20324 (expand_compare_loop): New function.
20325 (expand_block_compare): Call expand_compare_loop() when appropriate.
20326 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
20327 option description.
20328 (-mblock-compare-inline-loop-limit): New option.
20330 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20333 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
20334 Reverse order of second and third operands in first alternative.
20335 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
20336 of first and second elements in UNSPEC_VPERMR vector.
20337 (altivec_expand_vec_perm_le): Likewise.
20339 2018-01-08 Jeff Law <law@redhat.com>
20341 PR rtl-optimizatin/81308
20342 * tree-switch-conversion.c (cfg_altered): New file scoped static.
20343 (process_switch): If group_case_labels makes a change, then set
20345 (pass_convert_switch::execute): If a switch is converted, then
20346 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
20348 PR rtl-optimization/81308
20349 * recog.c (split_all_insns): Conditionally cleanup the CFG after
20352 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
20354 PR target/83663 - Revert r255946
20355 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
20356 generation for cases where splatting a value is not useful.
20357 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
20358 across a vec_duplicate and a paradoxical subreg forming a vector
20359 mode to a vec_concat.
20361 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20363 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
20364 -march=armv8.3-a variants.
20365 * config/arm/t-multilib: Likewise.
20366 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
20368 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20370 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
20372 (cceq_ior_compare_complement): Give it a name so I can use it, and
20373 change boolean_or_operator predicate to boolean_operator so it can
20374 be used to generate a crand.
20375 (eqne): New code iterator.
20376 (bd/bd_neg): New code_attrs.
20377 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
20378 a single define_insn.
20379 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
20380 decrement (bdnzt/bdnzf/bdzt/bdzf).
20381 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
20382 with the new names of the branch decrement patterns, and added the
20383 names of the branch decrement conditional patterns.
20385 2018-01-08 Richard Biener <rguenther@suse.de>
20387 PR tree-optimization/83563
20388 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
20391 2018-01-08 Richard Biener <rguenther@suse.de>
20393 PR middle-end/83713
20394 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
20396 2018-01-08 Richard Biener <rguenther@suse.de>
20398 PR tree-optimization/83685
20399 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
20400 references to abnormals.
20402 2018-01-08 Richard Biener <rguenther@suse.de>
20405 * dwarf2out.c (output_indirect_strings): Handle empty
20406 skeleton_debug_str_hash.
20407 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
20409 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20411 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
20412 (emit_store_direct): Likewise.
20413 (arc_trampoline_adjust_address): Likewise.
20414 (arc_asm_trampoline_template): New function.
20415 (arc_initialize_trampoline): Use asm_trampoline_template.
20416 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
20417 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
20418 * config/arc/arc.md (flush_icache): Delete pattern.
20420 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20422 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
20423 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
20426 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20429 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
20430 by not USED_FOR_TARGET.
20431 (make_pass_resolve_sw_modes): Likewise.
20433 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20435 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
20438 2018-01-08 Richard Biener <rguenther@suse.de>
20440 PR middle-end/83580
20441 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
20443 2018-01-08 Richard Biener <rguenther@suse.de>
20445 PR middle-end/83517
20446 * match.pd ((t * 2) / 2) -> t): Add missing :c.
20448 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
20450 PR middle-end/81897
20451 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
20452 basic blocks with a small number of successors.
20453 (convert_control_dep_chain_into_preds): Improve handling of
20455 (dump_predicates): Split apart into...
20456 (dump_pred_chain): ...here...
20457 (dump_pred_info): ...and here.
20458 (can_one_predicate_be_invalidated_p): Add debugging printfs.
20459 (can_chain_union_be_invalidated_p): Improve check for invalidation
20461 (uninit_uses_cannot_happen): Avoid unnecessary if
20462 convert_control_dep_chain_into_preds yielded nothing.
20464 2018-01-06 Martin Sebor <msebor@redhat.com>
20466 PR tree-optimization/83640
20467 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
20468 subtracting negative offset from size.
20469 (builtin_access::overlap): Adjust offset bounds of the access to fall
20470 within the size of the object if possible.
20472 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
20474 PR rtl-optimization/83699
20475 * expmed.c (extract_bit_field_1): Restrict the vector usage of
20476 extract_bit_field_as_subreg to cases in which the extracted
20477 value is also a vector.
20479 * lra-constraints.c (process_alt_operands): Test for the equivalence
20480 substitutions when detecting a possible reload cycle.
20482 2018-01-06 Jakub Jelinek <jakub@redhat.com>
20485 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
20486 by default if flag_selective_schedling{,2}. Formatting fixes.
20488 PR rtl-optimization/83682
20489 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
20490 if it has non-VECTOR_MODE element mode.
20491 (vec_duplicate_p): Likewise.
20493 PR middle-end/83694
20494 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
20495 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
20497 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20500 * config/i386/i386-builtin.def
20501 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
20502 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
20503 Require also OPTION_MASK_ISA_AVX512F in addition to
20504 OPTION_MASK_ISA_GFNI.
20505 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
20506 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
20507 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
20508 to OPTION_MASK_ISA_GFNI.
20509 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
20510 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
20511 OPTION_MASK_ISA_AVX512BW.
20512 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
20513 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
20514 addition to OPTION_MASK_ISA_GFNI.
20515 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
20516 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
20517 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
20518 to OPTION_MASK_ISA_GFNI.
20519 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
20520 a requirement for all ISAs rather than any of them with a few
20522 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
20524 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
20525 bitmasks to be enabled with 3 exceptions, instead of requiring any
20526 enabled ISA with lots of exceptions.
20527 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
20528 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
20529 Change avx512bw in isa attribute to avx512f.
20530 * config/i386/sgxintrin.h: Add license boilerplate.
20531 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
20532 to __AVX512F__ and __AVX512VL to __AVX512VL__.
20533 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
20534 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
20536 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
20537 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
20538 temporarily sse2 rather than sse if not enabled already.
20541 * config/i386/sse.md (VI248_VLBW): Rename to ...
20542 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
20543 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
20544 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
20545 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
20546 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
20547 mode iterator instead of VI248_VLBW.
20549 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
20551 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
20552 (record_modified): Skip clobbers; add debug output.
20553 (param_change_prob): Use sreal frequencies.
20555 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20557 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
20558 punt for user-aligned variables.
20560 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20562 * tree-chrec.c (chrec_contains_symbols): Return true for
20565 2018-01-05 Sudakshina Das <sudi.das@arm.com>
20568 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
20569 of (x|y) == x for BICS pattern.
20571 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20573 PR tree-optimization/83605
20574 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
20575 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
20578 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
20580 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
20581 * config/epiphany/rtems.h: New file.
20583 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20584 Uros Bizjak <ubizjak@gmail.com>
20587 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
20588 QIreg_operand instead of register_operand predicate.
20589 * config/i386/i386.c (ix86_rop_should_change_byte_p,
20590 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
20591 comments instead of -fmitigate[-_]rop.
20593 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20596 * cgraphunit.c (symbol_table::compile): Switch to text_section
20597 before calling assembly_start debug hook.
20598 * run-rtl-passes.c (run_rtl_passes): Likewise.
20601 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20603 * tree-vrp.c (extract_range_from_binary_expr_1): Check
20604 range_int_cst_p rather than !symbolic_range_p before calling
20605 extract_range_from_multiplicative_op_1.
20607 2018-01-04 Jeff Law <law@redhat.com>
20609 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
20610 redundant test in assertion.
20612 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20614 * doc/rtl.texi: Document machine_mode wrapper classes.
20616 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20618 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
20619 using tree_to_uhwi.
20621 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20623 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
20624 the VEC_PERM_EXPR fold to fail.
20626 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20629 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
20630 to switched_sections.
20632 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20635 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
20636 test for d.testing.
20638 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
20641 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
20642 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
20644 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20647 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
20648 is BLKmode and bitpos not zero or mode change is needed.
20650 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20653 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
20656 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
20659 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
20660 instead of MULT rtx. Update all corresponding splitters.
20661 (*saddl_se): Ditto.
20662 (*ssub<modesuffix>): Ditto.
20663 (*ssubl_se): Ditto.
20664 (*cmp_sadd_di): Update split patterns.
20665 (*cmp_sadd_si): Ditto.
20666 (*cmp_sadd_sidi): Ditto.
20667 (*cmp_ssub_di): Ditto.
20668 (*cmp_ssub_si): Ditto.
20669 (*cmp_ssub_sidi): Ditto.
20670 * config/alpha/predicates.md (const23_operand): New predicate.
20671 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
20672 Look for ASHIFT, not MULT inner operand.
20673 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
20675 2018-01-04 Martin Liska <mliska@suse.cz>
20677 PR gcov-profile/83669
20678 * gcov.c (output_intermediate_file): Add version to intermediate
20680 * doc/gcov.texi: Document new field 'version' in intermediate
20681 file format. Fix location of '-k' option of gcov command.
20683 2018-01-04 Martin Liska <mliska@suse.cz>
20686 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
20688 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20690 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
20692 2018-01-03 Martin Sebor <msebor@redhat.com>
20694 PR tree-optimization/83655
20695 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
20696 checking calls with invalid arguments.
20698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20700 * tree-vect-stmts.c (vect_get_store_rhs): New function.
20701 (vectorizable_mask_load_store): Delete.
20702 (vectorizable_call): Return false for masked loads and stores.
20703 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
20704 instead of gimple_assign_rhs1.
20705 (vectorizable_load): Handle IFN_MASK_LOAD.
20706 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
20708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20710 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
20712 (vectorizable_mask_load_store): ...here.
20713 (vectorizable_load): ...and here.
20715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20717 * tree-vect-stmts.c (vect_build_all_ones_mask)
20718 (vect_build_zero_merge_argument): New functions, split out from...
20719 (vectorizable_load): ...here.
20721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20723 * tree-vect-stmts.c (vect_check_store_rhs): New function,
20725 (vectorizable_mask_load_store): ...here.
20726 (vectorizable_store): ...and here.
20728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20730 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
20732 (vectorizable_mask_load_store): ...here.
20734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20736 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
20737 (vect_model_store_cost): Take a vec_load_store_type instead of a
20739 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
20740 (vect_model_store_cost): Take a vec_load_store_type instead of a
20742 (vectorizable_mask_load_store): Update accordingly.
20743 (vectorizable_store): Likewise.
20744 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
20746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20748 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
20749 IFN_MASK_LOAD calls here rather than...
20750 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
20752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20753 Alan Hayward <alan.hayward@arm.com>
20754 David Sherwood <david.sherwood@arm.com>
20756 * expmed.c (extract_bit_field_1): For vector extracts,
20757 fall back to extract_bit_field_as_subreg if vec_extract
20760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20761 Alan Hayward <alan.hayward@arm.com>
20762 David Sherwood <david.sherwood@arm.com>
20764 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
20765 they are variable or constant sized.
20766 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
20767 slots for constant-sized data.
20769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20770 Alan Hayward <alan.hayward@arm.com>
20771 David Sherwood <david.sherwood@arm.com>
20773 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
20774 handling COND_EXPRs with boolean comparisons, try to find a better
20775 basis for the mask type than the boolean itself.
20777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20779 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
20780 is calculated and how it can be overridden.
20781 * genmodes.c (max_bitsize_mode_any_mode): New variable.
20782 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
20784 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
20787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20788 Alan Hayward <alan.hayward@arm.com>
20789 David Sherwood <david.sherwood@arm.com>
20791 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
20792 Remove the mode argument.
20793 (aarch64_simd_valid_immediate): Remove the mode and inverse
20795 * config/aarch64/iterators.md (bitsize): New iterator.
20796 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
20797 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
20798 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
20799 aarch64_simd_valid_immediate.
20800 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
20801 (aarch64_reg_or_bic_imm): Likewise.
20802 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
20803 with an insn_type enum and msl with a modifier_type enum.
20804 Replace element_width with a scalar_mode. Change the shift
20805 to unsigned int. Add constructors for scalar_float_mode and
20806 scalar_int_mode elements.
20807 (aarch64_vect_float_const_representable_p): Delete.
20808 (aarch64_can_const_movi_rtx_p)
20809 (aarch64_simd_scalar_immediate_valid_for_move)
20810 (aarch64_simd_make_constant): Update call to
20811 aarch64_simd_valid_immediate.
20812 (aarch64_advsimd_valid_immediate_hs): New function.
20813 (aarch64_advsimd_valid_immediate): Likewise.
20814 (aarch64_simd_valid_immediate): Remove mode and inverse
20815 arguments. Rewrite to use the above. Use const_vec_duplicate_p
20816 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
20817 and aarch64_float_const_representable_p on the result.
20818 (aarch64_output_simd_mov_immediate): Remove mode argument.
20819 Update call to aarch64_simd_valid_immediate and use of
20820 simd_immediate_info.
20821 (aarch64_output_scalar_simd_mov_immediate): Update call
20824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20825 Alan Hayward <alan.hayward@arm.com>
20826 David Sherwood <david.sherwood@arm.com>
20828 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
20829 (mode_nunits): Likewise CONST_MODE_NUNITS.
20830 * machmode.def (ADJUST_NUNITS): Document.
20831 * genmodes.c (mode_data::need_nunits_adj): New field.
20832 (blank_mode): Update accordingly.
20833 (adj_nunits): New variable.
20834 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
20836 (emit_mode_size_inline): Set need_bytesize_adj for all modes
20837 listed in adj_nunits.
20838 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
20839 listed in adj_nunits. Don't emit case statements for such modes.
20840 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
20841 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
20842 nothing if adj_nunits is nonnull.
20843 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
20844 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
20845 (emit_mode_fbit): Update use of print_maybe_const_decl.
20846 (emit_move_size): Likewise. Treat the array as non-const
20848 (emit_mode_adjustments): Handle adj_nunits.
20850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20852 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
20853 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
20854 (VECTOR_MODES): Use it.
20855 (make_vector_modes): Take the prefix as an argument.
20857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20858 Alan Hayward <alan.hayward@arm.com>
20859 David Sherwood <david.sherwood@arm.com>
20861 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
20862 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
20863 for MODE_VECTOR_BOOL.
20864 * machmode.def (VECTOR_BOOL_MODE): Document.
20865 * genmodes.c (VECTOR_BOOL_MODE): New macro.
20866 (make_vector_bool_mode): New function.
20867 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
20869 * lto-streamer-in.c (lto_input_mode_table): Likewise.
20870 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
20872 * stor-layout.c (int_mode_for_mode): Likewise.
20873 * tree.c (build_vector_type_for_mode): Likewise.
20874 * varasm.c (output_constant_pool_2): Likewise.
20875 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
20876 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
20877 for MODE_VECTOR_BOOL.
20878 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
20879 of mode class checks.
20880 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
20881 instead of a list of mode class checks.
20882 (expand_vector_scalar_condition): Likewise.
20883 (type_for_widest_vector_mode): Handle BImode as an inner mode.
20885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20886 Alan Hayward <alan.hayward@arm.com>
20887 David Sherwood <david.sherwood@arm.com>
20889 * machmode.h (mode_size): Change from unsigned short to
20891 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
20892 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20893 or if measurement_type is not polynomial.
20894 (fixed_size_mode::includes_p): Check for constant-sized modes.
20895 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
20896 return a poly_uint16 rather than an unsigned short.
20897 (emit_mode_size): Change the type of mode_size from unsigned short
20898 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
20899 (emit_mode_adjustments): Cope with polynomial vector sizes.
20900 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20902 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20904 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
20905 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
20906 * caller-save.c (setup_save_areas): Likewise.
20907 (replace_reg_with_saved_mem): Likewise.
20908 * calls.c (emit_library_call_value_1): Likewise.
20909 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
20910 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
20911 (gen_lowpart_for_combine): Likewise.
20912 * convert.c (convert_to_integer_1): Likewise.
20913 * cse.c (equiv_constant, cse_insn): Likewise.
20914 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
20915 (cselib_subst_to_values): Likewise.
20916 * dce.c (word_dce_process_block): Likewise.
20917 * df-problems.c (df_word_lr_mark_ref): Likewise.
20918 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
20919 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
20920 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
20921 (rtl_for_decl_location): Likewise.
20922 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
20923 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
20924 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
20925 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
20926 (expand_expr_real_1): Likewise.
20927 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
20928 (pad_below): Likewise.
20929 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20930 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
20931 * ira.c (get_subreg_tracking_sizes): Likewise.
20932 * ira-build.c (ira_create_allocno_objects): Likewise.
20933 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
20934 (ira_sort_regnos_for_alter_reg): Likewise.
20935 * ira-costs.c (record_operand_costs): Likewise.
20936 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
20937 (resolve_simple_move): Likewise.
20938 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
20939 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
20940 (lra_constraints): Likewise.
20941 (CONST_POOL_OK_P): Reject variable-sized modes.
20942 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
20943 (add_pseudo_to_slot, lra_spill): Likewise.
20944 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20945 * optabs-query.c (get_best_extraction_insn): Likewise.
20946 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20947 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
20948 (expand_mult_highpart, valid_multiword_target_p): Likewise.
20949 * recog.c (offsettable_address_addr_space_p): Likewise.
20950 * regcprop.c (maybe_mode_change): Likewise.
20951 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
20952 * regrename.c (build_def_use): Likewise.
20953 * regstat.c (dump_reg_info): Likewise.
20954 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
20955 (find_reloads, find_reloads_subreg_address): Likewise.
20956 * reload1.c (eliminate_regs_1): Likewise.
20957 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
20958 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
20959 (simplify_binary_operation_1, simplify_subreg): Likewise.
20960 * targhooks.c (default_function_arg_padding): Likewise.
20961 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
20962 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
20963 (verify_gimple_assign_ternary): Likewise.
20964 * tree-inline.c (estimate_move_cost): Likewise.
20965 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20966 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
20967 (get_address_cost_ainc): Likewise.
20968 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
20969 (vect_supportable_dr_alignment): Likewise.
20970 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20971 (vectorizable_reduction): Likewise.
20972 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
20973 (vectorizable_operation, vectorizable_load): Likewise.
20974 * tree.c (build_same_sized_truth_vector_type): Likewise.
20975 * valtrack.c (cleanup_auto_inc_dec): Likewise.
20976 * var-tracking.c (emit_note_insn_var_location): Likewise.
20977 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
20978 (ADDR_VEC_ALIGN): Likewise.
20980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20981 Alan Hayward <alan.hayward@arm.com>
20982 David Sherwood <david.sherwood@arm.com>
20984 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
20986 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20987 or if measurement_type is polynomial.
20988 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
20989 * combine.c (make_extraction): Likewise.
20990 * dse.c (find_shift_sequence): Likewise.
20991 * dwarf2out.c (mem_loc_descriptor): Likewise.
20992 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
20993 (extract_bit_field, extract_low_bits): Likewise.
20994 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
20995 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
20996 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
20997 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
20998 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20999 * reload.c (find_reloads): Likewise.
21000 * reload1.c (alter_reg): Likewise.
21001 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
21002 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
21003 * tree-if-conv.c (predicate_mem_writes): Likewise.
21004 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
21005 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
21006 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
21007 * valtrack.c (dead_debug_insert_temp): Likewise.
21008 * varasm.c (mergeable_constant_section): Likewise.
21009 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
21011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21012 Alan Hayward <alan.hayward@arm.com>
21013 David Sherwood <david.sherwood@arm.com>
21015 * expr.c (expand_assignment): Cope with polynomial mode sizes
21016 when assigning to a CONCAT.
21018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21019 Alan Hayward <alan.hayward@arm.com>
21020 David Sherwood <david.sherwood@arm.com>
21022 * machmode.h (mode_precision): Change from unsigned short to
21024 (mode_to_precision): Return a poly_uint16 rather than an unsigned
21026 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
21027 or if measurement_type is not polynomial.
21028 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
21029 in which the mode is already known to be a scalar_int_mode.
21030 * genmodes.c (emit_mode_precision): Change the type of mode_precision
21031 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
21033 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21034 for GET_MODE_PRECISION.
21035 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21036 for GET_MODE_PRECISION.
21037 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
21039 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
21040 (expand_field_assignment, make_extraction): Likewise.
21041 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
21042 (get_last_value): Likewise.
21043 * convert.c (convert_to_integer_1): Likewise.
21044 * cse.c (cse_insn): Likewise.
21045 * expr.c (expand_expr_real_1): Likewise.
21046 * lra-constraints.c (simplify_operand_subreg): Likewise.
21047 * optabs-query.c (can_atomic_load_p): Likewise.
21048 * optabs.c (expand_atomic_load): Likewise.
21049 (expand_atomic_store): Likewise.
21050 * ree.c (combine_reaching_defs): Likewise.
21051 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
21052 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
21053 * tree.h (type_has_mode_precision_p): Likewise.
21054 * ubsan.c (instrument_si_overflow): Likewise.
21056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21057 Alan Hayward <alan.hayward@arm.com>
21058 David Sherwood <david.sherwood@arm.com>
21060 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
21061 polynomial numbers of units.
21062 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
21063 (valid_vector_subparts_p): New function.
21064 (build_vector_type): Remove temporary shim and take the number
21065 of units as a poly_uint64 rather than an int.
21066 (build_opaque_vector_type): Take the number of units as a
21067 poly_uint64 rather than an int.
21068 * tree.c (build_vector_from_ctor): Handle polynomial
21069 TYPE_VECTOR_SUBPARTS.
21070 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
21071 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
21072 (build_vector_from_val): If the number of units is variable,
21073 use build_vec_duplicate_cst for constant operands and
21074 VEC_DUPLICATE_EXPR otherwise.
21075 (make_vector_type): Remove temporary is_constant ().
21076 (build_vector_type, build_opaque_vector_type): Take the number of
21077 units as a poly_uint64 rather than an int.
21078 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
21080 * cfgexpand.c (expand_debug_expr): Likewise.
21081 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
21082 (store_constructor, expand_expr_real_1): Likewise.
21083 (const_scalar_mask_from_tree): Likewise.
21084 * fold-const-call.c (fold_const_reduction): Likewise.
21085 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
21086 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
21087 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
21088 (fold_relational_const): Likewise.
21089 (native_interpret_vector): Likewise. Change the size from an
21090 int to an unsigned int.
21091 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
21092 TYPE_VECTOR_SUBPARTS.
21093 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
21094 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
21095 duplicating a non-constant operand into a variable-length vector.
21096 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
21097 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
21098 * ipa-icf.c (sem_variable::equals): Likewise.
21099 * match.pd: Likewise.
21100 * omp-simd-clone.c (simd_clone_subparts): Likewise.
21101 * print-tree.c (print_node): Likewise.
21102 * stor-layout.c (layout_type): Likewise.
21103 * targhooks.c (default_builtin_vectorization_cost): Likewise.
21104 * tree-cfg.c (verify_gimple_comparison): Likewise.
21105 (verify_gimple_assign_binary): Likewise.
21106 (verify_gimple_assign_ternary): Likewise.
21107 (verify_gimple_assign_single): Likewise.
21108 * tree-pretty-print.c (dump_generic_node): Likewise.
21109 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21110 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
21111 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
21112 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
21113 (vect_shift_permute_load_chain): Likewise.
21114 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
21115 (expand_vector_condition, optimize_vector_constructor): Likewise.
21116 (lower_vec_perm, get_compute_type): Likewise.
21117 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
21118 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
21119 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
21120 (vect_recog_mask_conversion_pattern): Likewise.
21121 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
21122 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
21123 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21124 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
21125 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
21126 (vectorizable_shift, vectorizable_operation, vectorizable_store)
21127 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
21128 (supportable_widening_operation): Likewise.
21129 (supportable_narrowing_operation): Likewise.
21130 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
21132 * varasm.c (output_constant): Likewise.
21134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21135 Alan Hayward <alan.hayward@arm.com>
21136 David Sherwood <david.sherwood@arm.com>
21138 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
21139 so that both the length == 3 and length != 3 cases set up their
21140 own permute vectors. Add comments explaining why we know the
21141 number of elements is constant.
21142 (vect_permute_load_chain): Likewise.
21144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21145 Alan Hayward <alan.hayward@arm.com>
21146 David Sherwood <david.sherwood@arm.com>
21148 * machmode.h (mode_nunits): Change from unsigned char to
21150 (ONLY_FIXED_SIZE_MODES): New macro.
21151 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
21152 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
21153 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
21155 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
21156 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
21157 or if measurement_type is not polynomial.
21158 * genmodes.c (ZERO_COEFFS): New macro.
21159 (emit_mode_nunits_inline): Make mode_nunits_inline return a
21161 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
21162 Use ZERO_COEFFS when emitting initializers.
21163 * data-streamer.h (bp_pack_poly_value): New function.
21164 (bp_unpack_poly_value): Likewise.
21165 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
21166 for GET_MODE_NUNITS.
21167 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
21168 for GET_MODE_NUNITS.
21169 * tree.c (make_vector_type): Remove temporary shim and make
21170 the real function take the number of units as a poly_uint64
21171 rather than an int.
21172 (build_vector_type_for_mode): Handle polynomial nunits.
21173 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
21174 * emit-rtl.c (const_vec_series_p_1): Likewise.
21175 (gen_rtx_CONST_VECTOR): Likewise.
21176 * fold-const.c (test_vec_duplicate_folding): Likewise.
21177 * genrecog.c (validate_pattern): Likewise.
21178 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
21179 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
21180 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
21181 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
21182 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
21183 * rtlanal.c (subreg_get_info): Likewise.
21184 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21185 (vect_grouped_load_supported): Likewise.
21186 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
21187 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
21188 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
21189 (simplify_const_unary_operation, simplify_binary_operation_1)
21190 (simplify_const_binary_operation, simplify_ternary_operation)
21191 (test_vector_ops_duplicate, test_vector_ops): Likewise.
21192 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
21193 instead of CONST_VECTOR_NUNITS.
21194 * varasm.c (output_constant_pool_2): Likewise.
21195 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
21196 explicit-encoded elements in the XVEC for variable-length vectors.
21198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21200 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
21202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21203 Alan Hayward <alan.hayward@arm.com>
21204 David Sherwood <david.sherwood@arm.com>
21206 * coretypes.h (fixed_size_mode): Declare.
21207 (fixed_size_mode_pod): New typedef.
21208 * builtins.h (target_builtins::x_apply_args_mode)
21209 (target_builtins::x_apply_result_mode): Change type to
21210 fixed_size_mode_pod.
21211 * builtins.c (apply_args_size, apply_result_size, result_vector)
21212 (expand_builtin_apply_args_1, expand_builtin_apply)
21213 (expand_builtin_return): Update accordingly.
21215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21217 * cse.c (hash_rtx_cb): Hash only the encoded elements.
21218 * cselib.c (cselib_hash_rtx): Likewise.
21219 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
21220 CONST_VECTOR encoding.
21222 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21223 Jeff Law <law@redhat.com>
21226 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
21227 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
21228 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
21229 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
21232 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
21233 explicitly probe *sp in a noreturn function if there were any callee
21234 register saves or frame pointer is needed.
21236 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21239 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
21240 BLKmode for ternary, binary or unary expressions.
21243 * var-tracking.c (delete_vta_debug_insn): New inline function.
21244 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
21245 insns from get_insns () to NULL instead of each bb separately.
21246 Use delete_vta_debug_insn. No longer static.
21247 (vt_debug_insns_local, variable_tracking_main_1): Adjust
21248 delete_vta_debug_insns callers.
21249 * rtl.h (delete_vta_debug_insns): Declare.
21250 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
21251 instead of variable_tracking_main.
21253 2018-01-03 Martin Sebor <msebor@redhat.com>
21255 PR tree-optimization/83603
21256 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
21257 arguments past the endof the argument list in functions declared
21258 without a prototype.
21259 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
21260 Avoid checking when arguments are null.
21262 2018-01-03 Martin Sebor <msebor@redhat.com>
21265 * doc/extend.texi (attribute const): Fix a typo.
21266 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
21267 issuing -Wsuggest-attribute for void functions.
21269 2018-01-03 Martin Sebor <msebor@redhat.com>
21271 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
21272 offset_int::from instead of wide_int::to_shwi.
21273 (maybe_diag_overlap): Remove assertion.
21274 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
21275 * gimple-ssa-sprintf.c (format_directive): Same.
21276 (parse_directive): Same.
21277 (sprintf_dom_walker::compute_format_length): Same.
21278 (try_substitute_return_value): Same.
21280 2018-01-03 Jeff Law <law@redhat.com>
21282 PR middle-end/83654
21283 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
21284 non-constant residual for zero at runtime and avoid probing in
21285 that case. Reorganize code for trailing problem to mirror handling
21288 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21290 PR tree-optimization/83501
21291 * tree-ssa-strlen.c (get_string_cst): New.
21292 (handle_char_store): Call get_string_cst.
21294 2018-01-03 Martin Liska <mliska@suse.cz>
21296 PR tree-optimization/83593
21297 * tree-ssa-strlen.c: Include tree-cfg.h.
21298 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
21299 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
21300 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
21302 (strlen_dom_walker::before_dom_children): Call
21303 gimple_purge_dead_eh_edges. Dump tranformation with details
21305 (strlen_dom_walker::before_dom_children): Update call by adding
21306 new argument cleanup_eh.
21307 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
21309 2018-01-03 Martin Liska <mliska@suse.cz>
21312 * cif-code.def (VARIADIC_THUNK): New enum value.
21313 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
21316 2018-01-03 Jan Beulich <jbeulich@suse.com>
21318 * sse.md (mov<mode>_internal): Tighten condition for when to use
21319 vmovdqu<ssescalarsize> for TI and OI modes.
21321 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21323 Update copyright years.
21325 2018-01-03 Martin Liska <mliska@suse.cz>
21328 * ipa-visibility.c (function_and_variable_visibility): Skip
21329 functions with noipa attribure.
21331 2018-01-03 Jakub Jelinek <jakub@redhat.com>
21333 * gcc.c (process_command): Update copyright notice dates.
21334 * gcov-dump.c (print_version): Ditto.
21335 * gcov.c (print_version): Ditto.
21336 * gcov-tool.c (print_version): Ditto.
21337 * gengtype.c (create_file): Ditto.
21338 * doc/cpp.texi: Bump @copying's copyright year.
21339 * doc/cppinternals.texi: Ditto.
21340 * doc/gcc.texi: Ditto.
21341 * doc/gccint.texi: Ditto.
21342 * doc/gcov.texi: Ditto.
21343 * doc/install.texi: Ditto.
21344 * doc/invoke.texi: Ditto.
21346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21348 * vector-builder.h (vector_builder::m_full_nelts): Change from
21349 unsigned int to poly_uint64.
21350 (vector_builder::full_nelts): Update prototype accordingly.
21351 (vector_builder::new_vector): Likewise.
21352 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
21353 (vector_builder::operator ==): Likewise.
21354 (vector_builder::finalize): Likewise.
21355 * int-vector-builder.h (int_vector_builder::int_vector_builder):
21356 Take the number of elements as a poly_uint64 rather than an
21358 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
21359 from unsigned int to poly_uint64.
21360 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
21361 (vec_perm_indices::new_vector): Likewise.
21362 (vec_perm_indices::length): Likewise.
21363 (vec_perm_indices::nelts_per_input): Likewise.
21364 (vec_perm_indices::input_nelts): Likewise.
21365 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
21366 number of elements per input as a poly_uint64 rather than an
21367 unsigned int. Use the original encoding for variable-length
21368 vectors, rather than clamping each individual element.
21369 For the second and subsequent elements in each pattern,
21370 clamp the step and base before clamping their sum.
21371 (vec_perm_indices::series_p): Handle polynomial element counts.
21372 (vec_perm_indices::all_in_range_p): Likewise.
21373 (vec_perm_indices_to_tree): Likewise.
21374 (vec_perm_indices_to_rtx): Likewise.
21375 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
21376 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
21377 (tree_vector_builder::new_binary_operation): Handle polynomial
21378 element counts. Return false if we need to know the number
21379 of elements at compile time.
21380 * fold-const.c (fold_vec_perm): Punt if the number of elements
21381 isn't known at compile time.
21383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21385 * vec-perm-indices.h (vec_perm_builder): Change element type
21386 from HOST_WIDE_INT to poly_int64.
21387 (vec_perm_indices::element_type): Update accordingly.
21388 (vec_perm_indices::clamp): Handle polynomial element_types.
21389 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21390 (vec_perm_indices::all_in_range_p): Likewise.
21391 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
21393 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
21394 polynomial vec_perm_indices element types.
21395 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
21396 * fold-const.c (fold_vec_perm): Likewise.
21397 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
21398 * tree-vect-generic.c (lower_vec_perm): Likewise.
21399 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21400 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
21401 element type to HOST_WIDE_INT.
21403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21404 Alan Hayward <alan.hayward@arm.com>
21405 David Sherwood <david.sherwood@arm.com>
21407 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
21408 rather than an int. Use plus_constant.
21409 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
21410 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
21412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21413 Alan Hayward <alan.hayward@arm.com>
21414 David Sherwood <david.sherwood@arm.com>
21416 * calls.c (emit_call_1, expand_call): Change struct_value_size from
21417 a HOST_WIDE_INT to a poly_int64.
21419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21420 Alan Hayward <alan.hayward@arm.com>
21421 David Sherwood <david.sherwood@arm.com>
21423 * calls.c (load_register_parameters): Cope with polynomial
21424 mode sizes. Require a constant size for BLKmode parameters
21425 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
21426 forces a parameter to be padded at the lsb end in order to
21427 fill a complete number of words, require the parameter size
21428 to be ordered wrt UNITS_PER_WORD.
21430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21431 Alan Hayward <alan.hayward@arm.com>
21432 David Sherwood <david.sherwood@arm.com>
21434 * reload1.c (spill_stack_slot_width): Change element type
21435 from unsigned int to poly_uint64_pod.
21436 (alter_reg): Treat mode sizes as polynomial.
21438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21439 Alan Hayward <alan.hayward@arm.com>
21440 David Sherwood <david.sherwood@arm.com>
21442 * reload.c (complex_word_subreg_p): New function.
21443 (reload_inner_reg_of_subreg, push_reload): Use it.
21445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21446 Alan Hayward <alan.hayward@arm.com>
21447 David Sherwood <david.sherwood@arm.com>
21449 * lra-constraints.c (process_alt_operands): Reject matched
21450 operands whose sizes aren't ordered.
21451 (match_reload): Refer to this check here.
21453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21454 Alan Hayward <alan.hayward@arm.com>
21455 David Sherwood <david.sherwood@arm.com>
21457 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
21458 that the mode size is in the set {1, 2, 4, 8, 16}.
21460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21461 Alan Hayward <alan.hayward@arm.com>
21462 David Sherwood <david.sherwood@arm.com>
21464 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
21465 Use plus_constant instead of gen_rtx_PLUS.
21467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21468 Alan Hayward <alan.hayward@arm.com>
21469 David Sherwood <david.sherwood@arm.com>
21471 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
21472 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
21473 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
21474 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
21475 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
21476 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
21477 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
21478 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
21479 * config/i386/i386.c (ix86_push_rounding): ...this new function.
21480 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
21482 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
21483 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
21484 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
21485 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
21486 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
21487 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
21488 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
21489 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
21490 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
21491 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
21493 * expr.c (emit_move_resolve_push): Treat the input and result
21494 of PUSH_ROUNDING as a poly_int64.
21495 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
21496 (emit_push_insn): Likewise.
21497 * lra-eliminations.c (mark_not_eliminable): Likewise.
21498 * recog.c (push_operand): Likewise.
21499 * reload1.c (elimination_effects): Likewise.
21500 * rtlanal.c (nonzero_bits1): Likewise.
21501 * calls.c (store_one_arg): Likewise. Require the padding to be
21502 known at compile time.
21504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21505 Alan Hayward <alan.hayward@arm.com>
21506 David Sherwood <david.sherwood@arm.com>
21508 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
21509 Use plus_constant instead of gen_rtx_PLUS.
21511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21512 Alan Hayward <alan.hayward@arm.com>
21513 David Sherwood <david.sherwood@arm.com>
21515 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
21516 rather than an int.
21518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21519 Alan Hayward <alan.hayward@arm.com>
21520 David Sherwood <david.sherwood@arm.com>
21522 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
21523 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
21524 via stack temporaries. Treat the mode size as polynomial too.
21526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21527 Alan Hayward <alan.hayward@arm.com>
21528 David Sherwood <david.sherwood@arm.com>
21530 * expr.c (expand_expr_real_2): When handling conversions involving
21531 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
21532 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
21533 as a poly_uint64 too.
21535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21536 Alan Hayward <alan.hayward@arm.com>
21537 David Sherwood <david.sherwood@arm.com>
21539 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
21541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21542 Alan Hayward <alan.hayward@arm.com>
21543 David Sherwood <david.sherwood@arm.com>
21545 * combine.c (can_change_dest_mode): Handle polynomial
21546 REGMODE_NATURAL_SIZE.
21547 * expmed.c (store_bit_field_1): Likewise.
21548 * expr.c (store_constructor): Likewise.
21549 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
21550 and polynomial REGMODE_NATURAL_SIZE.
21551 (gen_lowpart_common): Likewise.
21552 * reginfo.c (record_subregs_of_mode): Likewise.
21553 * rtlanal.c (read_modify_subreg_p): Likewise.
21555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21556 Alan Hayward <alan.hayward@arm.com>
21557 David Sherwood <david.sherwood@arm.com>
21559 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
21560 numbers of elements.
21562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21563 Alan Hayward <alan.hayward@arm.com>
21564 David Sherwood <david.sherwood@arm.com>
21566 * match.pd: Cope with polynomial numbers of vector elements.
21568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21569 Alan Hayward <alan.hayward@arm.com>
21570 David Sherwood <david.sherwood@arm.com>
21572 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
21573 in a POINTER_PLUS_EXPR.
21575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21576 Alan Hayward <alan.hayward@arm.com>
21577 David Sherwood <david.sherwood@arm.com>
21579 * omp-simd-clone.c (simd_clone_subparts): New function.
21580 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
21581 (ipa_simd_modify_function_body): Likewise.
21583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21584 Alan Hayward <alan.hayward@arm.com>
21585 David Sherwood <david.sherwood@arm.com>
21587 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
21588 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
21589 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
21590 (expand_vector_condition, vector_element): Likewise.
21591 (subparts_gt): New function.
21592 (get_compute_type): Use subparts_gt.
21593 (count_type_subparts): Delete.
21594 (expand_vector_operations_1): Use subparts_gt instead of
21595 count_type_subparts.
21597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21598 Alan Hayward <alan.hayward@arm.com>
21599 David Sherwood <david.sherwood@arm.com>
21601 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
21602 (vect_compile_time_alias): ...this new function. Do the calculation
21603 on poly_ints rather than trees.
21604 (vect_prune_runtime_alias_test_list): Update call accordingly.
21606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21607 Alan Hayward <alan.hayward@arm.com>
21608 David Sherwood <david.sherwood@arm.com>
21610 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
21612 (vect_schedule_slp_instance): Likewise.
21614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21615 Alan Hayward <alan.hayward@arm.com>
21616 David Sherwood <david.sherwood@arm.com>
21618 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
21619 constant and extern definitions for variable-length vectors.
21620 (vect_get_constant_vectors): Note that the number of units
21621 is known to be constant.
21623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21624 Alan Hayward <alan.hayward@arm.com>
21625 David Sherwood <david.sherwood@arm.com>
21627 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
21628 of units as polynomial. Choose between WIDE and NARROW based
21631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21632 Alan Hayward <alan.hayward@arm.com>
21633 David Sherwood <david.sherwood@arm.com>
21635 * tree-vect-stmts.c (simd_clone_subparts): New function.
21636 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
21638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21639 Alan Hayward <alan.hayward@arm.com>
21640 David Sherwood <david.sherwood@arm.com>
21642 * tree-vect-stmts.c (vectorizable_call): Treat the number of
21643 vectors as polynomial. Use build_index_vector for
21644 IFN_GOMP_SIMD_LANE.
21646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21647 Alan Hayward <alan.hayward@arm.com>
21648 David Sherwood <david.sherwood@arm.com>
21650 * tree-vect-stmts.c (get_load_store_type): Treat the number of
21651 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
21652 for variable-length vectors.
21653 (vectorizable_mask_load_store): Treat the number of units as
21654 polynomial, asserting that it is constant if the condition has
21655 already been enforced.
21656 (vectorizable_store, vectorizable_load): Likewise.
21658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21659 Alan Hayward <alan.hayward@arm.com>
21660 David Sherwood <david.sherwood@arm.com>
21662 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
21663 of units as polynomial. Punt if we can't tell at compile time
21664 which vector contains the final result.
21666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21667 Alan Hayward <alan.hayward@arm.com>
21668 David Sherwood <david.sherwood@arm.com>
21670 * tree-vect-loop.c (vectorizable_induction): Treat the number
21671 of units as polynomial. Punt on SLP inductions. Use an integer
21672 VEC_SERIES_EXPR for variable-length integer reductions. Use a
21673 cast of such a series for variable-length floating-point
21676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21677 Alan Hayward <alan.hayward@arm.com>
21678 David Sherwood <david.sherwood@arm.com>
21680 * tree.h (build_index_vector): Declare.
21681 * tree.c (build_index_vector): New function.
21682 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
21683 of units as polynomial, forcibly converting it to a constant if
21684 vectorizable_reduction has already enforced the condition.
21685 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
21686 to create a {1,2,3,...} vector.
21687 (vectorizable_reduction): Treat the number of units as polynomial.
21688 Choose vectype_in based on the largest scalar element size rather
21689 than the smallest number of units. Enforce the restrictions
21692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21693 Alan Hayward <alan.hayward@arm.com>
21694 David Sherwood <david.sherwood@arm.com>
21696 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
21697 number of units as polynomial.
21699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21700 Alan Hayward <alan.hayward@arm.com>
21701 David Sherwood <david.sherwood@arm.com>
21703 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
21704 * target.def (autovectorize_vector_sizes): Return the vector sizes
21705 by pointer, using vector_sizes rather than a bitmask.
21706 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
21707 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
21708 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
21710 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
21711 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
21712 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
21713 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
21714 * omp-general.c (omp_max_vf): Likewise.
21715 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21716 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
21717 * tree-vect-loop.c (vect_analyze_loop): Likewise.
21718 * tree-vect-slp.c (vect_slp_bb): Likewise.
21719 * doc/tm.texi: Regenerate.
21720 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
21722 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
21723 the vector size as a poly_uint64 rather than an unsigned int.
21724 (current_vector_size): Change from an unsigned int to a poly_uint64.
21725 (get_vectype_for_scalar_type): Update accordingly.
21726 * tree.h (build_truth_vector_type): Take the size and number of
21727 units as a poly_uint64 rather than an unsigned int.
21728 (build_vector_type): Add a temporary overload that takes
21729 the number of units as a poly_uint64 rather than an unsigned int.
21730 * tree.c (make_vector_type): Likewise.
21731 (build_truth_vector_type): Take the number of units as a poly_uint64
21732 rather than an unsigned int.
21734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21735 Alan Hayward <alan.hayward@arm.com>
21736 David Sherwood <david.sherwood@arm.com>
21738 * target.def (get_mask_mode): Take the number of units and length
21739 as poly_uint64s rather than unsigned ints.
21740 * targhooks.h (default_get_mask_mode): Update accordingly.
21741 * targhooks.c (default_get_mask_mode): Likewise.
21742 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
21743 * doc/tm.texi: Regenerate.
21745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21746 Alan Hayward <alan.hayward@arm.com>
21747 David Sherwood <david.sherwood@arm.com>
21749 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
21750 * omp-general.c (omp_max_vf): Likewise.
21751 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
21752 (expand_omp_simd): Handle polynomial safelen.
21753 * omp-low.c (omplow_simd_context): Add a default constructor.
21754 (omplow_simd_context::max_vf): Change from int to poly_uint64.
21755 (lower_rec_simd_input_clauses): Update accordingly.
21756 (lower_rec_input_clauses): Likewise.
21758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21759 Alan Hayward <alan.hayward@arm.com>
21760 David Sherwood <david.sherwood@arm.com>
21762 * tree-vectorizer.h (vect_nunits_for_cost): New function.
21763 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
21764 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
21765 (vect_analyze_slp_cost): Likewise.
21766 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
21767 (vect_model_load_cost): Likewise.
21769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21770 Alan Hayward <alan.hayward@arm.com>
21771 David Sherwood <david.sherwood@arm.com>
21773 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
21774 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
21775 from an unsigned int * to a poly_uint64_pod *.
21776 (calculate_unrolling_factor): New function.
21777 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
21779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21780 Alan Hayward <alan.hayward@arm.com>
21781 David Sherwood <david.sherwood@arm.com>
21783 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
21784 from an unsigned int to a poly_uint64.
21785 (_loop_vec_info::slp_unrolling_factor): Likewise.
21786 (_loop_vec_info::vectorization_factor): Change from an int
21788 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
21789 (vect_get_num_vectors): New function.
21790 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
21791 (vect_get_num_copies): Use vect_get_num_vectors.
21792 (vect_analyze_data_ref_dependences): Change max_vf from an int *
21793 to an unsigned int *.
21794 (vect_analyze_data_refs): Change min_vf from an int * to a
21796 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21797 than an unsigned HOST_WIDE_INT.
21798 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
21799 (vect_analyze_data_ref_dependence): Change max_vf from an int *
21800 to an unsigned int *.
21801 (vect_analyze_data_ref_dependences): Likewise.
21802 (vect_compute_data_ref_alignment): Handle polynomial vf.
21803 (vect_enhance_data_refs_alignment): Likewise.
21804 (vect_prune_runtime_alias_test_list): Likewise.
21805 (vect_shift_permute_load_chain): Likewise.
21806 (vect_supportable_dr_alignment): Likewise.
21807 (dependence_distance_ge_vf): Take the vectorization factor as a
21808 poly_uint64 rather than an unsigned HOST_WIDE_INT.
21809 (vect_analyze_data_refs): Change min_vf from an int * to a
21811 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
21812 vfm1 as a poly_uint64 rather than an int. Make the same change
21813 for the returned bound_scalar.
21814 (vect_gen_vector_loop_niters): Handle polynomial vf.
21815 (vect_do_peeling): Likewise. Update call to
21816 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
21817 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
21819 * tree-vect-loop.c (vect_determine_vectorization_factor)
21820 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
21821 (vect_get_known_peeling_cost): Likewise.
21822 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
21823 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
21824 (vect_transform_loop): Likewise. Use the lowest possible VF when
21825 updating the upper bounds of the loop.
21826 (vect_min_worthwhile_factor): Make static. Return an unsigned int
21827 rather than an int.
21828 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
21829 polynomial unroll factors.
21830 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
21831 (vect_make_slp_decision): Likewise.
21832 (vect_supported_load_permutation_p): Likewise, and polynomial
21834 (vect_analyze_slp_cost): Handle polynomial vf.
21835 (vect_slp_analyze_node_operations): Likewise.
21836 (vect_slp_analyze_bb_1): Likewise.
21837 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21838 than an unsigned HOST_WIDE_INT.
21839 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
21840 (vectorizable_load): Handle polynomial vf.
21841 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
21843 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
21845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21846 Alan Hayward <alan.hayward@arm.com>
21847 David Sherwood <david.sherwood@arm.com>
21849 * match.pd: Handle bit operations involving three constants
21850 and try to fold one pair.
21852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21854 * tree-vect-loop-manip.c: Include gimple-fold.h.
21855 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
21856 niters_maybe_zero parameters. Handle other cases besides a step of 1.
21857 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
21858 Add a path that uses a step of VF instead of 1, but disable it
21860 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
21861 and niters_no_overflow parameters. Update calls to
21862 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
21863 Create a new SSA name if the latter choses to use a ste other
21864 than zero, and return it via niters_vector_mult_vf_var.
21865 * tree-vect-loop.c (vect_transform_loop): Update calls to
21866 vect_do_peeling, vect_gen_vector_loop_niters and
21867 slpeel_make_loop_iterate_ntimes.
21868 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
21869 (vect_gen_vector_loop_niters): Update declarations after above changes.
21871 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
21873 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
21874 128-bit round to integer instructions.
21875 (ceil<mode>2): Likewise.
21876 (btrunc<mode>2): Likewise.
21877 (round<mode>2): Likewise.
21879 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21881 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
21882 unaligned VSX load/store on P8/P9.
21883 (expand_block_clear): Allow the use of unaligned VSX
21884 load/store on P8/P9.
21886 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21888 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
21890 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
21891 swap associated with both a load and a store.
21893 2018-01-02 Andrew Waterman <andrew@sifive.com>
21895 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
21896 * config/riscv/riscv.md (clear_cache): Use it.
21898 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
21900 * web.c: Remove out-of-date comment.
21902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21904 * expr.c (fixup_args_size_notes): Check that any existing
21905 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
21906 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
21907 (emit_single_push_insn): ...here.
21909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21911 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
21912 (const_vector_encoded_nelts): New function.
21913 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
21914 (const_vector_int_elt, const_vector_elt): Declare.
21915 * emit-rtl.c (const_vector_int_elt_1): New function.
21916 (const_vector_elt): Likewise.
21917 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
21918 of CONST_VECTOR_ELT.
21920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21922 * expr.c: Include rtx-vector-builder.h.
21923 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
21924 directly on the tree encoding.
21925 (const_vector_from_tree): Likewise.
21926 * optabs.c: Include rtx-vector-builder.h.
21927 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
21928 sequence of "u" values.
21929 * vec-perm-indices.c: Include rtx-vector-builder.h.
21930 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
21931 directly on the vec_perm_indices encoding.
21933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21935 * doc/rtl.texi (const_vector): Describe new encoding scheme.
21936 * Makefile.in (OBJS): Add rtx-vector-builder.o.
21937 * rtx-vector-builder.h: New file.
21938 * rtx-vector-builder.c: Likewise.
21939 * rtl.h (rtx_def::u2): Add a const_vector field.
21940 (CONST_VECTOR_NPATTERNS): New macro.
21941 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
21942 (CONST_VECTOR_DUPLICATE_P): Likewise.
21943 (CONST_VECTOR_STEPPED_P): Likewise.
21944 (CONST_VECTOR_ENCODED_ELT): Likewise.
21945 (const_vec_duplicate_p): Check for a duplicated vector encoding.
21946 (unwrap_const_vec_duplicate): Likewise.
21947 (const_vec_series_p): Check for a non-duplicated vector encoding.
21948 Say that the function only returns true for integer vectors.
21949 * emit-rtl.c: Include rtx-vector-builder.h.
21950 (gen_const_vec_duplicate_1): Delete.
21951 (gen_const_vector): Call gen_const_vec_duplicate instead of
21952 gen_const_vec_duplicate_1.
21953 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
21954 (gen_const_vec_duplicate): Use rtx_vector_builder.
21955 (gen_const_vec_series): Likewise.
21956 (gen_rtx_CONST_VECTOR): Likewise.
21957 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
21958 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21959 Build a new vector rather than modifying a CONST_VECTOR in-place.
21960 (handle_special_swappables): Update call accordingly.
21961 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
21962 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21963 Build a new vector rather than modifying a CONST_VECTOR in-place.
21964 (handle_special_swappables): Update call accordingly.
21966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21968 * simplify-rtx.c (simplify_const_binary_operation): Use
21969 CONST_VECTOR_ELT instead of XVECEXP.
21971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21973 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
21974 the selector elements to be different from the data elements
21975 if the selector is a VECTOR_CST.
21976 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
21977 ssizetype for the selector.
21979 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21981 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
21982 before testing each element individually.
21983 * tree-vect-generic.c (lower_vec_perm): Likewise.
21985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21987 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
21988 * selftest-run-tests.c (selftest::run_tests): Call it.
21989 * vector-builder.h (vector_builder::operator ==): New function.
21990 (vector_builder::operator !=): Likewise.
21991 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
21992 (vec_perm_indices::all_from_input_p): New function.
21993 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21994 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
21995 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
21996 instead of reading the VECTOR_CST directly. Detect whether both
21997 vector inputs are the same before constructing the vec_perm_indices,
21998 and update the number of inputs argument accordingly. Use the
21999 utility functions added above. Only construct sel2 if we need to.
22001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22003 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
22004 the broadcast of the low byte.
22005 (expand_mult_highpart): Use an explicit encoding for the permutes.
22006 * optabs-query.c (can_mult_highpart_p): Likewise.
22007 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
22008 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22009 (vectorizable_bswap): Likewise.
22010 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
22011 explicit encoding for the power-of-2 permutes.
22012 (vect_permute_store_chain): Likewise.
22013 (vect_grouped_load_supported): Likewise.
22014 (vect_permute_load_chain): Likewise.
22016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22018 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
22019 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
22020 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
22021 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22022 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
22023 (vect_gen_perm_mask_any): Likewise.
22025 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22027 * int-vector-builder.h: New file.
22028 * vec-perm-indices.h: Include int-vector-builder.h.
22029 (vec_perm_indices): Redefine as an int_vector_builder.
22030 (auto_vec_perm_indices): Delete.
22031 (vec_perm_builder): Redefine as a stand-alone class.
22032 (vec_perm_indices::vec_perm_indices): New function.
22033 (vec_perm_indices::clamp): Likewise.
22034 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
22035 (vec_perm_indices::new_vector): New function.
22036 (vec_perm_indices::new_expanded_vector): Update for new
22037 vec_perm_indices class.
22038 (vec_perm_indices::rotate_inputs): New function.
22039 (vec_perm_indices::all_in_range_p): Operate directly on the
22040 encoded form, without computing elided elements.
22041 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
22042 encoding. Update for new vec_perm_indices class.
22043 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
22044 the given vec_perm_builder.
22045 (expand_vec_perm_var): Update vec_perm_builder constructor.
22046 (expand_mult_highpart): Use vec_perm_builder instead of
22047 auto_vec_perm_indices.
22048 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
22049 vec_perm_indices instead of auto_vec_perm_indices. Use a single
22050 or double series encoding as appropriate.
22051 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
22052 vec_perm_indices instead of auto_vec_perm_indices.
22053 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22054 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22055 (vect_permute_store_chain): Likewise.
22056 (vect_grouped_load_supported): Likewise.
22057 (vect_permute_load_chain): Likewise.
22058 (vect_shift_permute_load_chain): Likewise.
22059 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22060 (vect_transform_slp_perm_load): Likewise.
22061 (vect_schedule_slp_instance): Likewise.
22062 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22063 (vectorizable_mask_load_store): Likewise.
22064 (vectorizable_bswap): Likewise.
22065 (vectorizable_store): Likewise.
22066 (vectorizable_load): Likewise.
22067 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
22068 vec_perm_indices instead of auto_vec_perm_indices. Use
22069 tree_to_vec_perm_builder to read the vector from a tree.
22070 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
22071 vec_perm_builder instead of a vec_perm_indices.
22072 (have_whole_vector_shift): Use vec_perm_builder and
22073 vec_perm_indices instead of auto_vec_perm_indices. Leave the
22074 truncation to calc_vec_perm_mask_for_shift.
22075 (vect_create_epilog_for_reduction): Likewise.
22076 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
22077 from auto_vec_perm_indices to vec_perm_indices.
22078 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22079 instead of changing individual elements.
22080 (aarch64_vectorize_vec_perm_const): Use new_vector to install
22081 the vector in d.perm.
22082 * config/arm/arm.c (expand_vec_perm_d::perm): Change
22083 from auto_vec_perm_indices to vec_perm_indices.
22084 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
22085 instead of changing individual elements.
22086 (arm_vectorize_vec_perm_const): Use new_vector to install
22087 the vector in d.perm.
22088 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
22089 Update vec_perm_builder constructor.
22090 (rs6000_expand_interleave): Likewise.
22091 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
22092 (rs6000_expand_interleave): Likewise.
22094 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22096 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
22097 to qimode could truncate the indices.
22098 * optabs.c (expand_vec_perm_var): Likewise.
22100 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22102 * Makefile.in (OBJS): Add vec-perm-indices.o.
22103 * vec-perm-indices.h: New file.
22104 * vec-perm-indices.c: Likewise.
22105 * target.h (vec_perm_indices): Replace with a forward class
22107 (auto_vec_perm_indices): Move to vec-perm-indices.h.
22108 * optabs.h: Include vec-perm-indices.h.
22109 (expand_vec_perm): Delete.
22110 (selector_fits_mode_p, expand_vec_perm_var): Declare.
22111 (expand_vec_perm_const): Declare.
22112 * target.def (vec_perm_const_ok): Replace with...
22113 (vec_perm_const): ...this new hook.
22114 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
22115 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
22116 * doc/tm.texi: Regenerate.
22117 * optabs.def (vec_perm_const): Delete.
22118 * doc/md.texi (vec_perm_const): Likewise.
22119 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
22120 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
22121 expand_vec_perm for constant permutation vectors. Assert that
22122 the mode of variable permutation vectors is the integer equivalent
22123 of the mode that is being permuted.
22124 * optabs-query.h (selector_fits_mode_p): Declare.
22125 * optabs-query.c: Include vec-perm-indices.h.
22126 (selector_fits_mode_p): New function.
22127 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
22128 is defined, instead of checking whether the vec_perm_const_optab
22129 exists. Use targetm.vectorize.vec_perm_const instead of
22130 targetm.vectorize.vec_perm_const_ok. Check whether the indices
22131 fit in the vector mode before using a variable permute.
22132 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
22133 vec_perm_indices instead of an rtx.
22134 (expand_vec_perm): Replace with...
22135 (expand_vec_perm_const): ...this new function. Take the selector
22136 as a vec_perm_indices rather than an rtx. Also take the mode of
22137 the selector. Update call to shift_amt_for_vec_perm_mask.
22138 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
22139 Use vec_perm_indices::new_expanded_vector to expand the original
22140 selector into bytes. Check whether the indices fit in the vector
22141 mode before using a variable permute.
22142 (expand_vec_perm_var): Make global.
22143 (expand_mult_highpart): Use expand_vec_perm_const.
22144 * fold-const.c: Includes vec-perm-indices.h.
22145 * tree-ssa-forwprop.c: Likewise.
22146 * tree-vect-data-refs.c: Likewise.
22147 * tree-vect-generic.c: Likewise.
22148 * tree-vect-loop.c: Likewise.
22149 * tree-vect-slp.c: Likewise.
22150 * tree-vect-stmts.c: Likewise.
22151 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
22153 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
22154 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
22155 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
22156 (aarch64_vectorize_vec_perm_const): ...this new function.
22157 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22158 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22159 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
22160 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
22161 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22162 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22163 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
22165 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
22166 check for NEON modes.
22167 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
22168 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
22169 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
22170 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
22172 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
22173 the old VEC_PERM_CONST conditions.
22174 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
22175 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
22176 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
22177 (ia64_vectorize_vec_perm_const_ok): Merge into...
22178 (ia64_vectorize_vec_perm_const): ...this new function.
22179 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
22180 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
22181 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
22182 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
22183 * config/mips/mips.c (mips_expand_vec_perm_const)
22184 (mips_vectorize_vec_perm_const_ok): Merge into...
22185 (mips_vectorize_vec_perm_const): ...this new function.
22186 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
22187 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
22188 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
22189 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
22190 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
22191 (rs6000_expand_vec_perm_const): Delete.
22192 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
22194 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22195 (altivec_expand_vec_perm_const_le): Take each operand individually.
22196 Operate on constant selectors rather than rtxes.
22197 (altivec_expand_vec_perm_const): Likewise. Update call to
22198 altivec_expand_vec_perm_const_le.
22199 (rs6000_expand_vec_perm_const): Delete.
22200 (rs6000_vectorize_vec_perm_const_ok): Delete.
22201 (rs6000_vectorize_vec_perm_const): New function.
22202 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22203 an element count and rtx array.
22204 (rs6000_expand_extract_even): Update call accordingly.
22205 (rs6000_expand_interleave): Likewise.
22206 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
22207 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
22208 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
22209 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
22210 (rs6000_expand_vec_perm_const): Delete.
22211 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
22212 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22213 (altivec_expand_vec_perm_const_le): Take each operand individually.
22214 Operate on constant selectors rather than rtxes.
22215 (altivec_expand_vec_perm_const): Likewise. Update call to
22216 altivec_expand_vec_perm_const_le.
22217 (rs6000_expand_vec_perm_const): Delete.
22218 (rs6000_vectorize_vec_perm_const_ok): Delete.
22219 (rs6000_vectorize_vec_perm_const): New function. Remove stray
22220 reference to the SPE evmerge intructions.
22221 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
22222 an element count and rtx array.
22223 (rs6000_expand_extract_even): Update call accordingly.
22224 (rs6000_expand_interleave): Likewise.
22225 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
22226 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
22228 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
22230 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22232 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
22233 vector mode and that that mode matches the mode of the data
22235 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
22236 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
22237 directly using expand_vec_perm_1 when forcing selectors into
22239 (expand_vec_perm_var): New function, split out from expand_vec_perm.
22241 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22243 * optabs-query.h (can_vec_perm_p): Delete.
22244 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
22245 * optabs-query.c (can_vec_perm_p): Split into...
22246 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
22247 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
22248 particular selector is valid.
22249 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22250 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22251 (vect_grouped_load_supported): Likewise.
22252 (vect_shift_permute_load_chain): Likewise.
22253 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
22254 (vect_transform_slp_perm_load): Likewise.
22255 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22256 (vectorizable_bswap): Likewise.
22257 (vect_gen_perm_mask_checked): Likewise.
22258 * fold-const.c (fold_ternary_loc): Likewise. Don't take
22259 implementations of variable permutation vectors into account
22260 when deciding which selector to use.
22261 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
22262 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
22263 with a false third argument.
22264 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
22265 to test whether the constant selector is valid and can_vec_perm_var_p
22266 to test whether a variable selector is valid.
22268 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22270 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
22271 * optabs-query.c (can_vec_perm_p): Likewise.
22272 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
22273 instead of vec_perm_indices.
22274 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
22275 (vect_gen_perm_mask_checked): Likewise,
22276 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
22277 (vect_gen_perm_mask_checked): Likewise,
22279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
22281 * optabs-query.h (qimode_for_vec_perm): Declare.
22282 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
22283 (qimode_for_vec_perm): ...this new function.
22284 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
22286 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22288 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
22289 does not have a conditional at the top.
22291 2018-01-02 Richard Biener <rguenther@suse.de>
22293 * ipa-inline.c (big_speedup_p): Fix expression.
22295 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
22298 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
22301 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
22305 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
22306 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
22307 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
22308 cond_taken_branch_cost 3->4.
22310 2018-01-01 Jakub Jelinek <jakub@redhat.com>
22312 PR tree-optimization/83581
22313 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
22314 TODO_cleanup_cfg if any changes have been made.
22316 PR middle-end/83608
22317 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
22318 convert_modes if target mode has the right side, but different mode
22321 PR middle-end/83609
22322 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
22323 last argument when extracting from CONCAT. If either from_real or
22324 from_imag is NULL, use expansion through memory. If result is not
22325 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
22326 the parts directly to inner mode, if even that fails, use expansion
22329 PR middle-end/83623
22330 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
22331 check for bswap in mode rather than HImode and use that in expand_unop
22334 Copyright (C) 2018 Free Software Foundation, Inc.
22336 Copying and distribution of this file, with or without modification,
22337 are permitted in any medium without royalty provided the copyright
22338 notice and this notice are preserved.