1 2018-08-27 Martin Liska <mliska@suse.cz>
3 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4 selected expansion strategy.
6 2018-08-27 Martin Liska <mliska@suse.cz>
8 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
9 * builtins.c (is_builtin_fn): Likewise.
10 * attribs.c (diag_attr_exclusions): Use new function
11 fndecl_built_in_p and remove check for FUNCTION_DECL if
13 (builtin_mathfn_code): Likewise.
14 (fold_builtin_expect): Likewise.
15 (fold_call_expr): Likewise.
16 (fold_builtin_call_array): Likewise.
17 (fold_call_stmt): Likewise.
18 (set_builtin_user_assembler_name): Likewise.
19 (is_simple_builtin): Likewise.
20 * calls.c (gimple_alloca_call_p): Likewise.
21 (maybe_warn_nonstring_arg): Likewise.
22 * cfgexpand.c (expand_call_stmt): Likewise.
23 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
24 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
25 (cgraph_node::verify_node): Likewise.
26 * cgraphclones.c (build_function_decl_skip_args): Likewise.
27 (cgraph_node::create_clone): Likewise.
28 * config/arm/arm.c (arm_insert_attributes): Likewise.
29 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
30 * dse.c (scan_insn): Likewise.
31 * expr.c (expand_expr_real_1): Likewise.
32 * fold-const.c (operand_equal_p): Likewise.
33 (fold_binary_loc): Likewise.
34 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
35 * gimple-low.c (lower_stmt): Likewise.
36 * gimple-pretty-print.c (dump_gimple_call): Likewise.
37 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
38 * gimple.c (gimple_build_call_from_tree): Likewise.
39 (gimple_call_builtin_p): Likewise.
40 (gimple_call_combined_fn): Likewise.
41 * gimplify.c (gimplify_call_expr): Likewise.
42 (gimple_boolify): Likewise.
43 (gimplify_modify_expr): Likewise.
44 (gimplify_addr_expr): Likewise.
45 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
46 * ipa-cp.c (determine_versionability): Likewise.
47 * ipa-fnsummary.c (compute_fn_summary): Likewise.
48 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
49 * ipa-split.c (visit_bb): Likewise.
50 (split_function): Likewise.
51 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
52 * lto-cgraph.c (input_node): Likewise.
53 * lto-streamer-out.c (write_symbol): Likewise.
54 * omp-low.c (setjmp_or_longjmp_p): Likewise.
55 (lower_omp_1): Likewise.
56 * predict.c (strip_predict_hints): Likewise.
57 * print-tree.c (print_node): Likewise.
58 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
59 * trans-mem.c (is_tm_irrevocable): Likewise.
60 (is_tm_load): Likewise.
61 (is_tm_simple_load): Likewise.
62 (is_tm_store): Likewise.
63 (is_tm_simple_store): Likewise.
64 (is_tm_abort): Likewise.
65 (tm_region_init_1): Likewise.
66 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
67 * tree-cfg.c (verify_gimple_call): Likewise.
68 (move_stmt_r): Likewise.
69 (stmt_can_terminate_bb_p): Likewise.
70 * tree-eh.c (lower_eh_constructs_2): Likewise.
71 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
72 * tree-inline.c (remap_gimple_stmt): Likewise.
74 (estimate_num_insns): Likewise.
75 (fold_marked_statements): Likewise.
76 * tree-sra.c (scan_function): Likewise.
77 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
78 (optimize_stack_restore): Likewise.
79 (pass_fold_builtins::execute): Likewise.
80 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
81 (mark_all_reaching_defs_necessary_1): Likewise.
82 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
83 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
84 (pass_forwprop::execute): Likewise.
85 * tree-ssa-loop-im.c (stmt_cost): Likewise.
86 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
87 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
88 * tree-ssa-strlen.c (get_string_length): Likewise.
89 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
90 (find_func_aliases_for_call): Likewise.
91 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
92 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
93 * tree-tailcall.c (find_tail_calls): Likewise.
94 * tree.c (need_assembler_name_p): Likewise.
95 (free_lang_data_in_decl): Likewise.
96 (get_call_combined_fn): Likewise.
97 * ubsan.c (is_ubsan_builtin_p): Likewise.
98 * varasm.c (incorporeal_function_p): Likewise.
99 * tree.h (DECL_BUILT_IN): Remove and replace with
101 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
102 (fndecl_built_in_p): New.
104 2018-08-27 Martin Liska <mliska@suse.cz>
106 PR tree-optimization/86847
107 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
108 Dump also subtree probability.
109 (switch_decision_tree::do_jump_if_equal): New function.
110 (switch_decision_tree::emit_case_nodes): Handle special
111 situations in balanced tree that can be emitted much simpler.
112 Fix calculation of probabilities that happen in tree expansion.
113 * tree-switch-conversion.h (struct cluster): Add
115 (struct simple_cluster): Likewise.
116 (struct case_tree_node): Add new function has_child.
117 (do_jump_if_equal): New.
119 2018-08-27 Martin Liska <mliska@suse.cz>
121 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
122 Add new argument to bit_test_cluster constructor.
123 (bit_test_cluster::emit): Set bits really number of values
125 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
126 probability argument.
127 * tree-switch-conversion.h (struct bit_test_cluster):
128 Add m_handles_entire_switch.
130 2018-08-27 Martin Liska <mliska@suse.cz>
132 PR tree-optimization/86702
133 * tree-switch-conversion.c (jump_table_cluster::emit):
134 Make probabilities even for values in jump table
135 according to number of cases handled.
136 (switch_decision_tree::compute_cases_per_edge): Pass
137 argument to reset_out_edges_aux function.
138 (switch_decision_tree::analyze_switch_statement): Likewise.
139 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
142 2018-08-27 Martin Liska <mliska@suse.cz>
144 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
145 cfun argument explicitly.
146 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
147 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
148 function gimple_switch_default_bb.
149 (convert_switch_statements):
151 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
152 * stmt.c (label_to_block_fn): Use label_to_block and pass
153 cfun argument explicitly and use gimple_switch_label_bb.
154 (expand_case): Likewise.
155 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
156 cfun argument explicitly. Likewise.
157 (make_edges_bb): Likewise.
158 (make_cond_expr_edges): Likewise.
159 (get_cases_for_edge): Likewise.
160 (make_gimple_switch_edges): Likewise.
161 (label_to_block_fn): Likewise.
162 (label_to_block): Likewise.
163 (make_goto_expr_edges): Likewise.
164 (make_gimple_asm_edges): Likewise.
165 (main_block_label): Likewise.
166 (group_case_labels_stmt): Likewise.
167 (find_taken_edge_computed_goto): Likewise.
168 (find_taken_edge_switch_expr): Likewise.
169 (gimple_verify_flow_info): Likewise.
170 (gimple_redirect_edge_and_branch): Likewise.
171 (gimple_switch_label_bb): New function.
172 (gimple_switch_default_bb): Likewise.
173 (gimple_switch_edge): Likewise.
174 (gimple_switch_default_edge): Likewise.
175 * tree-cfg.h (label_to_block_fn): Remove and replace ...
176 (label_to_block): ... with this.
177 (gimple_switch_label_bb): New.
178 (gimple_switch_default_bb): Likewise.
179 (gimple_switch_edge): Likewise.
180 (gimple_switch_default_edge): Likewise.
181 * tree-cfgcleanup.c (convert_single_case_switch): Use
182 new gimple functions and pass new argument to label_to_block.
183 (cleanup_control_flow_bb):
184 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
185 cfun argument explicitly.
186 (make_eh_edges): Likewise.
187 (redirect_eh_dispatch_edge): Likewise.
188 (lower_resx): Likewise.
189 (lower_eh_dispatch): Likewise.
190 (maybe_remove_unreachable_handlers): Likewise.
191 (unsplit_eh): Likewise.
192 (cleanup_empty_eh): Likewise.
193 (verify_eh_edges): Likewise.
194 (verify_eh_dispatch_edge): Likewise.
195 * tree-ssa-dom.c (record_edge_info): Likewise.
196 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
197 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
198 (thread_through_normal_block): Likewise.
199 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
200 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
201 * tree-switch-conversion.c (switch_conversion::collect): Use new
203 (switch_conversion::check_final_bb): Likewise.
204 (switch_conversion::gather_default_values): Pass new argument
206 (switch_conversion::build_constructors): Likewise.
207 (switch_decision_tree::compute_cases_per_edge): Use new
208 gimple_switch_edge function.
209 (switch_decision_tree::analyze_switch_statement): Pass new argument
211 (switch_decision_tree::try_switch_expansion): Use
212 gimple_switch_default_edge.
213 * tree-vrp.c (find_switch_asserts): Pass new argument
215 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
216 (vr_values::simplify_switch_using_ranges): Likewise.
218 2018-08-27 Richard Biener <rguenther@suse.de>
220 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
221 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
223 * tree-ssa-sccvn.h (struct vn_pval): New structure.
224 (struct vn_nary_op_s): Add unwind_to member. Add
225 predicated_values flag and put result into a union together
226 with a linked list of vn_pval.
227 (struct vn_ssa_aux): Add name member to make maintaining
228 a map of SSA name to vn_ssa_aux possible. Remove no longer
229 needed info, dfsnum, low, visited, on_sccstack, use_processed
230 and range_info_anti_range_p members.
231 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
232 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
234 (vn_valueize): New global.
235 (vn_context_bb): Likewise.
236 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
237 VN_INFO_PTR_INFO): Remove.
238 * tree-ssa-sccvn.c: ... (rewrite)
239 (pass_fre::execute): For -O2+ initialize loops and run
240 RPO VN in optimistic mode (iterating). For -O1 and -Og
241 run RPO VN in non-optimistic mode.
242 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
243 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
244 * doc/invoke.texi (sccvn-max-scc-size): Remove.
245 (rpo-vn-max-loop-depth): Document.
246 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
247 when valuezing the VUSE signals we walked out of the region.
248 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
249 (phi_translate): Set VN context block to use for availability
251 (compute_avail): Likewise.
252 (pre_valueize): New function.
253 (pass_pre::execute): Adjust to the RPO VN API.
255 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
256 (propagate_constants_for_unrolling): Remove.
257 (tree_unroll_loops_completely): Perform value-numbering
258 on the unrolled bodies loop parent.
260 2018-08-27 Richard Biener <rguenther@suse.de>
262 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
263 for partial antic compute.
265 2018-08-27 Jakub Jelinek <jakub@redhat.com>
267 PR rtl-optimization/87065
268 * combine.c (simplify_if_then_else): Formatting fix.
269 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
271 (known_cond): Don't return const_true_rtx for vector modes. Use
272 CONST0_RTX instead of const0_rtx. Formatting fixes.
274 2018-08-27 Martin Liska <mliska@suse.cz>
276 PR gcov-profile/87069
277 * gcov.c (process_file): Record files already processed
278 and warn about a file being processed multiple times.
280 2018-08-27 Martin Liska <mliska@suse.cz>
283 * config/aarch64/aarch64.c (aarch64_override_options_internal):
284 Set default values for x_aarch64_*_string strings.
285 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
286 prefix. For -mabi do not print '=ABI' in help and use
287 <option_value> format for -msve-vector-bits and -moverride
290 2018-08-26 Jeff Law <law@redhat.com>
292 * config/mips/frame-header-opt.c: Include "backend.h" rather than
295 2018-08-26 Marek Polacek <polacek@redhat.com>
297 PR c++/87029, Implement -Wredundant-move.
298 * doc/invoke.texi: Document -Wredundant-move.
300 2018-08-25 Martin Sebor <msebor@redhat.com>
302 PR tree-optimization/87059
303 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
304 to the same type as the other.
305 * fold-const.c (fold_binary_loc): Assert expectation.
307 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
309 * config/darwin.c (machopic_legitimize_pic_address): Clean up
310 extraneous parentheses, dead code section and formatting.
312 2018-08-24 David Malcolm <dmalcolm@redhat.com>
315 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
316 wide enough for jumps in the line-numbering to be visible.
317 (layout::print_gap_in_line_numbering): New member function.
318 (layout::calculate_line_spans): When using line numbering, merge
319 line spans that are only 1 line apart.
320 (diagnostic_show_locus): When printing line numbers, show gaps in
321 line numbering directly, rather than printing headers.
322 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
323 line-numbering with multiple line spans.
324 (selftest::test_fixit_insert_containing_newline_2): Add test of
325 line-numbering, in which the spans are close enough to be merged.
327 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
329 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
330 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
331 * tree-vrp.c (range_is_nonnull): Remove.
332 (range_includes_zero_p): Accept value_range instead of min/max.
333 (extract_range_from_binary_expr_1): Do not early bail on
335 Use range_includes_zero_p instead of range_is_nonnull.
336 (extract_range_from_unary_expr): Use range_includes_zero_p instead
338 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
339 special case VR_ANTI_RANGE.
340 (vrp_finalize): Same.
341 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
343 (range_is_nonnull): Remove.
344 * vr-values.c (vrp_stmt_computes_nonzero): Use
345 range_includes_zero_p instead of range_is_nonnull.
346 (extract_range_basic): Pass value_range to range_includes_zero_p
347 instead of range_is_nonnull.
349 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
351 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
352 * emit-rtl.h (rtl_data): Remove return_bnd.
353 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
354 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
355 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
356 (POINTER_BOUNDS_MODE): Remove definition.
357 (make_pointer_bounds_mode): Remove.
358 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
359 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
360 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
361 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
362 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
363 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
364 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
366 * config/i386/i386-modes.def (BND32, BND64): Remove.
367 * config/i386/i386.c (dbx_register_map): Remove bound registers.
368 (dbx64_register_map): Ditto.
369 (svr4_dbx_register_map): Ditto.
370 (indirect_thunk_bnd_needed): Remove.
371 (indirect_thunks_bnd_used): Ditto.
372 (indirect_return_bnd_needed): Ditto.
373 (indirect_return_via_cx_bnd): Ditto.
374 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
375 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
376 (output_indirect_thunk): Ditto. Remove need_prefix argument.
377 (output_indirect_thunk_function): Remove handling of
378 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
379 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
380 (ix86_save_reg): Remove handling of crtl->return_bnd.
381 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
382 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
383 and UNSPEC_BNDLX_ADDR.
384 (ix86_output_indirect_branch_via_reg): Remove handling of
385 indirect_thunk_prefix_bnd.
386 (ix86_output_indirect_branch_via_push): Ditto.
387 (ix86_output_function_return): Ditto.
388 (ix86_output_indirect_function_return): Ditto.
389 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
390 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
391 (CALL_USED_REGISTERS): Ditto.
392 (REG_ALLOC_ORDER): Update for removal of bound registers.
393 (HI_REGISTER_NAMES): Ditto.
394 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
395 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
396 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
397 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
398 (FIRST_PSEUDO_REG): Update.
399 (BND): Remove mode iterator.
400 * config/i386/predicates.md (bnd_mem_operator): Remove.
402 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
404 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
407 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
409 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
410 the case in which the permute needs only a single element and
411 repeats for every vector of the result. Extend that case to
412 handle variable-length vectors.
413 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
415 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
418 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
419 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
421 2018-08-24 Richard Biener <rguenther@suse.de>
423 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
424 bb_flags_allocated members.
425 (auto_flag): New RAII class for allocating flags.
426 (auto_edge_flag): New RAII class for allocating edge flags.
427 (auto_bb_flag): New RAII class for allocating bb flags.
428 * cfgloop.c (verify_loop_structure): Allocate temporary edge
430 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
431 in favor of temporarily allocated BB flag.
432 * hsa-brig.c: Re-order includes.
433 * hsa-dump.c: Likewise.
434 * hsa-regalloc.c: Likewise.
435 * print-rtl.c: Likewise.
436 * profile-count.c: Likewise.
438 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
441 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
444 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
447 * wide-int-range.cc (wide_int_range_div): Do not ignore result
448 from wide_int_range_multiplicative_op.
450 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
452 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
455 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
457 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
460 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
462 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
463 full GENERIC expression used for replacement.
465 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
467 * tree-vrp.c (abs_extent_range): Remove.
468 (extract_range_into_wide_ints): Pass wide ints by reference.
469 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
470 Pass wide ints by reference in all calls to
471 extract_range_into_wide_ints.
472 * wide-int-range.cc (wide_int_range_div): New.
473 * wide-int-range.h (wide_int_range_div): New.
474 (wide_int_range_includes_zero_p): New.
475 (wide_int_range_zero_p): New.
477 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
479 * config/aarch64/aarch64.md (arches): New enum.
480 (arch): New enum attr.
481 (arch_enabled): New attr.
482 (enabled): Now uses arch_enabled only.
483 (simd, sve, fp16): Removed attribute.
484 (fp): Attr now defined in terms of 'arch'.
485 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
486 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
487 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
488 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
489 attributes into 'arch'.
490 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
491 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
492 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
493 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
494 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
495 'simd' attribute into 'arch'.
496 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
497 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
498 Convert use of 'fp' attribute to 'arch'.
499 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
500 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
502 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
503 (different modes) Merge 'fp' and 'simd' into 'arch'.
504 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
507 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
509 PR rtl-optimization/87026
510 * expmed.c (canonicalize_comparison): If we can no longer create
511 pseudoregisters, don't.
513 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
516 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
518 * config/arm/arm.c (speculation_barrier_libfunc): New static
520 (arm_init_libfuncs): Initialize it.
521 (arm_emit_speculation_barrier): New function.
522 * config/arm/arm.md (speculation_barrier): Call
523 arm_emit_speculation_barrier for architectures that do not have
525 (speculation_barrier_insn): Only match on Armv7 or later.
527 2018-08-23 Richard Biener <rguenther@suse.de>
530 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
533 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
535 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
536 of single-vector TBLs.
537 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
540 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
543 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
544 aarch64_evpc_tbl guard.
546 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
548 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
551 2018-08-22 Martin Sebor <msebor@redhat.com>
554 * tree-pretty-print.c (pretty_print_string): Add argument.
555 (dump_generic_node): Call to pretty_print_string with string size.
557 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
559 PR rtl-optimization/86771
560 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
561 of two SETs into those two SETs, one to be placed at i2, if that SETs
562 destination is modified between i2 and i3.
564 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
566 PR tree-optimization/86725
567 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
569 (vect_analyze_scalar_cycles_1): Check it.
571 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
573 PR tree-optimization/86725
574 * tree-vect-loop.c (vect_is_simple_reduction): When treating
575 an outer loop phi as a double reduction, make sure that the
576 single user of the phi result is an inner loop phi.
578 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
580 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
581 grouped stores with gaps to a strided group.
583 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
585 * tree-vect-stmts.c (get_group_load_store_type)
586 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
587 first statement in a group.
589 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
591 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
592 the sequence used in gcc/gcc.c.
594 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
597 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
600 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
602 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
603 Darwin10-specific unwinder-shim.
604 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
605 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
606 New to cater for Darwin10 Rosetta.
608 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
610 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
611 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
613 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
618 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
619 (dwarf2out_switch_text_section): Generate a local label for the
620 second function sub-section and apply it as the second FDE start
622 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
623 second sub-section start.
625 2018-08-22 Richard Biener <rguenther@suse.de>
627 PR tree-optimization/86988
628 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
630 2018-08-22 Richard Biener <rguenther@suse.de>
632 PR tree-optimization/86945
633 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
635 2018-08-22 Alexandre Oliva <oliva@adacore.com>
637 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
638 a comment about how uses of r2 for .sdata2 come about.
640 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
642 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
644 2018-08-21 Marek Polacek <polacek@redhat.com>
646 PR c++/86981, Implement -Wpessimizing-move.
647 * doc/invoke.texi: Document -Wpessimizing-move.
649 2018-08-21 Jan Hubicka <jh@suse.cz>
651 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
652 * tree.h (is_redundant_typedef): Remove.
653 * dwarf2out.c (is_redundant_typedef): Turn into static function.
655 2018-08-21 Jan Hubicka <jh@suse.cz>
657 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
660 2018-08-21 Tamar Christina <tamar.christina@arm.com>
662 * expmed.c (extract_low_bits): Reject invalid subregs early.
664 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
667 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
670 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
672 * config/vxworks.h: Guard vxworks_asm_out_constructor and
673 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
674 * config/vxworks.c: Likewise.
676 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
678 * config/vxworks.c: Set targetm.have_ctors_dtors
679 if HAVE_INITFINI_ARRAY_SUPPORT.
680 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
681 if HAVE_INITFINI_ARRAY_SUPPORT.
683 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
685 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
686 default search path for VxWorks < 7.
688 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
690 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
691 (get_format_string): Refer to c_getstr.
693 2018-08-21 Tom de Vries <tdevries@suse.de>
695 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
696 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
697 (debuginfo_early_stop): Declare.
698 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
699 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
700 (debuginfo_early_stop): New function.
701 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
702 and debuginfo_early_stop.
703 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
704 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
705 (general_init): Call debuginfo_early_init.
706 (finalize): Call debuginfo_fini.
707 (do_compile): Call debuginfo_init.
708 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
710 (@item -fdump-debug, @item -fdump-earlydebug): Add.
712 2018-08-21 Tom de Vries <tdevries@suse.de>
714 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
715 flag_dump_noaddr and flag_dump_unnumbered.
717 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
719 * wide-int-range.cc (wide_int_range_abs): New.
720 (wide_int_range_order_set): Rename from wide_int_range_min_max.
721 * wide-int-range.h (wide_int_range_abs): New.
722 (wide_int_range_min_max): New.
723 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
724 case to call wide_int_range_abs.
725 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
726 (extract_range_from_abs_expr): Delete.
728 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
731 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
732 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
735 2018-08-20 Nathan Sidwell <nathan@acm.org>
736 Jeff Law <law@redhat.com>
738 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
739 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
741 2018-08-20 David Malcolm <dmalcolm@redhat.com>
744 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
745 (decl_attributes): Likewise.
746 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
748 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
749 * diagnostic-core.h (class auto_diagnostic_group): New class.
750 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
751 (diagnostic_report_diagnostic): Handle the first diagnostics within
753 (emit_diagnostic): Add auto_diagnostic_group instance.
755 (inform_n): Likewise.
757 (warning_at): Likewise.
758 (warning_n): Likewise.
760 (permerror): Likewise.
763 (error_at): Likewise.
765 (fatal_error): Likewise.
766 (internal_error): Likewise.
767 (internal_error_no_backtrace): Likewise.
768 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
769 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
770 * diagnostic.h (struct diagnostic_context): Add fields
771 "diagnostic_group_nesting_depth",
772 "diagnostic_group_emission_count", "begin_group_cb",
774 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
775 Add auto_diagnostic_group instance(s).
776 (find_explicit_erroneous_behavior): Likewise.
777 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
778 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
779 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
780 (gimplify_va_arg_expr): Likewise.
781 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
782 (HSA_SORRY_AT): Likewise.
783 * ipa-devirt.c (compare_virtual_tables): Likewise.
784 (warn_odr): Likewise.
785 * multiple_target.c (expand_target_clones): Likewise.
786 * opts-common.c (cmdline_handle_error): Likewise.
787 * reginfo.c (globalize_reg): Likewise.
788 * substring-locations.c (format_warning_n_va): Likewise.
789 * tree-inline.c (expand_call_inline): Likewise.
790 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
791 * tree-ssa-loop-niter.c
792 (do_warn_aggressive_loop_optimizations): Likewise.
793 * tree-ssa-uninit.c (warn_uninit): Likewise.
794 * tree.c (warn_deprecated_use): Likewise.
796 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
799 * config/i386/i386.md (eh_return): Always update EH return
800 address in word_mode.
802 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
804 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
805 TARGET_SPLIT_COMPLEX_ARG is defined.
807 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
809 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
811 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
814 * expr.c (expand_assignment): Assert that bitpos is positive.
815 (store_field): Likewise
816 (expand_expr_real_1): Make sure that bitpos is positive.
817 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
820 2018-08-20 Nathan Sidwell <nathan@acm.org>
822 * Makefile.in (CPP_ID_DATA_H): Delete.
823 (CPP_INTERNAL_H): Don't add it.
824 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
825 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
827 2018-08-20 Richard Biener <rguenther@suse.de>
829 PR tree-optimization/78655
830 * tree-vrp.c (extract_range_from_binary_expr_1): Make
831 pointer + offset nonnull if either operand is nonnull work.
833 2018-08-20 Tom de Vries <tdevries@suse.de>
835 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
836 unless the referenced die describes the added property using
837 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
838 Otherwise, add a DW_AT_location to the referenced die.
840 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
843 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
844 register_operand when calling ix86_set_reg_reg_cost.
845 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
846 Set *total to 0 for operands that satisfy x86_64_immediate_operand
847 predicate and to 1 otherwise.
849 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
851 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
852 emit a diagnostic that it is not supported and reset the option.
853 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
854 supported and consume the option. (ASM_FINAL_SPEC): New.
856 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
858 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
861 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
863 C-SKY port: Documentation
865 * doc/extend.texi (C-SKY Function Attributes): New section.
866 * doc/invoke.texi (Option Summary): Add C-SKY options.
867 (C-SKY Options): New section.
868 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
870 2018-08-17 Jojo <jijie_rong@c-sky.com>
871 Huibin Wang <huibin_wang@c-sky.com>
872 Sandra Loosemore <sandra@codesourcery.com>
873 Chung-Lin Tang <cltang@codesourcery.com>
875 C-SKY port: Backend implementation
877 * config/csky/*: New.
878 * common/config/csky/*: New.
880 2018-08-17 Jojo <jijie_rong@c-sky.com>
881 Huibin Wang <huibin_wang@c-sky.com>
882 Sandra Loosemore <sandra@codesourcery.com>
883 Chung-Lin Tang <cltang@codesourcery.com>
884 Andrew Jenner <andrew@codesourcery.com>
886 C-SKY port: Configury
888 * config.gcc (csky-*-*): New.
889 * configure.ac: Add csky to targets for dwarf2 debug_line support.
890 * configure: Regenerated.
892 2018-08-17 David Malcolm <dmalcolm@redhat.com>
894 * dump-context.h: Include "dumpfile.h".
895 (dump_context::dump_printf_va): Convert final param from va_list
896 to va_list *. Convert from ATTRIBUTE_PRINTF to
897 ATTRIBUTE_GCC_DUMP_PRINTF.
898 (dump_context::dump_printf_loc_va): Likewise.
899 * dumpfile.c: Include "stringpool.h".
900 (make_item_for_dump_printf_va): Delete.
901 (make_item_for_dump_printf): Delete.
902 (class dump_pretty_printer): New class.
903 (dump_pretty_printer::dump_pretty_printer): New ctor.
904 (dump_pretty_printer::emit_items): New member function.
905 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
907 (dump_pretty_printer::emit_item): New member function.
908 (dump_pretty_printer::stash_item): New member function.
909 (dump_pretty_printer::format_decoder_cb): New member function.
910 (dump_pretty_printer::decode_format): New member function.
911 (dump_context::dump_printf_va): Reimplement in terms of
913 (dump_context::dump_printf_loc_va): Convert final param from va_list
915 (dump_context::begin_scope): Reimplement call to
916 make_item_for_dump_printf.
917 (dump_printf): Update for change to dump_printf_va.
918 (dump_printf_loc): Likewise.
919 (selftest::test_capture_of_dump_calls): Convert "stmt" from
920 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
922 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
923 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
924 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
925 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
926 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
927 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
928 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
929 within a dump_printf_loc call to "%wu".
930 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
931 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
932 missing space after "=".
933 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
934 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
935 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
936 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
937 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
938 duplicate "vectorized" from message.
940 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
942 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
943 polyNxK_t element's TYPE_STRING_FLAG.
945 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
947 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
948 (they were unnamed before). Fix comments.
950 2018-08-17 Nathan Sidwell <nathan@acm.org>
952 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
954 2018-08-17 Richard Biener <rguenther@suse.de>
956 PR tree-optimization/86841
957 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
959 2018-08-17 Martin Liska <mliska@suse.cz>
961 * common.opt: Remove Warn, Init and Report for options with
962 Ignore/Deprecated flag. Warning is done automatically for
964 * config/i386/i386.opt: Likewise.
965 * config/ia64/ia64.opt: Likewise.
966 * config/rs6000/rs6000.opt: Likewise.
967 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
968 Remove usage of flag_check_pointer_bounds.
969 * lto-wrapper.c (merge_and_complain): Do not handle
970 OPT_fcheck_pointer_bounds.
971 (append_compiler_options): Likewise.
972 * opt-functions.awk: Do not handle Deprecated.
973 * optc-gen.awk: Check that Var, Report and Init are not
974 used for an option with Ignore/Deprecated flag.
975 * opts-common.c (decode_cmdline_option): Do not report
977 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
979 * opts.h (struct cl_option): Remove cl_deprecated flag.
980 (CL_ERR_DEPRECATED): Remove error enum value.
982 2018-08-17 Richard Biener <rguenther@suse.de>
985 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
986 across a va-arg-pack using call adjust its return value accordingly.
988 2018-08-16 Martin Sebor <msebor@redhat.com>
990 PR tree-optimization/86853
991 * gimple-ssa-sprintf.c (struct format_result): Rename member.
992 (struct fmtresult): Add member and initialize it in ctors.
993 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
994 (format_string): Handle %S the same as %ls. Set MAYFAIL.
995 (format_directive): Set POSUNDER4K when MAYFAIL is set.
996 (parse_directive): Handle %C same as %c.
997 (sprintf_dom_walker::compute_format_length): Adjust.
998 (is_call_safe): Adjust.
1000 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1002 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1003 for determining how to count the elements.
1004 * builtins.h (c_strlen): Adjust prototype.
1005 * expr.c (string_constant): Add new parameter mem_size.
1006 Set *mem_size appropriately.
1007 * expr.h (string_constant): Adjust protoype.
1008 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1009 * gimple-fold.h (get_range_strlen): Adjust prototype.
1010 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1011 (format_string): Call get_string_length with eltsize.
1013 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1015 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1016 to emit the span, rather than setting it as the prefix.
1018 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1020 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1021 "margin_char" parameter, defaulting to space. Use it in place
1022 of pp_space for the initial part of the margin.
1023 (layout::print_leading_fixits): Use '+' when filling the margin
1024 of line-insertion fix-it hints.
1026 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1028 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1031 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1033 * config/rs6000/altivec.md: Don't set length attribute to the default
1035 * config/rs6000/darwin.md: Ditto.
1036 * config/rs6000/dfp.md: Ditto.
1037 * config/rs6000/htm.md: Ditto.
1038 * config/rs6000/rs6000.md: Ditto.
1039 * config/rs6000/sync.md: Ditto.
1040 * config/rs6000/vsx.md: Ditto.
1042 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1044 * config/rs6000/altivec.md: Don't set length attribute to the default
1045 value, for branch instructions.
1046 * config/rs6000/darwin.md: Ditto.
1047 * config/rs6000/rs6000.md: Ditto.
1049 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1051 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1052 (unnamed conditional branch define_insn): Set length to 4 or 8
1053 depending on offset.
1054 (<bd>_<mode>): Similar, for alternative 0.
1055 (<bd>tf_<mode>): Ditto.
1057 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1059 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1061 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1063 * doc/rtl.texi: Replace old RTX class names with new names.
1066 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1068 * expmed.h (canonicalize_comparison): New declaration.
1069 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1070 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1071 * optabs.c (prepare_cmp_insn): Likewise.
1072 * rtl.h (unsigned_condition_p): New function which checks if a
1073 comparison operator is unsigned.
1075 2018-08-16 Nathan Sidwell <nathan@acm.org>
1077 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1078 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1080 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1083 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1084 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1085 (mov<mov>): ..this and enable unconditionally.
1087 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1089 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1091 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1093 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1094 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1095 "Common" with "Target".
1097 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1099 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1100 * doc/invoke.texi (mmitigate-rop): Remove.
1101 * config/i386/i386.c: Do not include "regrename.h".
1102 (ix86_rop_should_change_byte_p, reg_encoded_number)
1103 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1105 (ix86_reorg): Remove call to ix86_mitigate_rop.
1106 * config/i386/i386.md (attr "modrm_class"): Remove.
1107 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1108 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1109 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1111 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1113 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1114 allow folding of mergeh() and mergel() for the float and double types.
1115 (fold_mergehl_helper): Rework to handle building a permute tree
1118 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1120 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1123 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1125 * common.opt (fdiagnostics-show-labels): New option.
1126 * diagnostic-show-locus.c (class layout_range): Add field
1128 (class layout): Add field "m_show_labels_p".
1129 (layout_range::layout_range): Add param "label" and use it to
1131 (make_range): Pass in NULL for new "label" param of layout_range's
1133 (layout::layout): Initialize m_show_labels_p.
1134 (layout::maybe_add_location_range): Pass in loc_range->m_label
1135 when constructing layout_range instances.
1136 (struct line_label): New struct.
1137 (layout::print_any_labels): New member function.
1138 (layout::print_line): Call it if label-printing is enabled.
1139 (selftest::test_one_liner_labels): New test.
1140 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1141 * diagnostic.c (diagnostic_initialize): Initialize
1142 context->show_labels_p.
1143 * diagnostic.h (struct diagnostic_context): Add field
1145 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1146 -fno-diagnostics-show-labels.
1147 * dwarf2out.c (gen_producer_string): Add
1148 OPT_fdiagnostics_show_labels to the ignored options.
1149 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1151 (gcc_rich_location::maybe_add_expr): Likewise.
1152 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1153 label" param, defaulting to NULL.
1154 (gcc_rich_location::add_expr): Add "label" param.
1155 (gcc_rich_location::maybe_add_expr): Likewise.
1156 (class text_range_label): New class.
1157 (class range_label_for_type_mismatch): New class.
1158 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1159 of format_warning_va.
1160 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1161 * lto-wrapper.c (merge_and_complain): Add
1162 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1163 (append_compiler_options): Likewise to the dropped options.
1164 (append_diag_options): Likewise to the passed-on options.
1165 * opts.c (common_handle_option): Handle the new option.
1166 * selftest-diagnostic.c
1167 (test_diagnostic_context::test_diagnostic_context): Enable
1169 * substring-locations.c: Include "gcc-rich-location.h".
1170 (format_warning_n_va): Add "fmt_label" and "param_label" params
1171 and use them as appropriate.
1172 (format_warning_va): Add "fmt_label" and "param_label" params,
1173 passing them on to format_warning_n_va.
1174 (format_warning_at_substring): Likewise.
1175 (format_warning_at_substring_n): Likewise.
1176 * substring-locations.h (format_warning_va): Add "fmt_label" and
1177 "param_label" params.
1178 (format_warning_n_va): Likewise.
1179 (format_warning_at_substring): Likewise.
1180 (format_warning_at_substring_n): Likewise.
1181 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1183 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1186 * builtins.c (expand_builtin_memcmp): Do not expand the call
1187 when overflow is detected.
1189 2018-08-15 Martin Sebor <msebor@redhat.com>
1191 PR tree-optimization/71625
1192 * config/aarch64/aarch64-builtins.c
1193 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
1195 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
1197 * config/s390/s390.c (s390_reorg): Remove loop.
1199 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1202 (darwin_function_switched_text_sections): Delete.
1203 * gcc/config/darwin.h
1204 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
1206 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1209 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1210 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1211 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1213 2018-08-15 Martin Liska <mliska@suse.cz>
1215 PR tree-optimization/86925
1216 * predict.c (expr_expected_value_1): When taking
1217 later predictor, assign also probability.
1218 Use fold_build2_initializer_loc in order to fold
1219 the expression in -frounding-math.
1221 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
1223 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
1225 (expand_vec_perm_1): Try the new method.
1227 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
1230 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
1231 Check whether lra_live_max_point is 0 before dividing.
1233 2018-08-14 Martin Sebor <msebor@redhat.com>
1235 PR tree-optimization/86650
1236 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
1237 (vrp_prop::check_mem_ref): Same.
1239 2018-08-13 Liu Hao <lh_mouse@126.com>
1241 * pretty-print.c (eat_esc_sequence): Swap the foreground and
1242 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
1243 and clear it thereafter, as it only works for DBCS.
1245 2018-08-13 Liu Hao <lh_mouse@126.com>
1247 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
1248 handle returned by _get_osf_handle().
1250 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1252 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
1253 for folding vec_perm.
1255 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1257 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
1258 Add support for gimple-folding of vec_pack() and vec_unpack()
1261 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1263 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
1264 vec_xst variants to the list.
1265 (rs6000_gimple_fold_builtin): Add support for folding unaligned
1266 vector loads and stores.
1268 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
1270 * config.gcc (rs6000-ibm-aix4.x): Delete.
1271 (rs6000-ibm-aix5.1): Delete.
1272 (rs6000-ibm-aix5.2): Delete.
1273 (rs6000-ibm-aix5.3): Delete.
1274 * config/rs6000/aix43.h: Delete.
1275 * config/rs6000/aix51.h: Delete.
1276 * config/rs6000/aix52.h: Delete.
1277 * config/rs6000/t-aix43: Delete.
1279 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
1281 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1283 (s390_decompose_address): Factor out constant pool ref
1286 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1288 * config/nds32/nds32-predicates.c
1289 (nds32_can_use_bclr_p): Change return type as bool.
1290 (nds32_can_use_bset_p): Ditto.
1291 (nds32_can_use_btgl_p): Ditto.
1292 (nds32_can_use_bitci_p): Ditto.
1293 * config/nds32/nds32-protos.h
1294 (nds32_can_use_bclr_p): Change declaration.
1295 (nds32_can_use_bset_p): Ditto.
1296 (nds32_can_use_btgl_p): Ditto.
1297 (nds32_can_use_bitci_p): Ditto.
1299 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1301 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
1302 Support -msched-prolog-epilog option.
1303 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
1305 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1307 * common/config/nds32/nds32-common.c
1308 (nds32_option_optimization_table): Enalbe -malways-align.
1310 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1312 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
1314 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
1315 OPT_misr_secure_ case.
1316 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
1317 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
1318 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
1320 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
1321 (nds32_isr_info): New field security_level.
1322 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
1323 * config/nds32/nds32.md (return_internal): Consider critical attribute.
1324 * config/nds32/nds32.opt (misr-secure): New option.
1325 * config/nds32/nds32_init.inc: New file.
1326 * config/nds32/nds32_isr.h: New file.
1328 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
1330 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
1331 Update comment for atomic instructions.
1332 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
1333 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
1335 (atomic_loaddi): Revise fence expansion to only emit fence prior to
1336 load for __ATOMIC_SEQ_CST model.
1337 (atomic_loaddi_1): Remove float register target.
1338 (atomic_storedi): Handle CONST_INT values.
1339 (atomic_storedi_1): Remove float register source. Add special case
1341 (memory_barrier): New expander and insn.
1343 2018-08-11 Jakub Jelinek <jakub@redhat.com>
1345 PR tree-optimization/86835
1346 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
1347 new_stmt after def_gsi, make sure to insert new_square_stmt after
1348 that stmt, not 2 stmts before it.
1350 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
1353 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
1356 2018-08-10 Martin Liska <mliska@suse.cz>
1359 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
1361 * builtins.c (expand_builtin_expect_with_probability):
1363 (expand_builtin_expect_with_probability): New function.
1364 (build_builtin_expect_predicate): Add new argumnet probability
1365 for BUILT_IN_EXPECT_WITH_PROBABILITY.
1366 (fold_builtin_expect):
1369 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
1370 * builtins.h (fold_builtin_expect): Set new argument.
1371 * doc/extend.texi: Document __builtin_expect_with_probability.
1372 * doc/invoke.texi: Likewise.
1373 * gimple-fold.c (gimple_fold_call): Pass new argument.
1374 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
1375 also BUILT_IN_EXPECT_WITH_PROBABILITY.
1376 * predict.c (get_predictor_value): New function.
1377 (expr_expected_value): Add new argument probability. Assume
1378 that predictor and probability are always non-null.
1379 (expr_expected_value_1): Likewise. For __builtin_expect and
1380 __builtin_expect_with_probability set probability. Handle
1381 combination in binary expressions.
1382 (tree_predict_by_opcode): Simplify code by simply calling
1383 get_predictor_value.
1384 (pass_strip_predict_hints::execute): Add handling of
1385 BUILT_IN_EXPECT_WITH_PROBABILITY.
1386 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
1388 * tree.h (DECL_BUILT_IN_P): New function.
1390 2018-08-10 Martin Liska <mliska@suse.cz>
1392 PR tree-optimization/85799
1393 * passes.def: Add argument for pass_strip_predict_hints.
1394 * predict.c (class pass_strip_predict_hints): Add new argument
1396 (strip_predictor_early): New function.
1397 (pass_strip_predict_hints::execute): Call the function to
1399 (strip_predict_hints): New function.
1400 * predict.def: Fix comment.
1402 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
1404 * Makefile.in: Clarify which tm.texi to copy over to assert the
1405 right to grant a GFDL license for all.
1407 2018-08-09 Jeff Law <law@redhat.com>
1409 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
1412 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
1414 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
1417 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1419 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
1420 reductions for variable-length vectors.
1422 2018-08-09 David Malcolm <dmalcolm@redhat.com>
1425 * common.opt (fdiagnostics-show-line-numbers): New option.
1426 * diagnostic-show-locus.c (class layout): Add fields
1427 "m_show_line_numbers_p" and "m_linenum_width";
1428 (num_digits): New function.
1429 (test_num_digits): New function.
1430 (layout::layout): Initialize new fields. Update m_x_offset
1431 logic to handle any left margin.
1432 (layout::print_source_line): Print line number when requested.
1433 (layout::start_annotation_line): New member function.
1434 (layout::print_annotation_line): Call it.
1435 (layout::print_leading_fixits): Likewise.
1436 (layout::print_trailing_fixits): Likewise. Update calls to
1437 move_to_column for new parameter.
1438 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
1439 it to potentially call start_annotation_line.
1440 (layout::show_ruler): Call start_annotation_line.
1441 (selftest::test_line_numbers_multiline_range): New selftest.
1442 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
1443 and selftest::test_line_numbers_multiline_range.
1444 * diagnostic.c (diagnostic_initialize): Initialize
1445 show_line_numbers_p.
1446 * diagnostic.h (struct diagnostic_context): Add field
1447 "show_line_numbers_p".
1448 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1449 -fno-diagnostics-show-line-numbers.
1450 * dwarf2out.c (gen_producer_string): Add
1451 OPT_fdiagnostics_show_line_numbers to the ignored options.
1452 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
1453 one setting" options.
1454 (append_compiler_options): Likewise to the dropped options.
1455 (append_diag_options): Likewise to the passed-on options.
1456 * opts.c (common_handle_option): Handle the new option.
1457 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
1459 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
1461 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
1462 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
1463 third argument of type "const signed char" to descriptions of
1464 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
1465 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
1466 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
1467 __builtin_bcdsub_ov functions.
1469 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1471 PR tree-optimization/86858
1472 * tree-vect-loop.c (vect_is_simple_reduction): Restore
1473 flow_bb_inside_loop_p calls.
1475 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1477 PR tree-optimization/86871
1478 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
1479 instead of gimple_assign_lhs.
1481 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
1484 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
1485 register constraint to operand 0.
1486 (add<mode>3_carryinC): Likewise.
1487 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
1489 2018-08-09 Martin Liska <mliska@suse.cz>
1492 * common.opt: Remove extra line.
1494 2018-08-09 Martin Liska <mliska@suse.cz>
1496 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
1497 at the end of a line, make first letter capital and end up
1498 a sentence with a dot.
1499 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
1500 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
1501 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
1502 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
1503 (PARAM_MAX_ISL_OPERATIONS): Likewise.
1504 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
1505 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
1506 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
1507 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
1508 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
1509 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
1510 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
1511 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
1512 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
1513 (PARAM_TREE_REASSOC_WIDTH): Likewise.
1514 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
1515 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
1516 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
1518 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
1521 * config/s390/s390.c (s390_option_override_internal): Reduce the
1522 stack-clash-protection-probe-interval param if it would be too big
1525 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
1528 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
1529 * config/m68k/m68k.c (handle_move_double): Don't call
1530 m68k_final_prescan_insn.
1531 (m68k_adjust_decorated_operand): Renamed from
1532 m68k_final_prescan_insn, remove first and third operand and
1534 (print_operand): Call it.
1535 (print_operand_address): Call it.
1537 2018-08-08 Nathan Sidwell <nathan@acm.org>
1539 * diagnostic.c (diagnostic_report_current_module): Use
1540 linemap_included_from & linemap_included_from_linemap.
1542 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
1544 * config/aarch64/aarch64-cores.def: Add phecda core.
1545 * config/aarch64/aarch64-tune.md: Regenerate.
1546 * doc/invoke.texi: Add phecda core.
1548 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1551 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
1553 * config/s390/s390.md ("movti"): Add more alternatives for
1554 constant to GPR copies.
1556 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1558 * config/s390/s390.c: Fix whitespace damage throughout the file.
1559 * config/s390/s390.h: Likewise.
1560 * config/s390/tpf.h: Likewise.
1562 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1564 * config/s390/s390.c (s390_loadrelative_operand_p):
1565 Remove TARGET_CPU_ZARCH usages.
1566 (s390_rtx_costs): Likewise.
1567 (s390_legitimate_constant_p): Likewise.
1568 (s390_cannot_force_const_mem): Likewise.
1569 (legitimate_reload_constant_p): Likewise.
1570 (s390_preferred_reload_class): Likewise.
1571 (legitimize_pic_address): Likewise.
1572 (legitimize_tls_address): Likewise.
1573 (s390_split_branches): Removed.
1574 (s390_add_execute): Removed.
1575 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
1576 (s390_mainpool_start): Likewise.
1577 (s390_mainpool_finish): Likewise.
1578 (s390_mainpool_cancel): Removed.
1579 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
1580 (s390_chunkify_cancel): Likewise.
1581 (s390_return_addr_rtx): Likewise.
1582 (s390_register_info): Remove split_branches_pending_p uages.
1583 (s390_optimize_register_info): Likewise.
1584 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
1585 split_branches_pending_p usages.
1586 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
1587 (s390_load_got): Likewise.
1588 (s390_expand_split_stack_prologue): Likewise.
1589 (output_asm_nops): Likewise.
1590 (s390_function_profiler): Likewise.
1591 (s390_emit_call): Likewise.
1592 (s390_conditional_register_usage): Likewise.
1593 (s390_optimize_prologue): Likewise.
1594 (s390_reorg): Remove TARGET_CPU_ZARCH and
1595 split_branches_pending_p usages.
1596 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
1598 (s390_output_indirect_thunk_function): Likewise.
1599 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
1600 (TARGET_CPU_ZARCH_P): Removed.
1601 (struct machine_function): Remove split_branches_pending_p.
1602 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
1604 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1606 * common/config/s390/s390-common.c (processor_flags_table):
1608 * config.gcc: Remove with_arch/with_tune support.
1609 * config/s390/2064.md: Remove cpu attribute comparisons.
1610 * config/s390/driver-native.c (s390_host_detect_local_cpu):
1612 * config/s390/linux.h (ASM_SPEC):
1613 Remove -march support.
1614 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
1615 Use a table to get an arch level.
1616 * config/s390/s390-opts.h (enum processor_type):
1618 * config/s390/s390.c
1619 (processor_table): Remove entries, add arch_level values.
1620 (s390_issue_rate): Remove cases.
1621 (s390_option_override): Adjust
1622 s390_option_override_internal() call.
1623 (s390_option_override_internal): Remove deprecation warning.
1624 (s390_valid_target_attribute_tree): Adjust
1625 s390_option_override_internal() call.
1626 * config/s390/s390.h (struct s390_processor):
1627 Share with s390-c.c, add arch_level field.
1628 * config/s390/s390.md:
1629 Remove occurrences in cpu attribute.
1630 * config/s390/s390.opt: Remove -march/-mtune support.
1631 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
1632 * doc/invoke.texi: Remove deprecation warning.
1634 2018-08-08 Luis Machado <luis.machado@linaro.org>
1636 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
1638 (qdf24xx_tunings): Set vector cost structure to
1639 qdf24xx_vector_cost.
1641 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
1642 <register_sextend>: Set to 3.
1644 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
1647 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
1648 * config/aarch64/aarch64-simd.md
1649 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
1650 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
1651 (aarch64_frecpx<mode>): ...this new pattern.
1652 * config/aarch64/aarch64-simd-builtins.def: Remove comment
1653 about aarch64_frecp<FRECP:frecp_suffix><mode>.
1655 2018-08-07 Martin Liska <mliska@suse.cz>
1658 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
1659 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
1660 * predict.def (PRED_MALLOC_NONNULL): New predictor.
1661 * doc/extend.texi: Document that malloc attribute adds
1664 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
1667 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1668 Define to speculation_safe_value_not_needed.
1670 2018-08-06 Jeff Law <law@redhat.com>
1672 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
1673 the vr_values instance to cprop_into_stmt.
1674 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
1675 (cprop_operand): Also query EVRP to determine if OP is a constant.
1677 2018-08-06 Nathan Sidwell <nathan@acm.org>
1679 * diagnostic.c (diagnostic_report_current_module): Reroll
1680 included-at loop. Translate text.
1682 2018-08-06 David Malcolm <dmalcolm@redhat.com>
1684 * function-tests.c (selftest::test_expansion_to_rtl): Call
1685 free_after_compilation.
1687 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1689 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
1691 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
1693 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
1694 loops with memory block operations from getting unrolled.
1696 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
1699 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1700 Define to speculation_safe_value_not_needed.
1702 2018-08-06 Jeff Law <law@redhat.com>
1704 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
1707 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1710 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
1711 with all enabled __intN types.
1713 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
1715 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1717 * alias.c (record_set): Check for clobber high.
1718 * cfgexpand.c (expand_gimple_stmt): Likewise.
1719 * combine-stack-adj.c (single_set_for_csa): Likewise.
1720 * combine.c (find_single_use_1): Likewise.
1721 (set_nonzero_bits_and_sign_copies): Likewise.
1722 (get_combine_src_dest): Likewise.
1723 (is_parallel_of_n_reg_sets): Likewise.
1724 (try_combine): Likewise.
1725 (record_dead_and_set_regs_1): Likewise.
1726 (reg_dead_at_p_1): Likewise.
1727 (reg_dead_at_p): Likewise.
1728 * dce.c (deletable_insn_p): Likewise.
1729 (mark_nonreg_stores_1): Likewise.
1730 (mark_nonreg_stores_2): Likewise.
1731 * df-scan.c (df_find_hard_reg_defs): Likewise.
1732 (df_uses_record): Likewise.
1733 (df_get_call_refs): Likewise.
1734 * dwarf2out.c (mem_loc_descriptor): Likewise.
1735 * haifa-sched.c (haifa_classify_rtx): Likewise.
1736 * ira-build.c (create_insn_allocnos): Likewise.
1737 * ira-costs.c (scan_one_insn): Likewise.
1738 * ira.c (equiv_init_movable_p): Likewise.
1739 (rtx_moveable_p): Likewise.
1740 (interesting_dest_for_shprep): Likewise.
1741 * jump.c (mark_jump_label_1): Likewise.
1742 * postreload-gcse.c (record_opr_changes): Likewise.
1743 * postreload.c (reload_cse_simplify): Likewise.
1744 (struct reg_use): Add source expr.
1745 (reload_combine): Check for clobber high.
1746 (reload_combine_note_use): Likewise.
1747 (reload_cse_move2add): Likewise.
1748 (move2add_note_store): Likewise.
1749 * print-rtl.c (print_pattern): Likewise.
1750 * recog.c (decode_asm_operands): Likewise.
1751 (store_data_bypass_p): Likewise.
1752 (if_test_bypass_p): Likewise.
1753 * regcprop.c (kill_clobbered_value): Likewise.
1754 (kill_set_value): Likewise.
1755 * reginfo.c (reg_scan_mark_refs): Likewise.
1756 * reload1.c (maybe_fix_stack_asms): Likewise.
1757 (eliminate_regs_1): Likewise.
1758 (elimination_effects): Likewise.
1759 (mark_not_eliminable): Likewise.
1760 (scan_paradoxical_subregs): Likewise.
1761 (forget_old_reloads_1): Likewise.
1762 * reorg.c (find_end_label): Likewise.
1763 (try_merge_delay_insns): Likewise.
1764 (redundant_insn): Likewise.
1765 (own_thread_p): Likewise.
1766 (fill_simple_delay_slots): Likewise.
1767 (fill_slots_from_thread): Likewise.
1768 (dbr_schedule): Likewise.
1769 * resource.c (update_live_status): Likewise.
1770 (mark_referenced_resources): Likewise.
1771 (mark_set_resources): Likewise.
1772 * rtl.c (copy_rtx): Likewise.
1773 * rtlanal.c (reg_referenced_p): Likewise.
1774 (single_set_2): Likewise.
1775 (noop_move_p): Likewise.
1776 (note_stores): Likewise.
1777 * sched-deps.c (sched_analyze_reg): Likewise.
1778 (sched_analyze_insn): Likewise.
1780 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1782 * cse.c (invalidate_reg): New function extracted from...
1783 (invalidate): ...here.
1784 (canonicalize_insn): Check for clobber high.
1785 (invalidate_from_clobbers): invalidate clobber highs.
1786 (invalidate_from_sets_and_clobbers): Likewise.
1787 (count_reg_usage): Check for clobber high.
1788 (insn_live_p): Likewise.
1789 * cselib.c (cselib_expand_value_rtx_1):Likewise.
1790 (cselib_invalidate_regno): Check for clobber in setter.
1791 (cselib_invalidate_rtx): Pass through setter.
1792 (cselib_invalidate_rtx_note_stores):
1793 (cselib_process_insn): Check for clobber high.
1794 * cselib.h (cselib_invalidate_rtx): Add operand.
1796 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1798 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
1799 (mark_not_eliminable): Likewise.
1800 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
1801 * lra-lives.c (process_bb_lives): Check for clobber high.
1802 * lra.c (new_insn_reg): Remember clobber highs.
1803 (collect_non_operand_hard_regs): Check for clobber high.
1804 (lra_set_insn_recog_data): Likewise.
1805 (add_regs_to_insn_regno_info): Likewise.
1806 (lra_update_insn_regno_info): Likewise.
1808 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1810 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
1811 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
1813 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1815 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
1816 (copy_insn_1): Likewise.
1817 (gen_hard_reg_clobber_high): New gen function.
1818 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
1819 * genemit.c (gen_exp): Likewise.
1820 (gen_emit_seq): Pass through info.
1821 (gen_insn): Check for CLOBBER_HIGH.
1822 (gen_expand): Pass through info.
1823 (gen_split): Likewise.
1824 (output_add_clobbers): Likewise.
1825 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
1826 (remove_clobbers): Likewise.
1827 * rtl.h (gen_hard_reg_clobber_high): New declaration.
1829 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1831 * doc/rtl.texi (clobber_high): Add.
1832 (parallel): Add in clobber high
1833 * rtl.c (rtl_check_failed_code3): Add function.
1834 * rtl.def (CLOBBER_HIGH): Add expression.
1835 * rtl.h (RTL_CHECKC3): Add macro.
1836 (rtl_check_failed_code3): Add declaration.
1837 (XC3EXP): Add macro.
1839 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
1842 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
1843 cfun->machine->max_used_stack_alignment if needed.
1845 2018-08-04 Martin Sebor <msebor@redhat.com>
1847 PR tree-optimization/86571
1848 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
1851 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
1854 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1857 2018-08-03 Jeff Law <law@redhat.com>
1860 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1861 Define to speculation_safe_value_not_needed.
1863 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1865 * doc/gcov.texi (-x): Remove duplicate "to".
1866 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1867 (-Wif-not-aligned): Remove duplicate "is".
1868 (-flto): Remove duplicate "the".
1869 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1870 duplicate "v5.00.b".
1871 (MSP430 Options): Remove duplicate "and" from the description
1872 of "-mgprel-sec=regexp".
1873 (x86 Options): Remove duplicate copies of "vmldLog102" and
1874 vmlsLog104 from description of "-mveclibabi=type".
1876 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
1878 * internal-fn.h (first_commutative_argument): Declare.
1879 * internal-fn.c (first_commutative_argument): New function.
1880 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
1881 restrictions for pattern statements. Use first_commutative_argument
1882 to look for commutative operands in calls to internal functions.
1884 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
1886 * Makefile.in (wide-int-range.o): New.
1887 * tree-vrp.c: Move all the wide_int_* functions to...
1888 * wide-int-range.cc: ...here.
1889 * tree-vrp.h: Move all the wide_int_* prototypes to...
1890 * wide-int-range.h: ...here.
1892 2018-08-03 Tom de Vries <tdevries@suse.de>
1894 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
1896 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
1897 * except.c (output_function_exception_table): Do early exit if
1898 targetm_common.except_unwind_info (&global_options) == UI_NONE.
1900 2018-08-03 Martin Liska <mliska@suse.cz>
1902 * predict.c (dump_prediction): Change to 2 digits
1905 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
1907 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
1909 (falkor_am_1_gtov_gtov): ... a new insn reservation.
1911 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
1913 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
1914 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
1915 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
1916 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
1917 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
1919 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1921 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1922 pp_take_prefix when saving the existing prefix.
1923 * diagnostic.c (diagnostic_append_note): Likewise.
1924 * langhooks.c (lhd_print_error_function): Likewise.
1925 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1926 param's type. Free the existing prefix.
1927 (pp_take_prefix): New function.
1928 (pretty_printer::pretty_printer): Drop the prefix parameter.
1929 Rename the length parameter to match the comment.
1930 (pretty_printer::~pretty_printer): Free the prefix.
1931 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1933 (struct pretty_printer): Drop the "const" from "prefix" field's
1934 type and clarify memory management.
1935 (pp_set_prefix): Drop the "const" from the 2nd param.
1936 (pp_take_prefix): New decl.
1938 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1940 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
1941 for word_mode_ok here instead of passing as argument.
1942 (expand_block_compare): Change select_block_compare_mode() call.
1943 (expand_strncmp_gpr_sequence): New function.
1944 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
1946 2018-08-02 Jeff Law <law@redhat.com>
1949 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1950 Define to speculation_safe_value_not_needed.
1953 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1954 Define to speculation_safe_value_not_needed.
1956 2018-08-02 Tom de Vries <tdevries@suse.de>
1959 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1960 function. Return UI_TARGET unconditionally.
1961 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1962 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1964 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1966 * genemit.c (print_overload_test): Fix typo.
1968 2018-08-02 Richard Biener <rguenther@suse.de>
1970 PR tree-optimization/86816
1971 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1972 which checks for value availability before querying it.
1973 (gvn_uses_equal): Use it.
1974 (same_succ_hash): Likewise.
1975 (gimple_equal_p): Likewise.
1977 2018-08-02 Nick Clifton <nickc@redhat.com>
1980 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1981 Define to speculation_safe_value_not_needed.
1984 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1985 Define to speculation_safe_value_not_needed.
1988 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1989 Define to speculation_safe_value_not_needed.
1992 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1993 Define to speculation_safe_value_not_needed.
1996 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1997 Define to speculation_safe_value_not_needed.
2000 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2001 Define to speculation_safe_value_not_needed.
2004 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2005 Define to speculation_safe_value_not_needed.
2008 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2009 Define to speculation_safe_value_not_needed.
2012 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2013 speculation_safe_value_not_needed.
2016 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2017 to speculation_safe_value_not_needed.
2019 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2021 * doc/md.texi: Expand the documentation of instruction names
2022 to mention port-local uses. Document '@' in pattern names.
2023 * read-md.h (overloaded_instance, overloaded_name): New structs.
2025 (md_reader::handle_overloaded_name): New member function.
2026 (md_reader::get_overloads): Likewise.
2027 (md_reader::m_first_overload): New member variable.
2028 (md_reader::m_next_overload_ptr): Likewise.
2029 (md_reader::m_overloads_htab): Likewise.
2030 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2031 m_next_overload_ptr and m_overloads_htab.
2032 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2033 (get_mode_token, get_code_token, get_int_token): New functions.
2034 (map_attr_string): Add an optional argument that passes back
2035 the associated iterator.
2036 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2037 (md_reader::handle_overloaded_name, add_overload_instance): New
2039 (apply_iterators): Handle '@' names. Report an error if '@'
2040 is used without iterators.
2041 (initialize_iterators): Initialize the new iterator_group fields.
2042 * genopinit.c (handle_overloaded_code_for)
2043 (handle_overloaded_gen): New functions.
2044 (main): Use them to print declarations of maybe_code_for_* and
2045 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2046 * genemit.c (print_overload_arguments, print_overload_test)
2047 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2048 (main): Use it to print definitions of maybe_code_for_* and
2049 maybe_gen_* functions.
2050 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2051 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2052 instead of explicit mode checks.
2053 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2054 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2055 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2056 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2057 (aarch64_expand_compare_and_swap): Likewise
2058 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2059 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2060 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2061 (aarch64_constant_pool_reload_icode): Delete.
2062 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2063 instead of aarch64_constant_pool_reload_icode. Use
2064 code_for_aarch64_reload_mov instead of explicit mode checks.
2065 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2066 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2067 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2068 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2069 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2070 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2071 (aarch64_atomic_load_op_code): Delete.
2072 (aarch64_emit_atomic_load_op): Likewise.
2073 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2074 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2075 instead of aarch64_emit_atomic_load_op.
2076 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2077 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2078 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2079 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2080 character before the pattern name.
2081 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2082 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2083 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2084 (aarch64_frecps<mode>): Likewise.
2085 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2086 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2087 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2088 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2089 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2091 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2093 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2094 Allow HFmode constants if TARGET_FP_F16INST.
2096 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2099 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2100 No longer check last store for clobber of address register.
2102 2018-08-02 Martin Liska <mliska@suse.cz>
2104 PR gcov-profile/86817
2105 * gcov.c (process_all_functions): New function.
2107 (process_file): Move functions processing to
2108 process_all_functions.
2110 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2112 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2113 "const" to the "gimple *" and "rtx_insn *" parameters.
2114 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2116 (dump_location_t::dump_location_t): Likewise.
2118 2018-08-01 Martin Sebor <msebor@redhat.com>
2120 PR tree-optimization/86650
2121 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2122 rather than a "gcall *". Directly pass the data of interest
2123 to percent_K_format, rather than building a temporary CALL_EXPR
2125 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2126 (gimple_fold_builtin_strncat): Adjust.
2127 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2128 gcall* argument with gimple*.
2129 * gimple-ssa-warn-restrict.c (check_call): Same.
2130 (wrestrict_dom_walker::before_dom_children): Same.
2131 (builtin_access::builtin_access): Same.
2132 (check_bounds_or_overlap): Same
2133 (maybe_diag_overlap): Same.
2134 (maybe_diag_offset_bounds): Same.
2135 * tree-diagnostic.c (default_tree_printer): Move usage of
2136 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2138 * tree-pretty-print.c (percent_K_format): Add argument.
2139 * tree-pretty-print.h: Add argument.
2140 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2142 (maybe_diag_stxncpy_trunc): Same.
2143 (handle_builtin_stxncpy): Same.
2144 (handle_builtin_strcat): Same.
2146 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2148 * match.pd: Optimise pointer range checks.
2150 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2152 PR tree-optimization/86758
2153 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2154 to remove pattern statements.
2156 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2158 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2159 result of dfs_enumerate_from when constructing stmt_vec_infos,
2160 instead of additionally calling get_loop_body.
2162 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2164 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2166 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2167 When creating an iv, assert that the step is not known to be zero.
2168 (vect_setup_realignment): Update call accordingly.
2169 * tree-vect-stmts.c (vectorizable_store): Likewise.
2170 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2172 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2174 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2175 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2176 (vectorizable_reduction): Likewise.
2177 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2178 (vect_detect_hybrid_slp_stmts): Likewise.
2179 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2181 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2183 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2184 (wide_int_set_zero_nonzero_bits): ...this.
2185 (zero_nonzero_bits_from_vr): Rename to...
2186 (vrp_set_zero_nonzero_bits): ...this.
2187 (extract_range_from_multiplicative_op_1): Abstract wide int
2189 (wide_int_range_multiplicative_op): ...here.
2190 (extract_range_from_binary_expr_1): Extract wide int binary
2191 operations into their own functions.
2192 (wide_int_range_lshift): New.
2193 (wide_int_range_can_optimize_bit_op): New.
2194 (wide_int_range_shift_undefined_p): New.
2195 (wide_int_range_bit_xor): New.
2196 (wide_int_range_bit_ior): New.
2197 (wide_int_range_bit_and): New.
2198 (wide_int_range_trunc_mod): New.
2199 (extract_range_into_wide_ints): New.
2200 (vrp_shift_undefined_p): New.
2201 (extract_range_from_multiplicative_op): New.
2202 (vrp_can_optimize_bit_op): New.
2203 * tree-vrp.h (value_range::dump): New.
2204 (wide_int_range_multiplicative_op): New.
2205 (wide_int_range_lshift):New.
2206 (wide_int_range_shift_undefined_p): New.
2207 (wide_int_range_bit_xor): New.
2208 (wide_int_range_bit_ior): New.
2209 (wide_int_range_bit_and): New.
2210 (wide_int_range_trunc_mod): New.
2211 (zero_nonzero_bits_from_bounds): Rename to...
2212 (wide_int_set_zero_nonzero_bits): ...this.
2213 (zero_nonzero_bits_from_vr): Rename to...
2214 (vrp_set_zero_nonzero_bits): ...this.
2215 (range_easy_mask_min_max): Rename to...
2216 (wide_int_range_can_optimize_bit_op): this.
2217 * vr-values.c (simplify_bit_ops_using_ranges): Rename
2218 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
2220 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2222 * tree-vectorizer.h (vect_orig_stmt): New function.
2223 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
2224 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
2225 (vect_create_epilog_for_reduction): Likewise.
2226 (vectorizable_live_operation): Likewise.
2227 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2228 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
2229 * tree-vect-stmts.c (vectorizable_call): Likewise.
2230 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
2232 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2234 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
2236 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
2237 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
2238 (vect_transform_loop): Likewise.
2239 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2241 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2243 * tree-vectorizer.h (vect_schedule_slp): Return void.
2244 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2245 (vect_schedule_slp): Likewise.
2247 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2249 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
2251 (vect_transform_loop): Update calls accordingly. Schedule SLP
2252 instances before the main loop, if any exist.
2254 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2256 PR tree-optimization/86749
2257 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
2258 If the lhs is used in a COND_EXPR, check that it is being used
2259 as the "then" or "else" value.
2261 2018-08-01 Tom de Vries <tdevries@suse.de>
2264 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2265 speculation_safe_value_not_needed.
2267 2018-08-01 Richard Biener <rguenther@suse.de>
2269 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
2272 2018-08-01 Martin Liska <mliska@suse.cz>
2274 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
2275 format how successful transformation is dumped.
2276 (gimple_mod_pow2_value_transform): Likewise.
2277 (gimple_mod_subtract_transform): Likewise.
2278 (gimple_stringops_transform): Likewise.
2280 2018-08-01 Martin Liska <mliska@suse.cz>
2283 * value-prof.c (interesting_stringop_to_profile_p):
2284 Simplify the code and add BUILT_IN_MEMMOVE.
2285 (gimple_stringops_transform): Likewise.
2287 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
2289 * config/aarch64/aarch64-simd.md
2290 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
2291 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
2292 use GPI iterator instead of SI mode.
2294 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2296 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
2297 rs6000_speculation_barrier.
2298 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
2299 new barrier pattern name.
2301 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2303 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
2304 (speculation_barrier): New insn.
2306 2018-08-01 Richard Biener <rguenther@suse.de>
2309 * graphite.h: Include isl/id.h and isl/space.h to allow build
2312 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
2315 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2316 mode for COFF targets.
2317 * defaults.h (TARGET_COFF): Define.
2318 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
2319 TARGET_COFF): Define.
2320 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
2321 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
2322 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
2324 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2326 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
2327 * gimple-streamer-out.c (output_bb): Save it.
2328 * lto-streamer-in.c (input_struct_function_base): Restore
2329 instance discriminator if available. Create map on demand.
2330 * lto-streamer-out.c (output_struct_function_base): Save it if
2332 * final.c (decl_to_instance_map): Document LTO strategy.
2334 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2335 Olivier Hainque <hainque@adacore.com>
2337 * debug.h (decl_to_instance_map_t): New type.
2338 (decl_to_instance_map): Declare.
2339 (maybe_create_decl_to_instance_map): New inline function.
2340 * final.c (bb_discriminator, last_bb_discriminator): New statics,
2341 to track basic block discriminators.
2342 (final_start_function_1): Initialize them.
2343 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
2345 (decl_to_instance_map): New variable.
2346 (map_decl_to_instance, maybe_set_discriminator): New functions.
2347 (notice_source_line): Set discriminator.
2349 2018-07-31 Ian Lance Taylor <iant@golang.org>
2351 * targhooks.c (default_have_speculation_safe_value): Add
2354 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2356 * dump-context.h: Include "pretty-print.h".
2357 (dump_context::refresh_dumps_are_enabled): New decl.
2358 (dump_context::emit_item): New decl.
2359 (class dump_context): Add fields "m_test_pp" and
2361 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2362 (temp_dump_context::get_dumped_text): New decl.
2363 (class temp_dump_context): Add field "m_pp".
2364 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
2365 (dump_context::refresh_dumps_are_enabled): ...and add a test for
2367 (set_dump_file): Update for above change.
2368 (set_alt_dump_file): Likewise.
2369 (dump_loc): New overload, taking a pretty_printer *.
2370 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
2371 to any test pretty-printer.
2372 (make_item_for_dump_gimple_stmt): New function, adapted from
2373 optinfo::add_gimple_stmt.
2374 (dump_context::dump_gimple_stmt): Call it, and use the result,
2375 eliminating the direct usage of dump_file and alt_dump_file in
2376 favor of indirectly using them via emit_item.
2377 (make_item_for_dump_gimple_expr): New function, adapted from
2378 optinfo::add_gimple_expr.
2379 (dump_context::dump_gimple_expr): Call it, and use the result,
2380 eliminating the direct usage of dump_file and alt_dump_file in
2381 favor of indirectly using them via emit_item.
2382 (make_item_for_dump_generic_expr): New function, adapted from
2384 (dump_context::dump_generic_expr): Call it, and use the result,
2385 eliminating the direct usage of dump_file and alt_dump_file in
2386 favor of indirectly using them via emit_item.
2387 (make_item_for_dump_printf_va): New function, adapted from
2388 optinfo::add_printf_va.
2389 (make_item_for_dump_printf): New function.
2390 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
2391 and use the result, eliminating the direct usage of dump_file and
2392 alt_dump_file in favor of indirectly using them via emit_item.
2393 (make_item_for_dump_dec): New function.
2394 (dump_context::dump_dec): Call it, and use the result,
2395 eliminating the direct usage of dump_file and alt_dump_file in
2396 favor of indirectly using them via emit_item.
2397 (make_item_for_dump_symtab_node): New function, adapted from
2398 optinfo::add_symtab_node.
2399 (dump_context::dump_symtab_node): Call it, and use the result,
2400 eliminating the direct usage of dump_file and alt_dump_file in
2401 favor of indirectly using them via emit_item.
2402 (dump_context::begin_scope): Reimplement, avoiding direct usage
2403 of dump_file and alt_dump_file in favor of indirectly using them
2405 (dump_context::emit_item): New member function.
2406 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2407 Set up test pretty-printer on the underlying context. Call
2408 refresh_dumps_are_enabled.
2409 (temp_dump_context::~temp_dump_context): Call
2410 refresh_dumps_are_enabled.
2411 (temp_dump_context::get_dumped_text): New member function.
2412 (selftest::verify_dumped_text): New function.
2413 (ASSERT_DUMPED_TEXT_EQ): New macro.
2414 (selftest::test_capture_of_dump_calls): Run all tests twice, with
2415 and then without optinfo enabled. Add uses of
2416 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
2417 * dumpfile.h: Update comment for the dump_* API.
2418 * optinfo-emit-json.cc
2419 (selftest::test_building_json_from_dump_calls): Update for new
2420 param for temp_dump_context ctor.
2421 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
2422 and "m_owned" field.
2423 (optinfo_item::~optinfo_item): Likewise.
2424 (optinfo::add_item): New member function.
2425 (optinfo::emit): Update comment.
2426 (optinfo::add_string): Delete.
2427 (optinfo::add_printf): Delete.
2428 (optinfo::add_printf_va): Delete.
2429 (optinfo::add_gimple_stmt): Delete.
2430 (optinfo::add_gimple_expr): Delete.
2431 (optinfo::add_tree): Delete.
2432 (optinfo::add_symtab_node): Delete.
2433 (optinfo::add_dec): Delete.
2434 * optinfo.h (class dump_context): New forward decl.
2435 (optinfo::add_item): New decl.
2436 (optinfo::add_string): Delete.
2437 (optinfo::add_printf): Delete.
2438 (optinfo::add_printf_va): Delete.
2439 (optinfo::add_gimple_stmt): Delete.
2440 (optinfo::add_gimple_expr): Delete.
2441 (optinfo::add_tree): Delete.
2442 (optinfo::add_symtab_node): Delete.
2443 (optinfo::add_dec): Delete.
2444 (optinfo::add_poly_int): Delete.
2445 (optinfo_item::optinfo_item): Remove "owned" param.
2446 (class optinfo_item): Remove field "m_owned".
2448 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2451 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
2452 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
2453 requested variable alignment.
2454 (expand_one_ssa_partition): Likewise.
2455 (expand_one_var): Likewise.
2457 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2459 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2460 to speculation_safe_value_not_needed.
2462 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2464 * targhooks.h (speculation_safe_value_not_needed): New prototype.
2465 * targhooks.c (speculation_safe_value_not_needed): New function.
2466 * target.def (have_speculation_safe_value): Update documentation.
2467 * doc/tm.texi: Regenerated.
2469 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2471 * config/aarch64/iterators.md (ALLI_TI): New iterator.
2472 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
2474 (despeculate_copy<ALLI:mode>_insn): New insn.
2475 (despeculate_copyti_insn): New insn.
2476 (despeculate_simple<ALLI:mode>): New insn
2477 (despeculate_simpleti): New insn.
2478 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
2480 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
2481 aarch64_speculation_safe_value.
2482 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
2484 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2486 * config/aarch64/aarch64-speculation.cc: New file.
2487 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
2488 before pass_reorder_blocks.
2489 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
2491 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
2492 X14 and X15 when tracking speculation.
2493 * config/aarch64/aarch64.md (register name constants): Add
2494 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
2495 (unspec): Add UNSPEC_SPECULATION_TRACKER.
2496 (speculation_barrier): New insn attribute.
2497 (cmp<mode>): Allow SP in comparisons.
2498 (speculation_tracker): New insn.
2499 (speculation_barrier): Add speculation_barrier attribute.
2500 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
2501 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
2502 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
2504 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2506 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
2507 aarch64_track_speculation is true.
2508 (tb<optab><mode>1): Likewise.
2509 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
2510 generate CB[N]Z when tracking speculation.
2511 (aarch64_split_compare_and_swap): Likewise.
2512 (aarch64_split_atomic_op): Likewise.
2514 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2516 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
2518 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2520 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
2521 (speculation_barrier): New insn.
2523 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2525 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
2526 * config/arm/arm.md (speculation_barrier): New expand.
2527 (speculation_barrier_insn): New pattern.
2529 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2531 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
2532 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
2533 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
2534 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
2536 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
2537 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
2538 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
2539 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
2540 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
2541 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
2542 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
2543 * builtins.c (expand_speculation_safe_value): New function.
2544 (expand_builtin): Call it.
2545 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
2546 * doc/extend.texi: Document __builtin_speculation_safe_value.
2547 * doc/md.texi: Document "speculation_barrier" pattern.
2548 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
2549 TARGET_HAVE_SPECULATION_SAFE_VALUE.
2550 * doc/tm.texi: Regenerated.
2551 * target.def (have_speculation_safe_value, speculation_safe_value): New
2553 * targhooks.c (default_have_speculation_safe_value): New function.
2554 (default_speculation_safe_value): New function.
2555 * targhooks.h (default_have_speculation_safe_value): Add prototype.
2556 (default_speculation_safe_value): Add prototype.
2558 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2560 * dump-context.h (dump_context::dump_loc): New decl.
2561 * dumpfile.c (dump_context::dump_loc): New member function.
2562 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
2563 and dump_gimple_stmt.
2564 (dump_context::dump_gimple_expr_loc): Likewise, using
2566 (dump_context::dump_generic_expr_loc): Likewise, using
2568 (dump_context::dump_printf_loc_va): Likewise, using
2570 (dump_context::begin_scope): Explicitly using the global function
2571 "dump_loc", rather than the member function.
2573 2018-07-31 Martin Sebor <msebor@redhat.com>
2575 PR tree-optimization/86741
2576 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
2578 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
2580 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
2581 depend on whether prefetch instructions will be emitted or not.
2582 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
2583 will be emitted or not.
2584 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
2585 (TARGET_SETMEM_PFD): New macros.
2587 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2589 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
2590 (NULL_STMT_VEC_INFO): Delete.
2591 (stmt_vec_info::operator*): Likewise.
2592 (stmt_vec_info::operator gimple *): Likewise.
2593 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
2594 of NULL_STMT_VEC_INFO.
2595 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2596 (vect_reassociating_reduction_p): Likewise.
2597 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2598 (vectorizable_store): Likewise.
2599 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
2600 (vec_info::free_stmt_vec_infos): Likewise.
2602 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2604 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
2605 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
2607 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2609 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
2610 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2611 (vec_info::free_stmt_vec_info): New private member functions.
2612 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
2613 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
2614 * tree-parloops.c (gather_scalar_reductions): Remove calls to
2615 set_stmt_vec_info_vec and free_stmt_vec_infos.
2616 * tree-vect-loop.c (_loop_vec_info): Remove call to
2617 set_stmt_vec_info_vec.
2618 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
2619 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
2620 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
2621 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2622 (vec_info::free_stmt_vec_info): ...these new functions. Remove
2623 assignments in {vec_info::,}new_stmt_vec_info that are redundant
2624 with the clearing in the xcalloc.
2625 (stmt_vec_info_vec): Delete.
2626 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
2627 (vectorize_loops): Likewise.
2628 (vec_info::~vec_info): Remove argument from call to
2629 free_stmt_vec_infos.
2630 (vec_info::add_stmt): Remove vinfo argument from call to
2633 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2635 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
2636 rather than a gimple stmt.
2637 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
2638 information for pattern statements when passed the original
2639 statement; instead wait to be passed the pattern statement itself.
2640 Don't call set_vinfo_for_stmt here.
2641 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
2642 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
2643 stmt_vec_infos here.
2644 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
2645 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
2646 stmt_vec_infos entry.
2648 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2650 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
2651 * tree-vectorizer.c (vec_info::replace_stmt): New function.
2652 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
2653 * tree-vect-stmts.c (vectorizable_call): Likewise.
2654 (vectorizable_simd_clone_call): Likewise.
2656 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2658 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
2659 * tree-vectorizer.c (vec_info::remove_stmt): New function.
2660 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
2661 * tree-vect-loop.c (vect_transform_loop): Likewise.
2662 * tree-vect-slp.c (vect_schedule_slp): Likewise.
2663 * tree-vect-stmts.c (vect_remove_stores): Likewise.
2665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2667 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
2668 (vect_dr_stmt): Delete.
2669 * tree-vectorizer.c (vec_info::lookup_dr): New function.
2670 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
2672 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
2673 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
2674 (vect_verify_datarefs_alignment, vect_peeling_supportable)
2675 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2676 (vect_analyze_data_refs): Likewise.
2677 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
2679 (vect_find_same_alignment_drs): Likewise.
2680 (vect_slp_analyze_node_dependences): Update calls accordingly.
2681 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2682 instead of DR_VECT_AUX.
2683 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
2684 of a vector data references. Use vec_info::lookup_dr instead of
2686 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
2687 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2688 instead of DR_VECT_AUX.
2690 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2692 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
2694 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2696 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2697 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2698 (vect_gen_prolog_loop_niters): Likewise.
2700 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2702 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
2703 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
2704 (vect_known_alignment_in_bytes, vect_dr_behavior)
2705 (vect_get_scalar_dr_size): Take references as dr_vec_infos
2706 instead of data_references. Update calls to other routines for
2707 which the same change has been made.
2708 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
2709 dr_vec_infos instead of stmt_vec_infos.
2710 (vect_analyze_data_ref_dependence): Update call accordingly.
2711 (vect_slp_analyze_data_ref_dependence)
2712 (vect_record_base_alignments): Use DR_VECT_AUX.
2713 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
2714 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
2715 (vector_alignment_reachable_p, vect_get_data_access_cost)
2716 (vect_peeling_supportable, vect_analyze_group_access_1)
2717 (vect_analyze_group_access, vect_analyze_data_ref_access)
2718 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
2719 (vect_compile_time_alias, vect_small_gap_p)
2720 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
2721 (vect_supportable_dr_alignment): Take references as dr_vec_infos
2722 instead of data_references. Update calls to other routines for
2723 which the same change has been made.
2724 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
2725 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2726 (vect_slp_analyze_and_verify_node_alignment)
2727 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2728 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2729 (vect_setup_realignment): Use dr_vec_infos. Update calls after
2731 (_vect_peel_info::dr): Replace with...
2732 (_vect_peel_info::dr_info): ...this new field.
2733 (vect_peeling_hash_get_most_frequent)
2734 (vect_peeling_hash_choose_best_peeling): Update accordingly.
2735 (vect_peeling_hash_get_lowest_cost):
2736 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
2737 routines for which the same change has been made.
2738 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
2740 * tree-vect-loop-manip.c (get_misalign_in_elems)
2741 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
2743 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2744 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
2745 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
2746 (get_group_load_store_type, get_negative_load_store_type)
2747 (vect_get_data_ptr_increment, vectorizable_store)
2748 (vectorizable_load): Likewise.
2749 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
2750 Update calls to other routines for which the same change has been made.
2752 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2754 * tree-vectorizer.h (vec_info::move_dr): New member function.
2755 (dataref_aux): Rename to...
2756 (dr_vec_info): ...this and add "dr" and "stmt" fields.
2757 (_stmt_vec_info::dr_aux): Update accordingly.
2758 (_stmt_vec_info::data_ref_info): Delete.
2759 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
2760 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
2761 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
2762 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
2764 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
2765 (STMT_VINFO_DR_INFO): New macro.
2766 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
2767 (set_dr_misalignment): Update after rename of dataref_aux.
2768 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
2769 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
2770 initialization of STMT_VINFO_DATA_REF.
2771 * tree-vectorizer.c (vec_info::move_dr): New function.
2772 * tree-vect-patterns.c (vect_recog_bool_pattern)
2773 (vect_recog_mask_conversion_pattern)
2774 (vect_recog_gather_scatter_pattern): Use it.
2775 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
2776 the "dr" and "stmt" fields of dr_vec_info instead of
2777 STMT_VINFO_DATA_REF.
2779 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2781 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
2782 (is_pattern_stmt_p): Use it.
2783 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
2784 on pattern statements.
2786 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2788 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
2789 original stmt as a stmt_vec_info rather than a gimple stmt.
2790 (vect_pattern_recog_1): Take the statement directly as a
2791 stmt_vec_info, rather than via a gimple_stmt_iterator.
2792 Update call to vect_mark_pattern_stmts.
2793 (vect_pattern_recog): Update calls accordingly.
2795 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2797 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
2798 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
2799 a vect_def_type for the first argument.
2800 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
2801 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
2802 operand if it isn't defined by a vectorized statement.
2803 (vect_build_gather_load_calls): Remove the mask_dt argument and
2804 update calls to vect_get_vec_def_for_stmt_copy.
2805 (vectorizable_bswap): Likewise the dt argument.
2806 (vectorizable_call): Update calls to vectorizable_bswap and
2807 vect_get_vec_def_for_stmt_copy.
2808 (vectorizable_simd_clone_call, vectorizable_assignment)
2809 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
2810 (vectorizable_comparison): Update calls to
2811 vect_get_vec_def_for_stmt_copy.
2812 (vectorizable_store): Likewise. Remove now-unnecessary calls to
2814 (vect_get_loop_based_defs): Remove dt argument and update call
2815 to vect_get_vec_def_for_stmt_copy.
2816 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
2817 and vect_get_vec_def_for_stmt_copy.
2818 (vectorizable_load): Update calls to vect_build_gather_load_calls
2819 and vect_get_vec_def_for_stmt_copy.
2820 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2821 (vectorizable_reduction, vectorizable_live_operation): Update calls
2822 to vect_get_vec_def_for_stmt_copy.
2824 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2826 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
2827 and gimple stmt arguments with a stmt_vec_info.
2828 (vect_record_base_alignments): Update calls accordingly.
2829 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
2830 and gimple stmt arguments with a stmt_vec_info.
2831 (vect_build_slp_tree_1): Remove vinfo argument and update call
2832 to vect_record_max_nunits.
2833 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
2834 and vect_record_max_nunits.
2836 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2838 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
2839 file and take a stmt_vec_info instead of a gimple stmt.
2840 (supportable_widening_operation, vect_finish_replace_stmt)
2841 (vect_finish_stmt_generation, vect_get_store_rhs)
2842 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
2843 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
2844 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
2845 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
2846 (vect_create_data_ref_ptr, bump_vector_ptr)
2847 (vect_permute_store_chain, vect_setup_realignment)
2848 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
2849 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
2850 (vectorizable_reduction, vectorizable_induction)
2851 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
2852 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
2853 than gimple stmts as arguments.
2854 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
2855 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
2856 (can_group_stmts_p, vect_check_gather_scatter)
2857 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2858 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
2859 (vect_permute_load_chain, vect_shift_permute_load_chain)
2860 (vect_transform_grouped_load)
2861 (vect_record_grouped_load_vectors): Likewise.
2862 * tree-vect-loop.c (vect_fixup_reduc_chain)
2863 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
2864 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
2865 (vectorizable_reduction, vectorizable_induction)
2866 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
2867 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
2868 (vect_get_load_store_mask): Likewise.
2869 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
2870 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
2871 * tree-vect-stmts.c (vect_mark_relevant)
2872 (is_simple_and_all_uses_invariant)
2873 (exist_non_indexing_operands_for_use_p, process_use)
2874 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
2875 (vect_get_vec_def_for_operand, vect_get_vec_defs)
2876 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
2877 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
2878 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
2879 (get_negative_load_store_type, get_load_store_type)
2880 (vect_check_load_store_mask, vect_check_store_rhs)
2881 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
2882 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2883 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2884 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2885 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
2886 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
2887 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
2888 (supportable_widening_operation): Likewise.
2890 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2892 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
2893 a stmt_vec_info instead of a gcall.
2894 (vect_check_gather_scatter): Update call accordingly.
2895 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
2897 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
2898 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
2899 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
2900 instead of a gimple stmt.
2901 (vect_transform_loop): Update calls accordingly.
2902 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
2903 stmt_vec_infos instead of gimple stmts.
2904 (vect_analyze_slp_instance): Update use accordingly.
2905 * tree-vect-stmts.c (read_vector_array, write_vector_array)
2906 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
2907 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
2908 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
2909 (vect_gen_widened_results_half, vect_get_loop_based_defs)
2910 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
2911 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
2912 down to subroutines.
2914 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2916 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
2917 of the worklist from a vector of gimple stmts to a vector of
2919 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2920 (vect_mark_stmts_to_be_vectorized): Likewise
2922 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2924 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
2925 statement before passing it to vect_analyze_stmt.
2926 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
2927 the chain of phi vector definitions. Track the exit phi via its
2929 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
2930 STMT_VINFO_REDUC_DEF.
2931 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
2932 stmt_vec_infos to handle the statement chains.
2933 (vect_get_slp_defs): Record the first statement in the node
2934 using a stmt_vec_info.
2935 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
2936 statements here and pass their stmt_vec_info down to subroutines.
2937 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
2938 down to vect_finish_stmt_generation.
2939 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
2940 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
2941 stmt_vec_infos to subroutines.
2942 (vect_remove_stores): Use stmt_vec_infos to handle the statement
2945 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2947 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
2948 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
2949 (vect_permute_store_chain, vect_setup_realignment)
2950 (vect_permute_load_chain, vect_shift_permute_load_chain)
2951 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
2952 stmts internally, and when passing values to other vectorizer routines.
2953 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2954 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2955 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2956 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2957 (vectorizable_reduction, vectorizable_induction)
2958 (vectorizable_live_operation, vect_transform_loop_stmt)
2959 (vect_transform_loop): Likewise.
2960 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2961 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2962 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2963 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2964 (vect_slp_analyze_node_operations_1): Likewise.
2965 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2966 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2967 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2968 (vect_finish_stmt_generation_1, get_group_load_store_type)
2969 (get_load_store_type, vect_build_gather_load_calls)
2970 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2971 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2972 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2973 (vectorizable_store, vectorizable_load, vectorizable_condition)
2974 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2975 (supportable_widening_operation): Likewise.
2976 (vect_get_vector_types_for_stmt): Likewise.
2977 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2979 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2981 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2982 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2983 (vect_permute_store_chain, vect_permute_load_chain)
2984 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2985 repeated stmt_vec_info lookups.
2986 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2987 (vect_update_ivs_after_vectorizer): Likewise.
2988 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2989 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2990 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2991 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2992 (vect_bb_slp_scalar_cost): Likewise.
2993 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2995 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2997 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2998 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2999 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3000 to get gassigns and gcalls, rather than operating on generc gimple
3002 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3003 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3004 (vectorizable_load, vect_analyze_stmt): Likewise.
3005 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3007 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3009 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3010 return stmt_vec_infos rather than gimple stmts. Do not accept
3012 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3014 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3015 Update use of get_later_stmt.
3016 (vect_get_constant_vectors): Update call accordingly.
3017 (vect_schedule_slp_instance): Likewise
3018 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3019 (vect_slp_analyze_instance_dependence): Likewise.
3020 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3022 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3024 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3025 (stmt_info_for_cost::stmt_info): ...this new field.
3026 (add_stmt_costs): Update accordingly.
3027 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3028 (vect_get_known_peeling_cost): Likewise.
3029 (vect_estimate_min_profitable_iters): Likewise.
3030 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3032 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3034 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3035 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3036 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3038 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3040 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3042 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3043 a gimple stmt to a stmt_vec_info.
3044 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3046 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3048 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3049 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3050 (_loop_vec_info::reduction_chains): Likewise.
3051 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3053 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3055 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3057 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3058 a gimple stmt to a stmt_vec_info.
3059 (_stmt_vec_info::next_element): Likewise.
3060 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3061 (vect_slp_analyze_and_verify_node_alignment)
3062 (vect_analyze_group_access_1, vect_analyze_group_access)
3063 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3064 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3065 (vect_supportable_dr_alignment): Update accordingly.
3066 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3067 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3068 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3069 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3070 * tree-vect-slp.c (vect_build_slp_tree_1)
3071 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3072 (vect_split_slp_store_group, vect_analyze_slp_instance)
3073 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3074 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3075 (get_group_load_store_type, get_load_store_type)
3076 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3077 (vect_transform_stmt, vect_remove_stores): Likewise.
3079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3081 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3083 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3084 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3085 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3086 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3087 (vector_alignment_reachable_p, vect_get_data_access_cost)
3088 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3089 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3090 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3091 (vect_analyze_group_access_1, vect_analyze_group_access)
3092 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3093 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3094 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3095 result of vect_dr_stmt and use the stmt_vec_info instead of
3096 the associated gimple stmt.
3097 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3098 (vect_gen_prolog_loop_niters): Likewise.
3099 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3101 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3103 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3104 to a vec<stmt_vec_info>.
3105 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3106 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3108 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3109 to a vec<stmt_vec_info>.
3110 (bst_traits::value_type, bst_traits::value_type): Likewise.
3111 (bst_traits::hash): Update accordingly.
3112 (vect_get_and_check_slp_defs): Change the stmts parameter from
3113 a vec<gimple *> to a vec<stmt_vec_info>.
3114 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3115 (vect_build_slp_tree): Likewise.
3116 (vect_build_slp_tree_2): Likewise. Update uses of
3117 SLP_TREE_SCALAR_STMTS.
3118 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3119 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3120 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3121 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3122 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3123 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3124 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3125 (vect_get_constant_vectors, vect_get_slp_defs)
3126 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3127 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3128 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3129 instead of gimple stmts.
3130 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3131 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3132 (vect_slp_analyze_instance_dependence): Update uses of
3133 SLP_TREE_SCALAR_STMTS.
3134 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3135 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3136 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3137 (get_initial_defs_for_reduction): Likewise.
3138 (vect_create_epilog_for_reduction): Likewise.
3139 (vectorize_fold_left_reduction): Likewise.
3140 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3141 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3142 (can_vectorize_live_stmts): Likewise.
3144 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3146 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3147 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3148 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3150 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3152 (gather_scalar_reductions): Update after above interface changes.
3153 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3154 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3156 (vect_force_simple_reduction): Likewise.
3157 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3158 LOOP_VINFO_REDUCTIONS.
3159 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3161 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3163 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3164 a gimple stmt to a stmt_vec_info.
3165 * tree-vect-loop.c (vect_active_double_reduction_p)
3166 (vect_force_simple_reduction, vectorizable_reduction): Update
3169 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3171 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3172 vec<gimple *> to a vec<stmt_vec_info>.
3173 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3174 the reduction_phis argument from a vec<gimple *> to a
3176 (vectorizable_reduction): Likewise the phis local variable that
3177 is passed to vect_create_epilog_for_reduction. Update for new type
3178 of SLP_TREE_VEC_STMTS.
3179 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3180 (vectorizable_live_operation): Likewise.
3181 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3182 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3184 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3186 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3187 a gimple stmt to a stmt_vec_info.
3188 (vectorizable_condition, vectorizable_live_operation)
3189 (vectorizable_reduction, vectorizable_induction): Pass back the
3190 vectorized statement as a stmt_vec_info.
3191 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
3192 use of STMT_VINFO_VEC_STMT.
3193 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
3194 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
3195 as stmt_vec_infos rather than gimple stmts.
3196 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
3198 (vectorizable_live_operation): Likewise.
3199 (vectorizable_reduction, vectorizable_induction): Likewise,
3200 updating use of STMT_VINFO_VEC_STMT.
3201 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
3202 of STMT_VINFO_VEC_STMT.
3203 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
3204 (vectorizable_simd_clone_call, vectorizable_conversion)
3205 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3206 (vectorizable_store, vectorizable_load, vectorizable_condition)
3207 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
3208 from a gimple stmt to a stmt_vec_info.
3209 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
3210 pointer to a stmt_vec_info to the vectorizable_* routines.
3212 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3214 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
3215 a gimple stmt to a stmt_vec_info.
3216 (is_pattern_stmt_p): Update accordingly.
3217 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
3218 (vect_record_grouped_load_vectors): Likewise.
3219 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3220 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
3221 (vect_model_reduction_cost): Likewise.
3222 (vect_create_epilog_for_reduction): Likewise.
3223 (vectorizable_reduction, vectorizable_induction): Likewise.
3224 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3225 Return the stmt_vec_info for the pattern statement.
3226 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
3227 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
3228 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3229 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
3230 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3231 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
3232 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
3233 (free_stmt_vec_info, vect_is_simple_use): Likewise.
3235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3237 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
3238 (vect_finish_stmt_generation): Likewise.
3239 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3240 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
3241 (vect_build_gather_load_calls): Use the return value of the above
3242 functions instead of a separate call to vinfo_for_stmt. Use narrow
3243 scopes for the input gimple stmt and wider scopes for the associated
3244 stmt_vec_info. Use vec_info::lookup_def when setting these
3245 stmt_vec_infos from an SSA_NAME definition.
3246 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3247 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3248 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3249 (vectorizable_store, vectorizable_load, vectorizable_condition)
3250 (vectorizable_comparison): Likewise.
3251 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
3252 (vectorizable_reduction): Likewise.
3254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3256 * tree-vectorizer.h (vect_is_simple_use): Add an optional
3257 stmt_vec_info * parameter before the optional gimple **.
3258 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3259 (process_use, vect_get_vec_def_for_operand_1): Update callers.
3260 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
3261 * tree-vect-loop.c (vectorizable_reduction): Likewise.
3262 (vectorizable_live_operation): Likewise.
3263 * tree-vect-patterns.c (type_conversion_p): Likewise.
3264 (vect_look_through_possible_promotion): Likewise.
3265 (vect_recog_rotate_pattern): Likewise.
3266 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3268 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3270 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
3271 a typedef to a wrapper class.
3272 (NULL_STMT_VEC_INFO): New macro.
3273 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
3274 (stmt_vec_info::operator*): New function.
3275 (stmt_vec_info::operator gimple *): Likewise.
3276 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
3277 (add_stmt_costs): Likewise.
3278 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
3279 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3280 (vect_get_known_peeling_cost): Likewise.
3281 (vect_estimate_min_profitable_iters): Likewise.
3282 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3283 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
3284 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3285 (vectorizable_store, free_stmt_vec_infos): Likewise.
3286 (new_stmt_vec_info): Change return type of xcalloc to
3289 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3291 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
3292 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
3293 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
3294 a single_imm_use-based sequence.
3295 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
3297 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3299 * tree-vectorizer.h (vec_info::lookup_def): Declare.
3300 * tree-vectorizer.c (vec_info::lookup_def): New function.
3301 * tree-vect-patterns.c (vect_get_internal_def): Use it.
3302 (vect_widened_op_tree): Likewise.
3303 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3304 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
3305 (vectorizable_reduction): Likewise.
3306 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
3308 (vect_is_slp_reduction): Update calls accordingly. Use
3309 vec_info::lookup_def.
3310 (vect_is_simple_reduction): Likewise
3311 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
3313 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3315 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
3316 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
3317 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
3319 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
3320 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
3321 (vect_update_vf_for_slp, vect_analyze_loop_operations)
3322 (vect_is_slp_reduction, vectorizable_induction)
3323 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
3324 * tree-vect-patterns.c (vect_init_pattern_stmt):
3325 (vect_determine_min_output_precision_1, vect_determine_precisions)
3326 (vect_pattern_recog): Likewise.
3327 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
3328 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
3329 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
3330 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3331 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
3332 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
3333 info field from a loop to a loop_vec_info.
3335 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3337 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
3338 (vec_info::add_stmt): Declare.
3339 * tree-vectorizer.c (vec_info::add_stmt): New function.
3340 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
3341 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
3342 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3343 (vectorizable_induction): Likewise.
3344 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
3345 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3346 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
3347 (vectorizable_load): Likewise.
3348 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3349 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
3350 (vect_recog_gather_scatter_pattern): Likewise.
3351 (append_pattern_def_seq): Likewise. Remove a check that is
3352 performed by add_stmt itself.
3354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3356 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
3357 which make_ssa_name was called with new_stmt before new_stmt
3360 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3362 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
3364 (vect_is_slp_reduction): ...here...
3365 (vect_is_simple_reduction): ...and here. Remove repetition of tests
3366 that are already known to be false.
3368 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3370 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
3371 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
3372 STMT_VINFO_NUM_SLP_USES when it's true.
3373 (vect_free_slp_instance): Add a final_p parameter and pass it to
3375 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
3376 (vect_analyze_slp_instance): Likewise.
3377 (vect_slp_analyze_operations): Likewise.
3378 (vect_slp_analyze_bb_1): Likewise.
3379 * tree-vectorizer.c (vec_info): Likewise.
3380 * tree-vect-loop.c (vect_transform_loop): Likewise.
3382 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3384 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3385 function is not called for second and subsequent members of
3388 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3390 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
3391 cases for nested loops from here to ...
3392 (vect_create_epilog_for_reduction): ...here. Only call
3393 vect_is_simple_use for inner-loop reductions.
3395 2018-07-31 Martin Liska <mliska@suse.cz>
3397 PR gcov-profile/85338
3398 PR gcov-profile/85350
3399 PR gcov-profile/85372
3400 * profile.c (struct location_triplet): New.
3401 (struct location_triplet_hash): Likewise.
3402 (output_location): Do not output a BB that
3403 is already recorded for a line.
3404 (branch_prob): Use streamed_locations.
3406 2018-07-31 Martin Liska <mliska@suse.cz>
3408 PR gcov-profile/85370
3409 * coverage.c (coverage_begin_function): Do not mark target
3410 clones as artificial functions.
3412 2018-07-31 Martin Liska <mliska@suse.cz>
3414 PR gcov-profile/83813
3415 PR gcov-profile/84758
3416 PR gcov-profile/85217
3417 PR gcov-profile/85332
3418 * profile.c (branch_prob): Do not record GOTO expressions
3419 for GIMPLE statements which locations are already streamed.
3421 2018-07-31 Olivier Hainque <hainque@adacore.com>
3423 * gcc.c (handle_spec_function): Accept a soft_matched_part
3424 argument, as do_spec_1. Pass it down to ...
3425 (eval_spec_function): Accept a soft_matched_part argument,
3426 and pass it down to ...
3427 (do_spec_2): Accept a soft_matched_part argument, and pass
3428 it down to do_spec_1.
3429 (do_spec_1): Pass soft_matched_part to handle_spec_function.
3430 (handle_braces): Update call to handle_spec_function.
3431 (driver::set_up_specs): Update calls to do_spec_2.
3432 (compare_debug_dump_opt_spec_function): Likewise.
3433 (compare_debug_self_opt_spec_function): Likewise.
3435 2018-07-31 Olivier Hainque <hainque@adacore.com>
3437 * common.opt (nolibc): New option.
3438 * doc/invoke.texi (Link Options): Document it.
3439 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
3440 * config/alpha/linux.h: Likewise.
3441 * config/arc/elf.h: Likewise.
3442 * config/arm/uclinux-elf.h: Likewise.
3443 * config/arm/unknown-elf.h: Likewise.
3444 * config/avr/avrlibc.h: Likewise.
3445 * config/bfin/bfin.h: Likewise.
3446 * config/bfin/linux.h: Likewise.
3447 * config/bfin/uclinux.h: Likewise.
3448 * config/darwin.h: Likewise.
3449 * config/darwin10.h: Likewise.
3450 * config/darwin12.h: Likewise.
3451 * config/gnu-user.h: Likewise.
3452 * config/lm32/uclinux-elf.h: Likewise.
3453 * config/pa/pa-hpux11.h: Likewise.
3454 * config/pa/pa64-hpux.h: Likewise.
3455 * config/sparc/sparc.h: Likewise.
3457 2018-07-31 Olivier Hainque <hainque@adacore.com>
3459 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
3460 undefined variables.
3462 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3465 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
3468 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3470 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3471 terminated string literal.
3473 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3475 PR rtl-optimization/85160
3476 * combine.c (is_just_move): New function.
3477 (try_combine): Allow combining two instructions into two if neither of
3478 the original instructions was a move.
3480 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
3483 * doc/extend.texi (Global Register Variables): Discourage use of type
3485 (Local Register Variables): Likewise.
3487 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
3489 PR tree-optimization/86506
3490 * hwint.h (ceil_log2): Resync with hwint.c implementation.
3492 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
3495 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
3496 hard_regno, make sure no insn between `from` and `to` clobbers it.
3498 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
3499 Tom de Vries <tdevries@suse.de>
3501 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
3502 (PTX_DEFAULT_RUNTIME_DIM): ... this.
3503 (nvptx_goacc_validate_dims): Set default worker and gang dims to
3504 PTX_DEFAULT_RUNTIME_DIM.
3505 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
3507 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
3509 * config/pa/pa.c (pa_output_addr_vec): Align address table.
3510 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
3511 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
3513 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
3515 * config/rs6000/constraints.md (wG constraint): Delete, no longer
3517 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
3518 predicate to reflect toc fusion has been deleted.
3519 (toc_fusion_mem_raw): Delete, no longer used.
3520 (toc_fusion_mem_wrapped): Likewise.
3521 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
3523 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
3524 Delete, no longer used.
3525 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
3526 meant to be used for toc fusion.
3527 (rs6000_debug_print_mode): Delete toc fusion debugging.
3528 (rs6000_debug_reg_global): Likewise.
3529 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
3530 fusion and secondary reload support that were never used.
3531 (rs6000_option_override_internal): Delete TOC fusion, that was only
3532 partially defined, and it did not work unless you also used the
3534 (rs6000_legitimate_address_p): Delete TOC fusion support.
3535 (rs6000_opt_masks): Likewise.
3536 (fusion_wrap_memory_address): Delete function, no longer used.
3537 (fusion_split_address); Delete TOC fusion support.
3538 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
3539 longer used with toc fusion being deleted.
3540 (TARGET_TOC_FUSION_FP): Likewise.
3541 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
3543 (toc fusion spliter): Delete TOC fusion support.
3544 (toc_fusionload_<mode>): Likewise.
3545 (toc_fusionload_di): Likewise.
3546 (fusion_gpr_load_<mode>): Delete generator function, this insn no
3547 longer needs to be named. Rename predicate to delete TOC fusion.
3548 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3549 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3550 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3551 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3552 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
3554 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3556 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
3557 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
3558 __int128 in built-in function prototypes.
3559 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
3560 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
3562 2018-07-27 Martin Sebor <msebor@redhat.com>
3564 PR tree-optimization/86696
3565 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
3566 types, including enums.
3567 (handle_char_store): Be prepared for the above function to fail.
3569 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
3571 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
3572 when optimization level is lower than 2 or optimize for size.
3574 2018-07-26 Martin Sebor <msebor@redhat.com>
3576 PR tree-optimization/86043
3577 PR tree-optimization/86042
3578 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
3579 (get_string_cst_length): Rename...
3580 (get_min_string_length): ...to this. Add argument.
3581 (handle_char_store): Extend to handle multi-character stores by
3583 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
3584 * tree.h (initializer_zerop): Add argument.
3586 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3589 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
3590 declare target to variables if they have always,{to,from,tofrom} map
3593 2018-07-26 Martin Liska <mliska@suse.cz>
3596 * lto-wrapper.c: Add linker_output as prefix
3597 for ltrans_output_file.
3599 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
3601 PR rtl-optimization/85805
3602 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
3603 value for hard registers if that was written in the same mode.
3605 2018-07-26 Martin Liska <mliska@suse.cz>
3607 PR gcov-profile/86536
3608 * gcov.c (format_gcov): Use printf format %.*f directly
3609 and do not handle special values.
3611 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3613 * common/config/arc/arc-common.c (arc_option_optimization_table):
3614 Update default optimizations for size.
3616 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3618 * config/arc/arc.md (movsf_insn): Add short instruction selection.
3619 * config/arc/constraints.md (CfZ): New constraint.
3620 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
3621 (subsf3_fpu): Likewise.
3622 (cmpsf_fpu): Likewise.
3623 (cmpsf_fpu_uneq): Likewise.
3625 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3627 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
3629 (arc_is_uncached_mem_p): uncached applies to both the variable and
3632 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3634 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
3637 2018-07-25 David Malcolm <dmalcolm@redhat.com>
3639 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
3640 field "m_scopes" from vec to auto_vec.
3642 2018-07-25 Martin Liska <mliska@suse.cz>
3644 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
3647 2018-07-25 Richard Biener <rguenther@suse.de>
3650 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
3651 special wrt context_die late.
3652 (gen_subprogram_die): Re-use DIEs in local scope.
3654 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
3656 PR tree-optimization/86644
3657 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
3659 2018-07-25 Martin Liska <mliska@suse.cz>
3662 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
3663 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
3665 2018-07-25 Martin Liska <mliska@suse.cz>
3668 * params.def: Explain ASan abbreviation and provide
3669 a documentation link.
3671 2018-07-24 Martin Sebor <msebor@redhat.com>
3673 PR tree-optimization/86622
3674 PR tree-optimization/86532
3675 * builtins.h (string_length): Declare.
3676 * builtins.c (c_strlen): Correct handling of non-constant offsets.
3677 (check_access): Be prepared for non-constant length ranges.
3678 (string_length): Make extern.
3679 * expr.c (string_constant): Only handle the minor non-constant
3680 array index. Use string_constant to compute the length of
3681 a generic string constant.
3683 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
3685 PR tree-optimization/86618
3686 * tree-vect-stmts.c (vectorizable_call): Don't take the address
3687 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
3689 2018-07-24 David Malcolm <dmalcolm@redhat.com>
3691 PR tree-optimization/86636
3692 * json.cc (json::object::set): Fix comment. Add assertions.
3693 (json::array::append): Move here from json.h. Add comment and an
3695 (json::string::string): Likewise.
3696 * json.h (json::array::append): Move to json.cc.
3697 (json::string::string): Likewise.
3698 * optinfo-emit-json.cc
3699 (optrecord_json_writer::impl_location_to_json): Assert that we
3700 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
3701 wrapper around it. Expand the location once, rather than three
3703 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
3704 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
3706 (optrecord_json_writer::optinfo_to_json): Likewise, in four
3707 places. Fix some overlong lines.
3709 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3711 * config/aarch64/aarch64-simd.md
3712 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
3713 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
3714 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
3715 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
3716 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
3717 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
3718 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
3719 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
3720 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
3722 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3725 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3726 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3727 and abs_d == d, do the power of two handling if profitable.
3729 2018-07-24 Richard Biener <rguenther@suse.de>
3731 * match.pd: Add BIT_FIELD_REF canonicalizations.
3733 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3736 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3738 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3740 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
3741 terminated STRING_CST object.
3743 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3745 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
3747 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3749 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3750 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
3751 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
3752 the elements into a register.
3753 (rs6000_split_v4si_init_di_reg): Delete.
3754 (rs6000_split_v4si_init): Delete.
3755 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
3756 (vsx_init_v4si): Rewrite as a define_expand.
3758 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3760 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
3761 zero_extend argument from memory): New.
3763 2018-07-22 Martin Sebor <msebor@redhat.com>
3766 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
3767 diagnosing calls with unknown arguments unless -Walloca-larger-than
3768 is restricted to less than PTRDIFF_MAX bytes.
3770 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
3772 * doc/gcov.texi (Invoking Gcov): Editorial changes.
3774 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3776 * pretty-print.c (text_info::set_location): Remove redundant
3777 "line_table" parameter from call to rich_location::set_range.
3779 2018-07-20 Martin Sebor <msebor@redhat.com>
3782 * builtins.c (expand_builtin_alloca): Adjust.
3783 * calls.c (alloc_max_size): Simplify.
3784 * cgraphunit.c (cgraph_node::expand): Adjust.
3785 * common.opt (larger_than_size, warn_frame_larger_than): Remove
3787 (frame_larger_than_size): Same.
3788 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
3789 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
3791 * doc/invoke.texi (GCC Command Options): Document option arguments.
3792 Explain byte-size arguments and suffixes.
3793 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
3794 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
3795 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
3796 * doc/options.texi (UInteger): Expand.
3797 (Host_Wide_Int, ByteSize): Document new properties.
3798 * final.c (final_start_function_1): Include sizes in an error message.
3799 * function.c (frame_offset_overflow): Same.
3800 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
3801 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
3802 Diagnose unbounded alloca calls only for limits of less than
3804 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
3805 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
3807 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
3808 only for limits of less than PTRDIFF_MAX.
3809 * langhooks-def.h (lhd_handle_option): Change function argument
3811 * langhooks.c (lhd_handle_option): Same.
3812 * langhooks.h (handle_option): Same.
3813 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
3815 (var_type, var_type_struct): Same.
3816 (var_set): Handle ByteSize flag.
3817 * optc-gen.awk: Add comments to output to ease debugging. Make
3818 use of HOST_WIDE_INT where appropriate.
3819 * opts-gen-save.awk: Use %lx to format unsigned long.
3820 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
3821 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
3822 arguments. Parse bytes-size suffixes.
3823 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
3824 (enum_value_to_arg): Same.
3825 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
3826 (handle_option): Adjust.
3827 (generate_option): Change function argument to HOST_WIDE_INT.
3828 (cmdline_handle_error): Adjust.
3829 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
3830 (set_option): Change function argument to HOST_WIDE_INT.
3831 (option_enabled): Handle cl_host_wide_int.
3832 (get_option_state): Handle CLVC_SIZE.
3833 (control_warning_option): Same.
3834 * opts.c (common_handle_option): Change function argument to
3835 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
3836 OPT_Wvla_larger_than_.
3837 * opts.h (enum cl_var_type): Add an enumerator.
3838 * stor-layout.c (layout_decl): Print a more meaningful warning.
3839 * toplev.c (output_stack_usage): Adjust.
3841 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
3843 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
3844 call to inline_expand_builtin_string_cmp.
3845 (expand_builtin_strcmp): Likewise.
3846 (expand_builtin_strncmp): Likewise.
3847 (inline_string_cmp): Delete the last parameter, change char_type_node
3848 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
3850 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
3851 the inlining expansion on target where the type of the call has same or
3852 narrower precision than unsigned char.
3854 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3856 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
3857 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
3858 * common.opt (fsave-optimization-record): New option.
3859 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
3860 * doc/invoke.texi (-fsave-optimization-record): New option.
3861 * dumpfile.c: Include "optinfo-emit-json.h".
3862 (struct kv_pair): Move to coretypes.h.
3863 (optgroup_options): Make non-static.
3864 (dump_context::end_scope): Call
3865 optimization_records_maybe_pop_dump_scope.
3866 * dumpfile.h (optgroup_options): New decl.
3867 * json.cc: New file.
3869 * optinfo-emit-json.cc: New file.
3870 * optinfo-emit-json.h: New file.
3871 * optinfo.cc: Include "optinfo-emit-json.h".
3872 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
3873 (optinfo_enabled_p): Check optimization_records_enabled_p.
3874 (optinfo_wants_inlining_info_p): Likewise.
3875 * optinfo.h: Update comment.
3876 * profile-count.c (profile_quality_as_string): New function.
3877 * profile-count.h (profile_quality_as_string): New decl.
3878 (profile_count::quality): New accessor.
3879 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
3880 and optinfo_emit_json_cc_tests.
3881 * selftest.h (selftest::json_cc_tests): New decl.
3882 (selftest::optinfo_emit_json_cc_tests): New decl.
3883 * toplev.c: Include "optinfo-emit-json.h".
3884 (compile_file): Call optimization_records_finish.
3885 (do_compile): Call optimization_records_start.
3886 * tree-ssa-live.c: Include optinfo.h.
3887 (remove_unused_scope_block_p): Retain inlining information if
3888 optinfo_wants_inlining_info_p returns true.
3890 2018-07-20 Richard Biener <rguenther@suse.de>
3893 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
3894 to cover -flto-partition=none.
3896 2018-07-20 Martin Liska <mliska@suse.cz>
3898 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
3899 (get_decl_source_range): Remove unused function.
3901 2018-07-20 Richard Biener <rguenther@suse.de>
3903 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
3904 (struct vn_phi_s): Likewise.
3905 (struct vn_reference_s): Likewise.
3906 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
3907 for searching the slot of an entry known to be in the hash itself.
3908 (vn_phi_hasher::equal): Likewise.
3909 (vn_reference_hasher::equal): Likewise.
3910 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
3912 (optimistic_info, current_info): Remove, keeping only valid_info.
3913 (vn_reference_lookup_1): Remove fallback lookup.
3914 (vn_reference_lookup_2): Likewise.
3915 (vn_nary_op_lookup_1): Likewise.
3916 (vn_phi_lookup): Likewise.
3917 (vn_nary_build_or_lookup_1): Make sure to not chain the built
3919 (vn_reference_insert): Adjust, chain the inserted hash element
3920 at last_inserted_ref.
3921 (vn_reference_insert_pieces): Likewise.
3922 (visit_reference_op_call): Likewise.
3923 (vn_nary_op_insert_into): Chain the inserted hash element at
3925 (vn_nary_op_insert_pieces): Adjust.
3926 (vn_nary_op_insert): Likewise.
3927 (vn_nary_op_insert_stmt): Likewise.
3928 (vn_phi_insert): Adjust, chain the inserted hash element at
3930 (process_scc): Remove clearing and copying the optimistic
3931 table. Instead remove elements inserted during an optimistic
3932 iteration from the single table we maintain.
3933 (init_scc_vn): Adjust.
3934 (free_scc_vn): Likewise.
3935 (sccvn_dom_walker::record_cond): Likewise.
3936 (sccvn_dom_walker::after_dom_children): Likewise.
3938 2018-07-19 Martin Sebor <msebor@redhat.com>
3940 PR tree-optimization/84047
3941 PR tree-optimization/83776
3942 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
3943 (check_array_bounds): Call it.
3945 2018-07-19 Martin Sebor <msebor@redhat.com>
3947 * align.h (align_flags): Use member initialization.
3949 2018-07-19 David Malcolm <dmalcolm@redhat.com>
3951 * Makefile.in (OBJS): Add optinfo.o.
3952 * coretypes.h (class symtab_node): New forward decl.
3953 (struct cgraph_node): New forward decl.
3954 (class varpool_node): New forward decl.
3955 * dump-context.h: New file.
3956 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3958 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3959 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3960 (set_alt_dump_file): Likewise.
3961 (dump_context::~dump_context): New dtor.
3962 (dump_gimple_stmt): Move implementation to...
3963 (dump_context::dump_gimple_stmt): ...this new member function.
3964 Add the stmt to any pending optinfo, creating one if need be.
3965 (dump_gimple_stmt_loc): Move implementation to...
3966 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3967 Start a new optinfo and add the stmt to it.
3968 (dump_gimple_expr): Move implementation to...
3969 (dump_context::dump_gimple_expr): ...this new member function.
3970 Add the stmt to any pending optinfo, creating one if need be.
3971 (dump_gimple_expr_loc): Move implementation to...
3972 (dump_context::dump_gimple_expr_loc): ...this new member function.
3973 Start a new optinfo and add the stmt to it.
3974 (dump_generic_expr): Move implementation to...
3975 (dump_context::dump_generic_expr): ...this new member function.
3976 Add the tree to any pending optinfo, creating one if need be.
3977 (dump_generic_expr_loc): Move implementation to...
3978 (dump_context::dump_generic_expr_loc): ...this new member
3979 function. Add the tree to any pending optinfo, creating one if
3981 (dump_printf): Move implementation to...
3982 (dump_context::dump_printf_va): ...this new member function. Add
3983 the text to any pending optinfo, creating one if need be.
3984 (dump_printf_loc): Move implementation to...
3985 (dump_context::dump_printf_loc_va): ...this new member function.
3986 Start a new optinfo and add the stmt to it.
3987 (dump_dec): Move implementation to...
3988 (dump_context::dump_dec): ...this new member function. Add the
3989 value to any pending optinfo, creating one if need be.
3990 (dump_context::dump_symtab_node): New member function.
3991 (dump_context::get_scope_depth): New member function.
3992 (dump_context::begin_scope): New member function.
3993 (dump_context::end_scope): New member function.
3994 (dump_context::ensure_pending_optinfo): New member function.
3995 (dump_context::begin_next_optinfo): New member function.
3996 (dump_context::end_any_optinfo): New member function.
3997 (dump_context::s_current): New global.
3998 (dump_context::s_default): New global.
3999 (dump_scope_depth): Delete global.
4000 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4001 (dump_symtab_node): New function.
4002 (get_dump_scope_depth): Reimplement in terms of dump_context.
4003 (dump_begin_scope): Likewise.
4004 (dump_end_scope): Likewise.
4005 (selftest::temp_dump_context::temp_dump_context): New ctor.
4006 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4007 (selftest::verify_item): New function.
4008 (ASSERT_IS_TEXT): New macro.
4009 (ASSERT_IS_TREE): New macro.
4010 (ASSERT_IS_GIMPLE): New macro.
4011 (selftest::test_capture_of_dump_calls): New test.
4012 (selftest::dumpfile_c_tests): Call it.
4013 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4014 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4015 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4016 descriptive comment.
4017 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4018 (dump_node, dump_bb): Move these unrelated decls.
4019 (class dump_manager): Add leading comment.
4020 * optinfo.cc: New file.
4021 * optinfo.h: New file.
4023 2018-07-19 Michael Collison <michael.collison@arm.com>
4024 Richard Henderson <rth@redhat.com>
4026 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4027 (subti): Handle op1 zero.
4028 (subvti4, usub4ti4): New.
4029 (*sub<GPI>3_compare1_imm): New.
4030 (sub<GPI>3_carryinCV): New.
4031 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4032 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4034 2018-07-19 Michael Collison <michael.collison@arm.com>
4035 Richard Henderson <rth@redhat.com>
4037 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4038 (addti3): Create simpler code if low part is already known to be 0.
4039 (addvti4, uaddvti4): New.
4040 (*add<GPI>3_compareC_cconly_imm): New.
4041 (*add<GPI>3_compareC_cconly): New.
4042 (*add<GPI>3_compareC_imm): New.
4043 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4044 handle constants within this pattern..
4045 (*add<GPI>3_compareV_cconly_imm): New.
4046 (*add<GPI>3_compareV_cconly): New.
4047 (*add<GPI>3_compareV_imm): New.
4048 (add<GPI>3_compareV): New.
4049 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4050 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4051 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4052 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4053 with 'comparison' operator.
4054 (*add<GPI>3_compareV_cconly_imm): Ditto.
4055 (*add<GPI>3_compareV_cconly): Ditto.
4056 (*add<GPI>3_compareV_imm): Ditto.
4057 (add<GPI>3_compareV): Ditto.
4058 (add<mode>3_carryinC): Ditto.
4059 (*add<mode>3_carryinC_zero): Ditto.
4060 (*add<mode>3_carryinC): Ditto.
4061 (add<mode>3_carryinV): Ditto.
4062 (*add<mode>3_carryinV_zero): Ditto.
4063 (*add<mode>3_carryinV): Ditto.
4065 2018-07-19 Michael Collison <michael.collison@arm.com>
4066 Richard Henderson <rth@redhat.com>
4068 * config/aarch64/aarch64-modes.def (CC_V): New.
4069 * config/aarch64/aarch64-protos.h
4070 (aarch64_addti_scratch_regs): Declare
4071 (aarch64_subvti_scratch_regs): Declare.
4072 (aarch64_expand_subvti): Declare.
4073 (aarch64_gen_unlikely_cbranch): Declare
4074 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4075 for signed overflow using CC_Vmode.
4076 (aarch64_get_condition_code_1): Handle CC_Vmode.
4077 (aarch64_gen_unlikely_cbranch): New function.
4078 (aarch64_addti_scratch_regs): New function.
4079 (aarch64_subvti_scratch_regs): New function.
4080 (aarch64_expand_subvti): New function.
4082 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4084 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4086 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4087 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4090 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4093 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4094 address check not strict.
4096 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4098 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4100 (store_pair_lanes<mode>): Likewise.
4101 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4102 enum value 'ADDR_QUERY_LDP_STP_N'.
4103 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4104 (aarch64_print_address_internal): Add declaration.
4105 (aarch64_print_ldpstp_address): Remove.
4106 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4107 (aarch64_print_operand): Change printing of 'y'.
4108 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4109 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4110 'true' rather than '1'.
4111 * config/aarch64/constraints.md (Uml): Likewise.
4112 (Uml): Rename to Umn.
4115 2018-07-19 Richard Biener <rguenther@suse.de>
4117 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4119 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4120 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4121 (vn_reference_hasher): Likewise.
4122 (struct vn_tables_s): Remove obstack and alloc-pool members.
4123 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4124 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4125 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4126 (vn_reference_insert_pieces): Likewise.
4127 (alloc_vn_nary_op_noinit): Adjust.
4128 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4129 (vn_phi_eq): Adjust.
4130 (shared_lookup_phiargs): Remove.
4131 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4132 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4133 (visit_reference_op_call): Likewise.
4134 (copy_nary, copy_phi, copy_reference): Remove.
4135 (process_scc): Rewind the obstack when iterating. Do not
4136 copy the elements to valid_info but just move them from one
4137 hashtable to the other.
4138 (allocate_vn_table): Adjust.
4139 (free_vn_table): Likewise.
4140 (init_scc_vn): Likewise.
4141 (free_scc_vn): Likewise.
4143 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4146 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4147 indirect_return as function type attribute.
4148 (ix86_attribute_table): Change indirect_return to function
4150 * doc/extend.texi: Update indirect_return attribute.
4152 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4154 * wide-int.h (widest2_int): New.
4155 * gimple-fold.c (arith_overflowed_p): Use it.
4156 * tree.h (widest2_int_cst): New.
4157 * tree-vrp.c (wide_int_binop_overflow): Rename from
4158 vrp_int_const_binop.
4159 Rewrite to work on trees.
4160 (extract_range_from_multiplicative_op_1): Abstract code to...
4161 (wide_int_range_min_max): ...here.
4162 (wide_int_range_cross_product): ...and here.
4163 (extract_range_from_binary_expr_1): Abstract overflow code to...
4164 (wide_int_range_mult_wrapping): ...here.
4165 * tree-vrp.h (wide_int_range_cross_product): New.
4166 (wide_int_range_mult_wrapping): New.
4168 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4169 Julia Koval <julia.koval@intel.com>
4171 * config/i386/x86-tune-costs.h (skylake_memcpy,
4172 skylake_memset): Replace rep_prefix with unrolling for size 512.
4174 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4177 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4178 comparision with EQ_EXPR in last stmt.
4180 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4182 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4183 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4184 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4185 previously known as "PowerPC AltiVec Built-in Functions". Move
4186 some material to new subsubsections "PowerPC AltiVec Built-in
4187 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
4189 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
4190 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
4191 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
4192 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
4194 2018-07-18 Richard Biener <rguenther@suse.de>
4196 PR tree-optimization/86557
4197 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
4200 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
4202 * config/s390/s390.c (s390_function_profiler): Generate CFI.
4204 2018-07-17 Jeff Law <law@redhat.com>
4206 * config/arm/arm.c (get_label_padding): Update for recent
4207 changes to label_to_alignment.
4209 PR tree-optimization/86010
4210 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
4212 * config/mips/mips.c (vr4130_align_insns): Update for recent
4213 changes to label_to_alignment.
4215 * config/frv/frv.c (frv_label_align): Update for recent changes
4216 to label_to_alignment.
4218 * config/nios2/nios2.c (nios2_label_align): Update for recent
4219 changes which dropped ALIGN_LABELS_LOG.
4221 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
4223 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
4224 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
4226 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
4228 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4229 deprecated align_labels_log.
4231 2018-07-17 Richard Biener <rguenther@suse.de>
4234 * dwarf2out.c (init_sections_and_labels): Always generate
4235 a debug_line_str_section for early LTO debug.
4236 (dwarf2out_finish): Reset debug_line_str_hash output early.
4237 Bump counter for extra dwarf5 .debug_loc labels to not conflict
4239 (dwarf2out_early_finish): Output debug_line_str.
4241 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4243 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
4244 index register on z196 or later.
4246 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4248 * config/s390/s390.c (s390_default_align): Set default function
4250 (s390_override_options_after_change): Call s390_default align.
4251 (s390_option_override_internal): Call s390_default_align.
4252 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
4254 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4257 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
4258 to _looptemp_ clauses, other than the first two.
4260 2018-07-17 Martin Liska <mliska@suse.cz>
4262 * opts.c: Do not enable OPT_falign_* for -Os.
4264 2018-07-17 Martin Liska <mliska@suse.cz>
4266 * align.h (MAX_CODE_ALIGN): New.
4267 (MAX_CODE_ALIGN_VALUE): New.
4268 * common/config/i386/i386-common.c (ix86_handle_option):
4269 (MAX_CODE_ALIGN): Moved to align.h.
4270 * final.c (MAX_CODE_ALIGN): Likewise.
4271 * opts.c (parse_and_check_align_values):
4272 (MAX_CODE_ALIGN): Likewise.
4273 (MAX_CODE_ALIGN_VALUE): Likewise.
4275 2018-07-17 Martin Liska <mliska@suse.cz>
4277 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
4278 in order to fulfil coding style.
4279 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
4280 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4281 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4282 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
4283 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
4284 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
4285 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
4286 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
4287 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4289 2018-07-17 Martin Liska <mliska@suse.cz>
4291 * align.h: New file.
4292 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
4294 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
4295 align_flags of label_to_alignment.
4296 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
4298 * config/m68k/m68k.c: Do not use removed align_labels_value and
4300 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
4301 (LOOP_ALIGN): Likewise.
4302 (LABEL_ALIGN): Likewise.
4303 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4304 Remove not used macro.
4305 (rs6000_loop_align): Change return type to align_flags.
4306 (rs6000_loop_align_max_skip): Remove.
4307 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
4308 Change return type to align_flags.
4309 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4310 Remove not used macro.
4311 (rs6000_loop_align): Change return type to align_flags.
4312 (rs6000_loop_align_max_skip): Remove.
4313 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
4314 * config/rx/rx-protos.h (rx_align_for_label): Make it
4316 * config/rx/rx.c (rx_align_for_label): Change return type
4318 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
4320 into align_flags class.
4321 (LABEL_ALIGN): Likewise.
4322 (LOOP_ALIGN): Likewise.
4323 * config/s390/s390.c (s390_label_align): Use align_flags
4325 (s390_asm_output_function_label): Likewise.
4326 * config/sh/sh.c (sh_override_options_after_change):
4327 Use align_flags class directly without macros.
4328 (find_barrier): Likewise.
4329 (barrier_align): Likewise.
4330 (sh_loop_align): Likewise.
4331 * config/spu/spu.c (spu_option_override):
4332 Use align_flags_tuple::get_value instead of removed macros.
4333 (spu_sched_init): Likewise.
4334 * config/spu/spu.h (GTY): Likewise.
4335 * config/visium/visium.c (visium_option_override):
4336 Set "8" as default secondary alignment.
4337 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
4338 in order to guarantee secondary alignment of 8.
4339 * coretypes.h: Include align.h header file.
4340 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
4341 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
4342 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
4343 * doc/tm.texi.in: Likewise.
4344 * final.c (struct label_alignment): Remove not used structure.
4345 (LABEL_ALIGN): Change type to align_flags.
4346 (LOOP_ALIGN): Likewise.
4347 (JUMP_ALIGN): Likewise.
4348 (default_loop_align_max_skip): Remove.
4349 (default_label_align_max_skip): Likewise.
4350 (default_jump_align_max_skip): Likewise.
4351 (default_label_align_after_barrier_max_skip):
4352 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
4353 (LABEL_TO_MAX_SKIP): Remove.
4354 (label_to_alignment): Return align_flags type instead of integer.
4355 (label_to_max_skip): Remove.
4356 (align_fuzz): Use align_flags type.
4357 (compute_alignments): Use align_flags type and use align_flags::max
4358 to combine multiple alignments.
4359 (grow_label_align): Grow vec instead of C array.
4360 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
4361 (shorten_branches): Use align_flags type and use align_flags::max
4362 to combine multiple alignments.
4363 (final_scan_insn_1): Remove usage of secondary alignment that comes
4364 from label alignment, but instead use proper secondary alignment
4365 which is computed in grow_label_align.
4366 * flags.h (struct align_flags_tuple): Move to align.h.
4367 (struct align_flags): Likewise.
4368 (state_align_loops): Rename to align_loops.
4369 (state_align_jumps): Rename to align_jumps.
4370 (state_align_labels): Rename to align_labels.
4371 (state_align_functions): Rename to align_functions.
4372 (align_loops_log): Remove.
4373 (align_jumps_log): Remove.
4374 (align_labels_log): Remove.
4375 (align_functions_log): Remove.
4376 (align_loops_max_skip): Remove.
4377 (align_jumps_max_skip): Remove.
4378 (align_labels_max_skip): Remove.
4379 (align_functions_max_skip): Remove.
4380 (align_loops_value): Remove.
4381 (align_jumps_value): Remove.
4382 (align_labels_value): Remove.
4383 (align_functions_value): Remove.
4384 * output.h (label_to_alignment): Change return type to align_flags.
4385 (label_to_max_skip): Remove.
4386 * target.def: Remove loop_align_max_skip, label_align_max_skip,
4387 jump_align_max_skip macros.
4388 * targhooks.h (default_loop_align_max_skip): Remove.
4389 (default_label_align_max_skip): Likewise.
4390 (default_jump_align_max_skip): Likewise.
4391 (default_label_align_after_barrier_max_skip): Remove.
4392 * toplev.c (read_log_maxskip): Use ::normalize function.
4393 (parse_N_M): Remove not used argument and also call ::normalize.
4394 (parse_alignment_opts): Do not pass unused arguments.
4395 * varasm.c (assemble_start_function): Use directly align_functions
4396 instead of removed macros.
4397 * system.h: Do not poison removed macros.
4399 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4402 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
4403 and cond temporaries don't have reference type if iterator has
4404 pointer type. For init use &for_pre_body instead of pre_p if
4405 for_pre_body is non-empty.
4407 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4409 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
4410 double-double modes to SFmode directly directly.
4411 (trunc<mode>sf2_fprs): Delete.
4413 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4415 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
4416 for conversions between IFmode and the decimal floating point modes.
4417 (init_float128_ieee): Use the correct names for conversions between
4418 KFmode and the decimal floating point modes.
4420 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4422 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
4423 for the conversions between TDmode and IFmode.
4424 (init_float128_ieee): Use more correct names for the conversions
4425 between TDmode and KFmode.
4427 2018-07-16 Jakub Jelinek <jakub@redhat.com>
4429 PR tree-optimization/86526
4430 * builtins.c (expand_builtin_memcmp): Formatting fixes.
4431 (inline_expand_builtin_string_cmp): Likewise.
4432 (inline_string_cmp): Likewise. Use c_readstr instead of
4433 builtin_memcpy_read_str. Add unit_mode temporary.
4435 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4438 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
4439 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
4441 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4443 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
4444 Alphabetize prototypes of built-in functions, separating out
4445 built-in functions that are listed in this section but should be
4446 described elsewhere.
4448 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
4451 * expmed.c (emit_store_flag): Do not emit setcc followed by a
4452 conditional move when trapping comparison was split to a
4453 non-trapping one (and vice versa).
4455 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4457 * config/s390/s390.c (s390_function_profiler): Generate nops
4458 instead of profiler call sequences.
4459 * config/s390/s390.opt: Add the new option.
4461 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4463 * config/s390/s390.c (s390_function_profiler): Generate
4464 __mcount_loc section.
4465 * config/s390/s390.opt: Add the new option.
4467 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4469 * common.opt: Add the new warning.
4470 * config/s390/s390.c (s390_function_profiler): Emit "brasl
4471 %r0,__fentry__" when -mfentry is specified.
4472 (s390_option_override_internal): Disallow -mfentry for 31-bit
4474 * config/s390/s390.opt: Add the new option.
4476 2018-07-16 Richard Biener <rguenther@suse.de>
4479 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
4480 for function-local FUNCTION_DECL and RESULT_DECL immediately.
4482 2018-07-16 Martin Liska <mliska@suse.cz>
4485 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
4488 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4490 * config/arc/arcHS.md: Update ARCHS scheduling rules.
4492 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4494 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
4496 * config/arc/arc-cpus.def (hs4x): New cpu.
4498 * config/arc/arc-tables.opt: Regenerate.
4499 * config/arc/arc.c (arc_sched_issue_rate): New function.
4500 (TARGET_SCHED_ISSUE_RATE): Define.
4501 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
4502 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
4504 (attr tune): Add ARCHS4x tune values.
4505 (attr tune_dspmpy): Define.
4506 (*tst): Correct instruction type.
4507 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
4508 * config/arc/arcHS4x.md: New file.
4509 * config/arc/fpu.md: Update instruction type attributes.
4510 * config/arc/t-multilib: Regenerate.
4512 2018-07-16 Tom de Vries <tdevries@suse.de>
4515 * var-tracking.c (vt_initialize): Fix pre_dec handling.
4517 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4519 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
4522 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
4524 PR tree-optimization/86514
4525 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
4526 conversion to a boolean type from a type with greater precision.
4528 2018-07-16 Tom de Vries <tdevries@suse.de>
4530 * var-tracking.c (vt_initialize): Print adjusted insn slim if
4531 dump_flags request TDF_SLIM.
4533 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
4535 * fold-const.c (int_const_binop_1): Abstract...
4536 (wide_int_binop): ...wide int code here.
4537 (poly_int_binop): ...poly int code here.
4538 Abstract the rest of int_const_binop_1 into int_const_binop.
4539 * fold-const.h (wide_int_binop): New.
4540 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
4541 Remove useless PLUS/MINUS_EXPR case.
4542 (zero_nonzero_bits_from_vr): Move wide int code...
4543 (zero_nonzero_bits_from_bounds): ...here.
4544 (extract_range_from_binary_expr_1): Move mask optimization code...
4545 (range_easy_mask_min_max): ...here.
4546 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
4547 (range_easy_mask_min_max): New.
4549 2018-07-15 Jeff Law <law@redhat.com>
4552 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
4555 2018-07-14 Jim Wilson <jimw@sifive.com>
4557 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
4559 2018-07-14 Paul Koning <ni1d@arrl.net>
4561 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
4563 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
4565 * lto-streamer-out.c (copy_function_or_variable): Dump info about
4568 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
4569 Steve Munroe <munroesj52@gmail.com>
4571 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
4572 (_mm_andnot_si128): Likewise.
4573 (_mm_or_si128): Likewise.
4574 (_mm_xor_si128): Likewise.
4576 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
4579 * builtins.c (expand_builtin_memcmp): Inline the calls first
4580 when result_eq is false.
4581 (expand_builtin_strcmp): Inline the calls first.
4582 (expand_builtin_strncmp): Likewise.
4583 (inline_string_cmp): New routine. Expand a string compare
4584 call by using a sequence of char comparison.
4585 (inline_expand_builtin_string_cmp): New routine. Inline expansion
4586 a call to str(n)cmp/memcmp.
4587 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
4589 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
4591 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
4593 * config/arm/driver-arm.c: Include arm-native.h.
4594 (host_detect_local_cpu): Use auto-generated data tables.
4595 (vendors, arm_cpu_table): Delete. Move part information to ...
4596 * config/arm/arm-cpus.in: ... here.
4597 * config/arm/parsecpu.awk (gen_native): New function.
4598 (vendor, part): New CPU fields.
4599 (END): Add support for building the native CPU detection tables.
4600 * config/arm/t-arm (arm-native.h): Add build rule.
4601 (driver-arm.o): Add dependency on arm-native.h.
4603 2018-07-13 Richard Biener <rguenther@suse.de>
4606 * match.pd (addr1 - addr2): Allow either of the operand to
4609 2018-07-13 Tom de Vries <tdevries@suse.de>
4611 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
4614 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
4616 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
4617 arrays instead of numbered variables.
4619 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
4621 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
4622 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
4624 2018-07-13 Richard Biener <rguenther@suse.de>
4627 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
4628 instead of get_context_die.
4630 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
4631 Richard Biener <rguenther@suse.de>
4634 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
4635 that the loop latch destination where phi is defined.
4637 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
4639 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
4640 (riscv_expand_epilogue): Add assertion to check interrupt mode.
4641 (riscv_set_current_function): Extract getting interrupt type to new
4643 (riscv_get_interrupt_type): New function.
4644 (riscv_merge_decl_attributes): New function, checking interrupt type is
4646 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
4648 2018-07-12 Paul Koning <ni1d@arrl.net>
4650 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
4653 2018-07-12 Paul Koning <ni1d@arrl.net>
4655 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
4656 zero reference, add doloop_end instead.
4657 * doc/md.texi (decrement_and_branch_until_zero): Remove.
4658 (Looping patterns): Remove decrement_and_branch_until_zero. Add
4659 detail for doloop_end.
4661 2018-07-12 Martin Sebor <msebor@redhat.com>
4664 * attribs.c (decl_attributes): Reject conflicting attributes before
4665 calling attribute handlers.
4667 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4669 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
4671 (gcc::dump_manager::get_dump_file_name): likewise.
4672 (dump_begin): Likewise.
4673 * dumpfile.h (dump_begin): Update prototype.
4674 (gcc::dump_manager::get_dump_file_name,
4675 gcc::dump_manager::get_dump_file_name): Update prototype.
4677 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4679 * internal-fn.h (vectorizable_internal_fn_p): New function.
4680 * tree-vect-slp.c (compatible_calls_p): Likewise.
4681 (vect_build_slp_tree_1): Remove nops argument. Handle calls
4682 to internal functions.
4683 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
4685 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4687 * fold-const.h (inverse_conditions_p): Declare.
4688 * fold-const.c (inverse_conditions_p): New function.
4689 * match.pd: Use inverse_conditions_p. Add folds of view_converts
4690 that test the inverse condition of a conditional internal function.
4691 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
4692 * internal-fn.c (internal_fn_mask_index): Handle conditional
4694 (vectorized_internal_fn_supported_p): New function.
4695 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
4696 (any_pred_load_store): Replace with...
4697 (need_to_predicate): ...this new variable.
4698 (redundant_ssa_names): New variable.
4699 (ifcvt_can_use_mask_load_store): Move initial checks to...
4700 (ifcvt_can_predicate): ...this new function. Handle tree codes
4701 for which a conditional internal function exists.
4702 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
4703 instead of ifcvt_can_use_mask_load_store. Update after variable
4705 (predicate_load_or_store): New function, split out from
4706 predicate_mem_writes.
4707 (check_redundant_cond_expr): New function.
4708 (value_available_p): Likewise.
4709 (predicate_rhs_code): Likewise.
4710 (predicate_mem_writes): Rename to...
4711 (predicate_statements): ...this. Use predicate_load_or_store
4712 and predicate_rhs_code.
4713 (combine_blocks, tree_if_conversion): Update after above name changes.
4714 (ifcvt_local_dce): Handle redundant_ssa_names.
4715 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
4716 general conditional functions.
4717 * tree-vect-stmts.c (vectorizable_call): Likewise.
4719 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4720 Alan Hayward <alan.hayward@arm.com>
4721 David Sherwood <david.sherwood@arm.com>
4723 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
4724 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
4725 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
4726 plus and minus and convert them into IFN_COND_FMA-based sequences.
4727 (convert_mult_to_fma): Handle conditional plus and minus.
4729 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4731 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
4732 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
4733 (cond_fnms_optab): New optabs.
4734 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
4736 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
4737 * internal-fn.h (get_conditional_internal_fn): Declare.
4738 (get_unconditional_internal_fn): Likewise.
4739 * internal-fn.c (cond_ternary_direct): New macro.
4740 (expand_cond_ternary_optab_fn): Likewise.
4741 (direct_cond_ternary_optab_supported_p): Likewise.
4742 (FOR_EACH_COND_FN_PAIR): Likewise.
4743 (get_conditional_internal_fn): New function.
4744 (get_unconditional_internal_fn): Likewise.
4745 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
4746 (gimple_match_op::gimple_match_op): Add a new overload for 5
4748 (gimple_match_op::set_op): Likewise.
4749 (gimple_resimplify5): Declare.
4750 * genmatch.c (decision_tree::gen): Generate simplifications for
4752 * gimple-match-head.c (gimple_simplify): Define an overload for
4753 5 operands. Handle calls with 5 arguments in the top-level overload.
4754 (convert_conditional_op): Handle conversions from unconditional
4755 internal functions to conditional ones.
4756 (gimple_resimplify5): New function.
4757 (build_call_internal): Pass a fifth operand.
4758 (maybe_push_res_to_seq): Likewise.
4759 (try_conditional_simplification): Try converting conditional
4760 internal functions to unconditional internal functions.
4761 Handle 3-operand unconditional forms.
4762 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
4763 Define ternary equivalents of the current rules for binary conditional
4765 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
4767 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
4768 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
4769 (optab): Handle them.
4770 (SVE_COND_FP_TERNARY): New int iterator.
4771 (sve_fmla_op, sve_fmad_op): New int attributes.
4772 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
4773 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
4774 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
4776 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4778 * target.def (preferred_else_value): New target hook.
4779 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
4780 * doc/tm.texi: Regenerate.
4781 * targhooks.h (default_preferred_else_value): Declare.
4782 * targhooks.c (default_preferred_else_value): New function.
4783 * internal-fn.h (conditional_internal_fn_code): Declare.
4784 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
4785 (get_conditional_internal_fn): Use it.
4786 (conditional_internal_fn_code): New function.
4787 * gimple-match.h (gimple_match_cond): New struct.
4788 (gimple_match_op): Add a cond member function.
4789 (gimple_match_op::gimple_match_op): Update all forms to take a
4791 * genmatch.c (expr::gen_transform): Use the same condition as res_op
4792 for the suboperation, but don't specify a particular else_value.
4793 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
4794 (visit_nary_op, visit_reference_op_load): Pass
4795 gimple_match_cond::UNCOND to the gimple_match_op constructor.
4796 * gimple-match-head.c: Include tree-eh.h
4797 (convert_conditional_op): New function.
4798 (maybe_resimplify_conditional_op): Likewise.
4799 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
4800 (gimple_resimplify2): Likewise.
4801 (gimple_resimplify3): Likewise.
4802 (gimple_resimplify4): Likewise.
4803 (maybe_push_res_to_seq): Return null for conditional operations.
4804 (try_conditional_simplification): New function.
4805 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
4807 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
4809 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
4811 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
4813 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4815 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4817 (hash_tree): Do not hash DECL_FCONTEXT
4818 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
4819 Do not stream DECL_FCONTEXT.
4820 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
4821 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
4823 2018-07-12 Richard Biener <rguenther@suse.de>
4826 * dwarf2out.c (gen_block_die): Only output blocks when they have
4827 at least one !DECL_IGNORED_P variable.
4829 2018-07-12 Richard Biener <rguenther@suse.de>
4832 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
4833 Remove -mieee-fp handling.
4835 2018-07-12 Richard Biener <rguenther@suse.de>
4837 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
4838 left-over from last patch.
4840 2018-07-12 Jakub Jelinek <jakub@redhat.com>
4842 PR tree-optimization/86492
4843 * gimple-ssa-store-merging.c
4844 (imm_store_chain_info::coalesce_immediate_stores): Call
4845 check_no_overlap even for the merge_overlapping case. Formatting fix.
4847 2018-07-12 Richard Biener <rguenther@suse.de>
4850 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
4851 move possibly trapping operations into the conditional.
4853 2018-07-12 Richard Biener <rguenther@suse.de>
4855 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4856 (vn_lookup_simplify_result): Remove recursion limit applied
4858 (vn_nary_build_or_lookup_1): Adjust.
4859 (try_to_simplify): Likewise.
4860 * gimple-match-head.c (gimple_resimplify1): Instead apply one
4862 (gimple_resimplify2): Likewise.
4863 (gimple_resimplify3): Likewise.
4864 (gimple_resimplify4): Likewise.
4866 2018-07-11 Jakub Jelinek <jakub@redhat.com>
4868 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4869 Use __mmask64 type instead of __mmask8 for __M argument.
4870 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
4871 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
4873 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
4874 __mmask16 for __M argument.
4875 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
4876 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
4877 to __mmask16 instead of __mmask8.
4878 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
4879 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
4880 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
4881 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
4882 instead of __mmask16 for __U argument.
4883 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
4884 __mmask16 instead of __mmask8 for __U argument.
4885 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
4887 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
4889 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
4891 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
4893 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
4895 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
4897 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
4899 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
4901 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
4903 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
4905 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
4906 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
4907 return type as well as __M argument type and all casts from __mmask8
4909 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
4910 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
4911 return type as well as __M argument type and all casts from __mmask8
4913 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
4914 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
4915 return type as well as __M argument type and all casts from __mmask8
4917 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
4918 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
4919 return type as well as __M argument type and all casts from __mmask8
4921 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
4922 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
4925 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
4927 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
4928 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
4931 2018-07-11 Paul Koning <ni1d@arrl.net>
4933 * doc/md.texi (define_subst): Document how multiple occurrences of
4934 the same argument in the replacement pattern are handled.
4936 2018-07-11 Paul Koning <ni1d@arrl.net>
4938 * doc/extend.texi (Common Variable Attributes): Move "mode" into
4940 (Common Type Attributes): Add "mode" attribute.
4942 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
4944 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
4945 stream DECL_ORIGINAL_TYPE.
4946 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
4947 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
4948 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
4949 Do not walk original type.
4950 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
4952 (write_ts_decl_non_common_tree_pointers): Do not stream
4954 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4955 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4957 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4959 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4960 thread twice from the same starting edge.
4962 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4964 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4965 * gimple.c (gimple_call_nonnull_result_p): ...here...
4966 (gimple_call_nonnull_arg): ...and here.
4967 * gimple.h (gimple_call_nonnull_result_p): New.
4968 (gimple_call_nonnull_arg): New.
4970 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4972 * config/arm/arm-cpus.in: Move information from fpu field of each
4973 cpu definition to the isa field.
4974 * config/arm/parsecpu.awk (fpu): Delete match rule.
4975 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4977 2018-07-11 Richard Biener <rguenther@suse.de>
4980 * dwarf2out.c (init_sections_and_labels): Use
4981 output_asm_line_debug_info consistently.
4982 (dwarf2out_early_finish): Likewise.
4983 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4986 2018-07-11 Richard Biener <rguenther@suse.de>
4988 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4989 Rework father_bb setting in a way to avoid propagating constants
4990 multiple times on a loop body.
4992 2018-07-10 Mark Wielaard <mark@klomp.org>
4995 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4997 2018-07-10 Richard Biener <rguenther@suse.de>
4999 * hash-map.h (hash_map::iterator::operator*): Return
5000 references to key and value.
5002 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5005 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5006 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5007 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5008 (gimplify_omp_for): For composite loops, move outer
5009 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5010 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5011 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5012 TREE_LIST for both the original class iterator and the "last" helper
5013 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5014 loop, remember has_decl_expr from outer composite loops for the
5015 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5017 2018-07-09 Martin Sebor <msebor@redhat.com>
5021 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5022 accessing implicitly initialized array elements.
5023 * expr.c (string_constant): Handle string initializers of
5024 character arrays within aggregates.
5025 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5026 Store element offset. As a special case, handle zero size.
5027 (fold_nonarray_ctor_reference): Same.
5028 (fold_ctor_reference): Add argument. Store subobject offset.
5029 * gimple-fold.h (fold_ctor_reference): Add argument.
5031 2018-07-09 Paul Koning <ni1d@arrl.net>
5033 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5034 (pdp11_insn_cost): New function.
5035 (pdp11_md_asm_adjust): New function.
5036 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5037 (pdp11_rtx_costs): Update to match machine better.
5038 (output_addr_const_pdp11): Correct format mismatch warnings.
5039 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5040 * config/pdp11/pdp11.md: General change to add base_cost and/or
5041 length attributes for use by new pdp11_insn_cost function.
5042 (MIN_BRANCH): Correct definition.
5044 (doloop_end): Use standard pattern name for looping pattern.
5045 (doloop_end_nocc): New.
5046 (movsf): Add another constraint alternative.
5047 (zero_extendqihi2): Add constraint alternatives for not in place
5049 (zero_extendhisi2): Remove.
5050 (shift patterns): Add CC handling variants.
5054 (define_peephole2): New peephole to recognize mov that sets CC for
5057 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5059 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5060 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5063 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5066 * real.c (real_nextafter): Return true if result is denormal.
5068 2018-07-09 Martin Liska <mliska@suse.cz>
5070 * common.opt: Add back wrongly removed attribute.
5072 2018-07-09 Richard Biener <rguenther@suse.de>
5075 * dwarf2out.c (gen_block_die): For an early generated DIE
5076 always output high/low PC attributes.
5078 2018-07-09 Tom de Vries <tdevries@suse.de>
5080 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5081 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5082 onto VAR_DECL with abstract origin.
5084 2018-07-07 Jim Wilson <jimw@sifive.com>
5086 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5088 2018-07-07 Tom de Vries <tdevries@suse.de>
5090 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5092 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5094 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5096 (combine_bound): Use wide-int overflow calculation instead of
5098 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5100 * fold-const.c (int_const_binop_2): Same.
5101 (extract_muldiv_1): Same.
5102 (fold_div_compare): Same.
5103 (fold_abs_const): Same.
5105 * poly-int.h (add): Same.
5109 * predict.c (predict_iv_comparison): Same.
5110 * profile-count.c (slow_safe_scale_64bit): Same.
5111 * simplify-rtx.c (simplify_const_binary_operation): Same.
5112 * tree-chrec.c (tree_fold_binomial): Same.
5113 * tree-data-ref.c (split_constant_offset_1): Same.
5114 * tree-if-conv.c (idx_within_array_bound): Same.
5115 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5116 * tree-ssa-phiopt.c (minmax_replacement): Same.
5117 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5118 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5119 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5120 * wide-int.cc (wi::add_large): Same.
5121 (wi::mul_internal): Same.
5122 (wi::sub_large): Same.
5123 (wi::divmod_internal): Same.
5124 * wide-int.h: Change overflow type to overflow_type for neg, add,
5125 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5126 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5127 mul_internal, divmod_internal.
5128 (overflow_type): New enum.
5129 (accumulate_overflow): New.
5131 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5133 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5134 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5136 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5138 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5139 argument is checked for zero before entering loop, avoid checking again.
5141 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5143 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5144 * gimplify.h (generic_expr_could_trap_p): Likwise.
5145 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5147 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5149 PR tree-optimization/86401
5150 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5151 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5152 (fold_bit_and_mask): ... here. New helper function for match.pd.
5153 * fold-const.h (fold_bit_and_mask): Declare.
5154 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5156 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5159 * target.def (translate_mode_attribute): New hook.
5160 * targhooks.h (default_translate_mode_attribute): Declare.
5161 * targhooks.c (default_translate_mode_attribute): New function.
5162 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5163 * doc/tm.texi: Regenerate.
5164 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5165 (rs6000_translate_mode_attribute): New function.
5167 2018-07-06 Paul Koning <ni1d@arrl.net>
5169 * doc/md.texi (define_split): Document DONE and FAIL.
5170 (define_peephole2): Ditto.
5172 2018-07-05 Jeff Law <law@redhat.com>
5174 PR tree-optimization/86010
5175 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5176 both the head and tail of mem* and str* calls.
5178 2018-07-05 Jim Wilson <jimw@sifive.com>
5180 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5182 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5184 * config/aarch64/aarch64-simd.md: correct flags text for
5185 MIN_EXPR replacement.
5187 2018-07-05 James Clarke <jrtc27@jrtc27.com>
5189 * configure: Regenerated.
5191 2018-07-05 Carl Love <cel@us.ibm.com>
5193 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
5194 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
5195 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
5196 VSX_BUILTIN_DOUBLEL_V4SF.
5198 2018-07-05 Martin Sebor <msebor@redhat.com>
5201 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5202 than its domain to compute its the upper bound of a char array.
5204 2018-07-05 Nathan Sidwell <nathan@acm.org>
5206 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
5207 * doc/cpp.texi: Update comment.
5208 * doc/tm.texi: Rebuilt.
5209 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
5210 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
5211 * doc/extend.texi (Backwards Compatibility): Clarify it is system
5212 headers affected by extern "C".
5213 * system.h: Poison NO_IMPLICIT_EXTERN_C.
5214 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
5215 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
5216 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
5217 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
5218 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
5219 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
5220 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
5221 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
5222 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
5223 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
5224 config/sparc/sp64-elf.h, config/spu/spu.h,
5225 config/stormy16/stormy16.h, config/v850/v850.h,
5226 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
5227 define NO_IMPLICIT_EXTERN_C.
5228 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
5230 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5233 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5234 instead of GET_MODE_SIZE when comparing Units.
5236 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5239 * rtlanal.c (set_noop_p): Constrain on mode change,
5240 include hard-reg-set.h
5242 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5244 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
5246 2018-07-05 Jakub Jelinek <jakub@redhat.com>
5249 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5252 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5254 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5256 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5259 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5261 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5263 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
5266 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
5268 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
5270 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
5271 neg_*_op* variables.
5273 2018-07-04 Martin Liska <mliska@suse.cz>
5275 * tree-switch-conversion.c: Define
5276 max_ratio_for_speed and max_ratio_for_size constants.
5278 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5279 Martin Liska <mliska@suse.cz>
5284 * common.opt: Rename align options with 'str_' prefix.
5285 * common/config/i386/i386-common.c (set_malign_value): New
5287 (ix86_handle_option): Use it to set -falign-* options/
5288 * config/aarch64/aarch64-protos.h (struct tune_params): Change
5289 type from int to string.
5290 * config/aarch64/aarch64.c: Update default values from int
5292 * config/alpha/alpha.c (alpha_override_options_after_change):
5294 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
5295 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5296 max skip conditionally.
5297 * config/i386/freebsd.h (SUBALIGN_LOG): New.
5298 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5299 max skip conditionally.
5300 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5301 max skip conditionally.
5302 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
5303 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5304 max skip conditionally.
5305 * config/i386/i386.c (struct ptt): Change type from int to
5307 (ix86_default_align): Set default values.
5308 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
5309 max skip conditionally.
5310 * config/i386/iamcu.h (SUBALIGN_LOG): New.
5311 (ASM_OUTPUT_MAX_SKIP_ALIGN):
5312 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
5313 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5314 max skip conditionally.
5315 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
5316 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
5317 * config/i386/x86-64.h (SUBALIGN_LOG): New.
5318 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5319 max skip conditionally.
5320 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5321 * config/ia64/ia64.c (ia64_option_override): Set default values
5322 for alignment options.
5323 * config/m68k/m68k.c: Handle new str_align_* options.
5324 * config/mips/mips.c (mips_set_compression_mode): Change
5326 (mips_option_override): Set default values for options.
5327 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5329 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5331 * config/rx/rx.c (rx_option_override): Likewise.
5332 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
5333 (LABEL_ALIGN): Use align_labels_log.
5334 (LOOP_ALIGN): Use align_loops_align.
5335 * config/s390/s390.c (s390_asm_output_function_label): Use new
5337 * config/sh/sh.c (sh_override_options_after_change):
5338 Change type of constants.
5339 * config/spu/spu.c (spu_sched_init): Likewise.
5340 * config/sparc/sparc.c (sparc_option_override): Set default
5342 * config/visium/visium.c (visium_option_override): Likewise.
5343 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
5344 emit p2align format with last argument if it's not needed.
5345 * doc/invoke.texi: Document extended format of -falign-*.
5346 * final.c: Use align_labels alignment.
5347 * flags.h (struct target_flag_state): Change type to use
5349 (struct align_flags_tuple): New.
5350 (struct align_flags): Likewise.
5351 (align_loops_log): Redefine macro to use new types.
5352 (align_loops_max_skip): Redefine macro to use new types.
5353 (align_jumps_log): Redefine macro to use new types.
5354 (align_jumps_max_skip): Redefine macro to use new types.
5355 (align_labels_log): Redefine macro to use new types.
5356 (align_labels_max_skip): Redefine macro to use new types.
5357 (align_functions_log): Redefine macro to use new types.
5358 (align_loops): Redefine macro to use new types.
5359 (align_jumps): Redefine macro to use new types.
5360 (align_labels): Redefine macro to use new types.
5361 (align_functions): Redefine macro to use new types.
5362 (align_functions_max_skip): Redefine macro to use new types.
5363 (align_loops_value): New macro.
5364 (align_jumps_value): New macro.
5365 (align_labels_value): New macro.
5366 (align_functions_value): New macro.
5367 * function.c (invoke_set_current_function_hook): Propagate
5368 alignment values from flags to global variables default in
5370 * ipa-icf.c (sem_function::equals_wpa): Use
5371 cl_optimization_option_eq instead of memcmp.
5372 * lto-streamer.h (cl_optimization_stream_out): Support streaming
5374 (cl_optimization_stream_in): Likewise.
5375 * optc-save-gen.awk: Support strings in cl_optimization.
5376 * opth-gen.awk: Likewise.
5377 * opts.c (finish_options): Remove error checking of invalid
5379 (MAX_CODE_ALIGN): Remove.
5380 (MAX_CODE_ALIGN_VALUE): Likewise.
5381 (parse_and_check_align_values): New function.
5382 (check_alignment_argument): Likewise.
5383 (common_handle_option): Use check_alignment_argument.
5384 * opts.h (parse_and_check_align_values): Declare.
5385 * toplev.c (init_alignments): Remove.
5386 (read_log_maxskip): New.
5387 (parse_N_M): Likewise.
5388 (parse_alignment_opts): Likewise.
5389 (backend_init_target): Remove usage of init_alignments.
5390 * toplev.h (parse_alignment_opts): Declare.
5391 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
5393 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
5394 * tree.c (cl_option_hasher::equal): New.
5395 * varasm.c: Use new global macros.
5397 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5399 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
5400 Use a simpler align directive also if MAXSKIP = ALIGN-1.
5401 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5402 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5403 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5404 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5405 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
5406 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
5407 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
5408 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5409 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5410 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5411 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5413 2018-07-04 Martin Liska <mliska@suse.cz>
5414 Jonathan Wakely <jwakely@redhat.com>
5416 * coverage.c: Use correct type.
5417 * doc/invoke.texi: Language correction.
5419 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
5422 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
5423 ENDBRANCH for non-tail call which may return via indirect branch.
5424 * doc/extend.texi: Document indirect_return attribute.
5426 2018-07-03 Martin Sebor <msebor@redhat.com>
5428 PR tree-optimization/86274
5429 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
5431 (format_floating): Correct handling of infinities and NaNs.
5433 2018-07-03 Martin Sebor <msebor@redhat.com>
5435 * print-tree.c (print_real_cst): New function.
5436 (print_node_brief): Call it.
5437 (print_node): Ditto.
5439 2018-07-03 Jeff Law <law@redhat.com>
5441 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
5442 into a single pattern.
5444 * config/h8300/h8300.md (ors code_iterator): New.
5445 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
5446 a single pattern and single splitter.
5447 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
5448 (iorqi3_1, xorqi3_1): Likewise.
5449 (iorqi3, xorqi3 expanders): Similarly.
5451 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
5452 (movmd_internal) into a single pattern using the P mode iterator.
5453 (movmd splitters): Similarly.
5454 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
5455 (movsd splitters): Similarly.
5457 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
5458 ADDB, ADDW and ADDL into a single ADD attribute which selects the
5459 right table based on the size of the operand.
5460 * config/h8300/h8300.md (length_table): Corresponding changes. All
5461 references to "addb", "addw" and "addl" changed to "add".
5462 (btst patterns): Merge two variants into a single pattern.
5463 (tstqi, tsthi): Likewise.
5464 (addhi3_incdec, addsi3_incdec): Likewise.
5465 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
5466 (mulhi3, mulsi3): Likewise.
5467 (udivhi3, udivsi3): Likewise.
5468 (divhi3, divsi3): Likewise.
5469 (andorqi3, andorhi3, andorsi3): Likewise.
5471 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
5474 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
5475 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
5477 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5479 PR tree-optimization/85694
5480 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
5481 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
5483 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
5484 (<u>avg<mode>3_ceil): New patterns.
5486 2018-07-03 David Malcolm <dmalcolm@redhat.com>
5488 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
5489 scan-tree-dump directive.
5490 * gcc.dg/vect/slp-perm-2.c: Likewise.
5491 * gcc.dg/vect/slp-perm-3.c: Likewise.
5492 * gcc.dg/vect/slp-perm-5.c: Likewise.
5493 * gcc.dg/vect/slp-perm-6.c: Likewise.
5494 * gcc.dg/vect/slp-perm-7.c: Likewise.
5495 * gcc.dg/vect/slp-perm-8.c: Likewise.
5497 2018-07-03 Marek Polacek <polacek@redhat.com>
5500 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
5502 2018-07-03 Richard Biener <rguenther@suse.de>
5505 * tree-ssa-structalias.c (find_func_clobbers): Properly
5506 handle indirect calls.
5508 2018-07-03 Jeff Law <law@redhat.com>
5510 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
5511 (shifts): New code iterator.
5512 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
5513 expander. Fix HImode handling on H8/SX.
5514 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
5515 (subqi3, subhi3, subsi3 expanders): Likewise.
5516 (andqi3, andhi3, andsi3 expanders): Likewise.
5517 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
5518 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
5519 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
5520 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
5521 (zero_extendqihi2, zero_extendqisi2): Likewise.
5522 (extendqihi2, extendqisi2): Likewise.
5523 (rotlqi3, rotlhi3, rotlsi3): Likewise.
5524 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
5525 (rotlqi3_1, rotlhi3_1): Likewise.
5526 (logicalhi3_sn, logicalsi3_sn): Likewise.
5527 (logicalhi3, logicalsi3): Likewise.
5529 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5531 * tree-vect-patterns.c (vect_recog_rotate_pattern)
5532 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5533 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
5534 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
5535 type to append_pattern_def_seq instead of creating a stmt_vec_info
5537 (build_mask_conversion): Likewise. Remove vinfo argument.
5538 (vect_add_conversion_to_patterm): Likewise, renaming to...
5539 (vect_add_conversion_to_pattern): ...this.
5540 (vect_recog_mask_conversion_pattern): Update call to
5541 build_mask_conversion. Pass the vector type to
5542 append_pattern_def_seq here too.
5543 (vect_recog_gather_scatter_pattern): Update call to
5544 vect_add_conversion_to_pattern.
5546 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5548 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
5549 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5550 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
5551 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
5552 STMT_VINFO_PATTERN_DEF_SEQ to null here.
5553 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
5554 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
5555 append_pattern_def_seq instead of new_pattern_def_seq.
5556 (vect_recog_divmod_pattern): Do both of the above.
5557 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
5560 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5562 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
5563 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
5564 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
5565 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
5566 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
5567 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5568 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
5569 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5570 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5571 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
5572 parameter with a single stmt_vec_info.
5573 (vect_recog_func_ptr): Likewise.
5574 (vect_recog_gather_scatter_pattern): Likewise, folding in...
5575 (vect_try_gather_scatter_pattern): ...this.
5576 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
5577 the stmt_vec_info of the statement to be matched. Don't clear
5578 STMT_VINFO_RELATED_STMT.
5579 (vect_pattern_recog): Update call accordingly.
5581 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5583 PR tree-optimization/85694
5584 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
5585 (uavgM3_ceil): Document new optabs.
5586 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
5587 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
5589 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
5590 (savg_ceil_optab): New optabs.
5591 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
5592 (vect_vect_recog_func_ptrs): Add it.
5593 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
5594 constant directly from the associated lhs.
5596 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5598 * tree-vect-patterns.c (vect_split_statement): New function.
5599 (vect_convert_input): Use it to try to split an existing cast.
5601 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5603 * poly-int.h (print_hex): New function.
5604 * dumpfile.h (dump_dec, dump_hex): Declare.
5605 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
5606 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
5607 min_input_precision, operation_precision and operation_sign.
5608 * tree-vect-patterns.c (vect_get_range_info): New function.
5609 (vect_same_loop_or_bb_p, vect_single_imm_use)
5610 (vect_operation_fits_smaller_type): Delete.
5611 (vect_look_through_possible_promotion): Add an optional
5612 single_use_p parameter.
5613 (vect_recog_over_widening_pattern): Rewrite to use new
5614 stmt_vec_info infomration. Handle one operation at a time.
5615 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
5616 (vect_truncatable_operation_p, vect_set_operation_type)
5617 (vect_set_min_input_precision): New functions.
5618 (vect_determine_min_output_precision_1): Likewise.
5619 (vect_determine_min_output_precision): Likewise.
5620 (vect_determine_precisions_from_range): Likewise.
5621 (vect_determine_precisions_from_users): Likewise.
5622 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
5623 (vect_vect_recog_func_ptrs): Put over_widening first.
5625 (vect_pattern_recog): Call vect_determine_precisions.
5627 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5629 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
5630 statements that have been replaced by further pattern statements.
5631 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
5633 2018-07-03 Richard Biener <rguenther@suse.de>
5635 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
5636 always set *dt. Dump vectype in vectype overload.
5637 * dumpfile.h (dump_gimple_expr): New function.
5638 (dump_gimple_expr_loc): Likewise.
5639 * dumpfile.c (dump_gimple_expr): New function.
5640 (dump_gimple_expr_loc): Likewise.
5642 2018-07-02 Jeff Law <law@redhat.com>
5644 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
5645 the H8/300, H8/300H and H8/S variants into a single pattern.
5646 (movhi_h8300, movqi_h8300hs): Similarly.
5647 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
5648 (QHI mode iterator): New.
5650 * config/h8300/h8300.md: Remove trailing whitespace.
5652 2018-07-02 Jim Wilson <jimw@sifive.com>
5654 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
5655 instead of emit_insn for interrupt returns.
5656 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
5657 (riscv_sret, riscv_uret): Likewise.
5659 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5661 * pretty-print.c (selftest::test_pp_format): Move save and restore
5662 of quotes to class auto_fix_quotes, and add an instance.
5663 * selftest.c: Include "intl.h".
5664 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
5665 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
5666 * selftest.h (selftest::auto_fix_quotes): New class.
5668 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
5670 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
5671 (aarch64_sve_prepare_conditional_op): Remove.
5672 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
5673 Allow aarch64_simd_reg_or_zero as select operand; remove
5674 the aarch64_sve_prepare_conditional_op call.
5675 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
5676 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5677 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
5678 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
5679 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
5680 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
5681 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
5682 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
5683 and a splitters to match all of the *_any patterns.
5684 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
5686 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
5687 (SVE_COND_FP_BINARY_REV): Remove.
5688 (sve_int_op_rev, sve_fp_op_rev): New.
5689 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
5690 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
5691 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
5692 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
5693 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
5694 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
5695 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
5696 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
5697 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
5699 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
5700 Remove match_dup 1 from the inner unspec.
5701 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5703 * config/aarch64/aarch64.md (movprfx): New attr.
5704 (length): Default movprfx to 8.
5705 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
5706 (*madd<SVE_I>, *msub<SVE_I): Likewise.
5707 (*<su>mul<SVE_I>3_highpart): Likewise.
5708 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
5709 (*v<ASHIFT><SVE_I>3): Likewise.
5710 (*<su><MAXMIN><SVE_I>3): Likewise.
5711 (*<su><MAXMIN><SVE_F>3): Likewise.
5712 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
5713 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
5714 (*div<SVE_F>4): Likewise.
5716 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
5718 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
5721 2018-07-02 Richard Biener <rguenther@suse.de>
5723 PR tree-optimization/86363
5724 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
5725 memset argument refers to a non-variable address.
5727 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5729 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
5730 {PLUS,MINUS}_EXPR code to...
5731 (adjust_symbolic_bound): ...here,
5732 (combine_bound): ...here,
5733 (set_value_range_with_overflow): ...and here.
5735 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5737 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
5739 (extract_range_from_abs_expr): ...here.
5741 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
5743 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
5744 -fno-omit-frame-pointer when not optimizing.
5746 2018-07-02 Martin Liska <mliska@suse.cz>
5749 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5750 (propagate_nothrow): Likewise.
5752 2018-07-02 Martin Liska <mliska@suse.cz>
5755 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
5757 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5759 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
5760 function in r262149, changing "loc" param from source_location to
5761 const dump_location_t &.
5762 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
5763 declaration, as above.
5765 2018-07-01 Paul Koning <ni1d@arrl.net>
5767 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5768 -munit-asm, -mgnu-asm, -mdec-asm.
5769 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
5770 (pdp11_output_labelref): New.
5771 (pdp11_output_def): New.
5772 (pdp11_output_addr_vec_elt): New.
5773 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
5774 %# and %@ format codes.
5775 (pdp11_option_override): New.
5776 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
5777 (pdp11_output_ident): New.
5778 (pdp11_asm_named_section): New.
5779 (pdp11_asm_init_sections): New.
5780 (pdp11_file_start): New.
5781 (pdp11_file_end): New.
5782 (output_ascii): Use .ascii/.asciz for -mdec-asm.
5783 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
5784 %o, like %c but octal.
5785 (pdp11_option_override): New.
5786 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
5788 (DATA_SECTION_ASM_OP): Ditto.
5789 (READONLY_DATA_SECTION_ASM_OP): New.
5790 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
5791 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
5792 (ASM_OUTPUT_LABELREF): Ditto.
5793 (ASM_OUTPUT_DEF): Ditto.
5794 (ASM_OUTPUT_EXTERNAL): New.
5795 (ASM_OUTPUT_SOURCE_FILENAME): New.
5796 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
5797 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
5798 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
5799 %# and %@ format codes.
5800 * config/pdp11/pdp11.opt (mgnu-asm): New.
5801 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
5802 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
5803 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
5805 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
5807 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
5808 dereferencing path[] beyond its length.
5810 (debug_all_paths): New.
5811 (rewire_first_differing_edge): New.
5812 (adjust_paths_after_duplication): New.
5813 (duplicate_thread_path): Call adjust_paths_after_duplication.
5815 (thread_through_all_blocks): Add new argument to
5816 duplicate_thread_path.
5818 2018-06-30 Jim Wilson <jimw@sifive.com>
5820 * config/riscv/predicates.md (p2m1_shift_operand): New.
5821 (high_mask_shift_operand): New.
5822 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
5823 pattern using p2m1_shift_operand.
5824 (lshsi3_zero_extend_3+2): New combiner pattern using
5825 high_mask_shift_operand.
5827 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5829 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
5831 (vect_recog_rotate_pattern): ...here.
5832 (vect_convert_input): Try to insert casts of invariants in the
5834 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
5835 preheader to be empty.
5837 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5839 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
5840 vector type. If given, install it in the new statement's
5842 (vect_element_precision): New function.
5843 (vect_unpromoted_value): New struct.
5844 (vect_unpromoted_value::vect_unpromoted_value): New function.
5845 (vect_unpromoted_value::set_op): Likewise.
5846 (vect_look_through_possible_promotion): Likewise.
5847 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
5848 (vect_widened_op_tree, vect_convert_input): Likewise.
5849 (vect_convert_inputs, vect_convert_output): Likewise.
5850 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
5851 to handle the optional cast of the multiplication result and
5852 vect_widened_op_tree to detect the widened multiplication itself.
5853 Do not require the input and output of promotion casts to have
5854 the same sign, but base the signedness of the operation on the
5855 input rather than the result. If the pattern includes two
5856 promotions, check that those promotions have the same sign.
5857 Do not restrict the MULT_EXPR handling to a double-width result;
5858 handle quadruple-width results and wider. Use vect_convert_inputs
5859 to convert the inputs to the common type.
5860 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
5861 to handle the optional cast of the ABS result. Also allow a sign
5862 change or a sign extension between the ABS and MINUS.
5863 Use vect_widened_op_tree to detect the widened subtraction and use
5864 vect_convert_inputs to convert the inputs to the common type.
5865 (vect_handle_widen_op_by_const): Delete.
5866 (vect_recog_widen_op_pattern): New function.
5867 (vect_recog_widen_mult_pattern): Use it.
5868 (vect_recog_widen_shift_pattern): Likewise.
5869 (vect_recog_widen_sum_pattern): Use
5870 vect_look_through_possible_promotion to handle the promoted
5873 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5875 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
5876 the containing gimple_seq *.
5877 * gimple-iterator.h (gsi_for_stmt): Declare it.
5878 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
5879 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
5880 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5881 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5882 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
5884 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
5886 (vect_mark_pattern_stmts): ...here. Handle cases in which the
5887 statement being replaced is part of an existing pattern
5888 definition sequence, inserting the new pattern statements before
5890 (vect_pattern_recog_1): Don't return a bool. If the statement
5891 is already part of a pattern, instead apply pattern matching
5892 to the pattern definition statements. Don't clear the
5893 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
5894 (vect_pattern_recog): Don't break after the first match;
5895 continue processing the pattern definition statements instead.
5896 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
5898 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5900 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5901 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5902 (vect_recog_widen_sum_pattern): Use it.
5904 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5906 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5907 phi is not a pattern statement and has not been replaced by
5908 a pattern statement.
5909 * tree-vect-patterns.c (type_conversion_p): Don't check
5910 STMT_VINFO_IN_PATTERN_P.
5911 (vect_recog_vector_vector_shift_pattern): Likewise.
5912 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
5913 the pattern statement rather than the original statement; check
5914 directly for a WIDEN_MULT_EXPR here.
5915 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
5916 vect_is_simple_use to return the pattern statement rather
5917 than the original statement; use is_pattern_stmt_p to check
5918 for such a pattern statement.
5919 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
5920 to return the pattern statement rather than the original statement;
5921 don't do the same transformation here.
5922 (vect_is_simple_use): If the defining statement has been replaced
5923 by a pattern statement, return the pattern statement instead.
5924 Remove the corresponding (local) transformation from the vectype
5927 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5929 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
5930 end and default to null.
5931 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5932 (vectorizable_reduction): Update calls accordingly, dropping the
5933 gimple ** argument if the passed-back statement isn't needed.
5934 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
5935 (vect_recog_rotate_pattern): Likewise.
5936 (vect_recog_mask_conversion_pattern): Likewise.
5937 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5938 (vect_mask_constant_operand_p): Likewise.
5939 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
5940 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
5941 (get_group_load_store_type, get_load_store_type): Likewise.
5942 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
5943 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
5944 (vectorizable_conversion, vectorizable_assignment): Likewise.
5945 (vectorizable_shift, vectorizable_operation): Likewise.
5946 (vectorizable_store, vect_is_simple_cond): Likewise.
5947 (vectorizable_condition, vectorizable_comparison): Likewise.
5948 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
5949 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
5950 and move it to the end. Cope with null def_stmt_outs.
5952 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5954 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5956 2018-06-29 Jeff Law <law@redhat.com>
5958 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5959 displacements for TARGET_V850E2V3 and newer.
5960 (TARGET_LRA_P): Remove. Defaults to LRA now.
5961 * config/v850/v850.md (sign23byte_load): Remove.
5962 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5963 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5965 2018-06-29 Martin Liska <mliska@suse.cz>
5968 * coverage.c (coverage_init): Mangle full path name.
5969 * doc/invoke.texi: Document the change.
5970 * gcov-io.c (mangle_path): New.
5971 * gcov-io.h (mangle_path): Likewise.
5972 * gcov.c (mangle_name): Use mangle_path for path mangling.
5974 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5976 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5977 if starting source register is not even.
5979 2018-06-29 Martin Liska <mliska@suse.cz>
5981 PR tree-optimization/86263
5982 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5983 Make edge redirection.
5985 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5987 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5988 (dump_scope_depth): New variable.
5989 (get_dump_scope_depth): New function.
5990 (dump_begin_scope): New function.
5991 (dump_end_scope): New function.
5992 * dumpfile.h (get_dump_scope_depth): New declaration.
5993 (dump_begin_scope): New declaration.
5994 (dump_end_scope): New declaration.
5995 (class auto_dump_scope): New class.
5996 (AUTO_DUMP_SCOPE): New macro.
5997 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6000 2018-06-29 Richard Biener <rguenther@suse.de>
6002 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6003 compute_all_dependences succeeds.
6004 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6005 exceed --param loop-max-datarefs-for-datadeps.
6007 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6009 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6011 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6014 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6015 alternative 0 in preferred_for_speed attribute.
6017 2018-06-28 Paul Koning <ni1d@arrl.net>
6019 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6020 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6021 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6022 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6023 for shift insn patterns.
6025 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6027 * cgraph.c (cgraph_node::get_body): Replace assignments to
6028 "dump_file" with calls to set_dump_file.
6029 * dumpfile.c (alt_dump_file): Make static, and group with...
6030 (alt_flags): ...this definition.
6031 (dumps_are_enabled): New variable.
6032 (refresh_dumps_are_enabled): New function.
6033 (set_dump_file): New function.
6034 (set_alt_dump_file): New function.
6035 (gcc::dump_manager::dump_start): Replace assignments to
6036 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6038 (gcc::dump_manager::dump_finish): Likewise.
6039 * dumpfile.h (alt_dump_file): Delete decl.
6040 (dumps_are_enabled): New variable decl.
6041 (set_dump_file): New function decl.
6042 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6044 * tree-nested.c (lower_nested_functions): Replace assignments to
6045 "dump_file" with calls to set_dump_file.
6047 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6049 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6050 goto_locus of each outgoing edge of each basic block.
6052 2018-06-28 Richard Biener <rguenther@suse.de>
6054 * dwarf2out.c (decl_scope_table): Remove.
6055 (push_decl_scope): Likewise.
6056 (pop_decl_scope): Likewise.
6057 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6058 (gen_struct_or_union_type_die): Likewise.
6059 (gen_tagged_type_die): Likewise.
6060 (dwarf2out_init): Do not initialize decl_scope_table.
6061 (dwarf2out_c_finalize): Do not free it.
6063 2018-06-28 Richard Biener <rguenther@suse.de>
6065 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6066 deciding whether to not re-use a DIE.
6068 2018-06-28 Richard Biener <rguenther@suse.de>
6070 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6071 DW_AT_abstract_origin attribute.
6073 2018-06-28 Martin Liska <mliska@suse.cz>
6075 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6076 Use newly introduced constants.
6077 * tree-switch-conversion.h (struct jump_table_cluster):
6078 Define max_ratio_for_size and max_ratio_for_speed.
6080 2018-06-28 Martin Liska <mliska@suse.cz>
6082 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6083 Add new checking assert to catch invalid state.
6084 (jump_table_cluster::can_be_handled): Handle single case
6086 (jump_table_cluster::is_beneficial): Bail out for such case.
6087 (bit_test_cluster::find_bit_tests):
6088 Add new checking assert to catch invalid state.
6089 (bit_test_cluster::can_be_handled): Handle single case
6091 (bit_test_cluster::is_beneficial): Bail out for such case.
6092 (switch_decision_tree::analyze_switch_statement):
6095 2018-06-28 Martin Liska <mliska@suse.cz>
6097 * common.opt: Introduce -completion option.
6098 * gcc.c (driver_handle_option): Handle it.
6099 (driver::main): Print completions if completion
6101 * opt-suggestions.c (option_proposer::get_completions):
6103 (option_proposer::suggest_completion): Likewise.
6104 (option_proposer::find_param_completions): Likewise.
6105 (verify_autocompletions): Likewise.
6106 (test_completion_valid_options): Likewise.
6107 (test_completion_valid_params): Likewise.
6108 (in_completion_p): Likewise.
6109 (empty_completion_p): Likewise.
6110 (test_completion_partial_match): Likewise.
6111 (test_completion_garbage): Likewise.
6112 (opt_proposer_c_tests): Likewise.
6113 * opt-suggestions.h: Declare new functions.
6114 * opts.c (common_handle_option): Handle OPT__completion_.
6115 * selftest-run-tests.c (selftest::run_tests): Add
6116 opt_proposer_c_tests.
6117 * selftest.c (assert_str_startswith): New.
6118 * selftest.h (assert_str_startswith): Likewise.
6119 (opt_proposer_c_tests): New.
6120 (ASSERT_STR_STARTSWITH): Likewise.
6122 2018-06-28 Martin Liska <mliska@suse.cz>
6124 * Makefile.in: Add opt-suggestions.o.
6125 * gcc-main.c: Include opt-suggestions.h.
6126 * gcc.c (driver::driver): Likewise.
6127 (driver::~driver): Remove m_option_suggestions.
6128 (driver::build_option_suggestions): Moved to option_proposer.
6129 (driver::suggest_option): Likewise.
6130 (driver::handle_unrecognized_options): Use option_proposer.
6131 * gcc.h (class driver): Add new memver m_option_proposer.
6132 * opt-suggestions.c: New file.
6133 * opt-suggestions.h: New file.
6135 2018-06-28 Martin Liska <mliska@suse.cz>
6137 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6138 (auto_string_vec::~auto_string_vec): Likewise.
6140 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6142 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6143 prevent_decl_creation_for_types fields up and add reset_location field.
6144 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6145 statement if id->reset_location is true.
6146 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6147 id->reset_location is true.
6148 (copy_phis_for_bb): Force input_location on the arguments if
6149 id->reset_location is true.
6150 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6151 is set on the function to be inlined.
6153 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6155 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6157 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6159 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6160 registers for Pmode.
6161 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6162 hard registers for the clobbered pseudo.
6164 2018-06-27 Paul Koning <ni1d@arrl.net>
6166 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6167 mutually exclusive options.
6168 * config/pdp11/constraints.md (h): New constraint.
6169 (O): Update definition to match shift code generation.
6170 (D): New constraint.
6171 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6173 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6175 (output_jump): Change arguments.
6176 (pdp11_fixed_cc_regs): New function.
6177 (pdp11_cc_mode): Ditto.
6178 (pdp11_expand_shift): Ditto.
6179 (pdp11_assemble_shift): Ditto.
6180 (pdp11_small_shift): Ditto.
6181 (pdp11_branch_cost): Remove.
6182 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6184 (pdp11_register_move_cost): Update for CC registers.
6185 (pdp11_rtx_costs): Add case for LSHIFTRT.
6186 (pdp11_output_jump): Add CCNZ mode conditional branches.
6187 (notice_update_cc_on_set): Remove.
6188 (pdp11_cc_mode): New function.
6189 (simple_memory_operand): Correct pre/post decrement case.
6190 (no_side_effect_operand): New function.
6191 (pdp11_regno_reg_class): Add CC_REGS class.
6192 (pdp11_fixed_cc_regs): New function.
6193 (pdp11_small_shift): New function.
6194 (pdp11_expand_shift): New function to expand shift insns.
6195 (pdp11_assemble_shift): New function to output shifts.
6196 (pdp11_branch_cost): Remove.
6197 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
6198 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
6199 (WCHAR_TYPE): Ditto.
6200 (PTRDIFF_TYPE): Ditto.
6201 (ADJUST_INSN_LENGTH): New macro.
6202 (FIXED_REGISTERS): Add CC registers.
6203 (CALL_USED_REGISTERS): Ditto.
6205 (REG_CLASS_NAMES): Ditto.
6206 (REG_CLASS_CONTENTS): Ditto.
6207 (SELECT_CC_MODE): Use new function.
6208 (TARGET_FLAGS_REGNUM): New macro.
6209 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
6210 (cc0_reg_rtx): Remove.
6211 (CC_STATUS_MDEP): Remove.
6212 (CC_STATUS_MDEFP_INIT): Remove.
6213 (CC_IN_FPU): Remove.
6214 (NOTICE_UPDATE_CC): Remove.
6215 (REGISTER_NAMES): Add CC registers.
6216 (BRANCH_COST): Change to constant 1.
6217 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
6219 * config/pdp11/pdp11.opt (mbcopy): Remove.
6220 (mbcopy-builtin): Remove.
6221 (mbranch-cheap): Remove.
6222 (mbranch-expensive): Remove.
6223 * config/pdp11/predicates.md (expand_shift_operand): Update to
6224 match shift code generation.
6225 (ccnz_operator): New predicate.
6226 * doc/invoke.texi (PDP-11 Options): Remove deleted options
6227 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
6228 Remove non-existent option -mabshi, -mno-abshi. Document mutually
6230 * doc/md.texi (PDP-11): Document new D and h constraints. Update
6231 description of O constraint.
6233 2018-06-27 Jeff Law <law@redhat.com>
6234 Austin Law <austinklaw@gmail.com>
6236 * config/v850/v850.md (addsi3_set_flags): New pattern.
6237 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
6238 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
6239 (zero_extendhisi2_v850_set_flags): Likewise.
6240 (zero_extendqisi2_v850_set_flags): Likewise.
6241 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
6242 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
6243 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
6245 * config/v850/v850-protos.h (notice_update_cc): Remove.
6246 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
6247 (v850_print_operand): Handle 'D' and "d".
6248 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
6249 Add handling of arithmetic/logical operations compared against zero.
6250 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
6251 Do not look at v850_compare_op, instead get mode from last argument.
6252 (v850_gen_compare): Remove
6253 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
6254 after reload for prologue insns.
6255 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
6257 (construct_save_jarl): Likewise.
6258 (TARGET_FLAGS_REGNUM): Define.
6259 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
6260 (NOTICE_UPDATE_CC): Remove.
6261 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
6262 than cc0. Conditionalize on reload_completed.
6263 (cmpsi_insn, setfcc_insn): Likewise.
6264 (tst1 splitter): Turn into define_and_split which sets the flags
6266 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
6267 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
6268 (cstoresf4, cstoredf4): Clobber the flags.
6269 (cmpsi, cmpsf, cmpdf): Remove expanders.
6270 (setf_insn): Remove pattern.
6271 (addsi3): Turn into define_and_split which clobbers the flags after
6272 reload and a suitable pattern (addsi3_clobber_flags) for use after
6274 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
6275 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
6276 (ashrsi3, ashrsi3_v850e2): Likewise.
6277 (bins): Clobber the flags.
6278 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
6279 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
6280 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
6281 (call_value_internal_short, call_value_internal_long): Likewise.
6282 (callt_save_interrupt, callt_return_interrupt): Likewise.
6283 (save_interrupt, return_interrupt): Likewise.
6284 (callt_save_all_interrupt, save_all_interrupt): Likewise.
6285 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
6286 (restore_all_interrupt, _restore_all_interrupt): Likewise.
6287 (All FP comparisons): Only allow after reload has completed.
6289 (divh, divhu): Tweak output template.
6290 (branch_z_normal, branch_z_invert): Remove
6291 (branch_nz_normal, branch_nz_invert): Likewise.
6292 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
6294 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
6295 * config/v850/v850.c (notice_update_cc): Remove.
6296 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
6297 (CC_NO_CARRY): Likewise.
6298 (NOTICE_UPDATE_CC): Define to nothing.
6299 * config/v850/v850.md: Remove block comment on cc0 handling
6300 Remove "cc" attribute from all patterns. Remove cc_status handling
6301 from all patterns. Minor formatting fixes.
6303 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6305 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
6306 (cortex-a76.cortex-a55): Likewise.
6307 * config/aarch64/aarch64-tune.md: Regenerate.
6308 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
6309 cortex-a76.cortex-a55.
6311 2018-06-27 Jeff Law <law@redhat.com>
6313 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
6314 (MULTILIB_DIRNAMES): Similarly.
6316 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
6318 * gimple.h (gimple_return_retbnd): Delete.
6319 (gimple_return_set_retbnd): Likewise.
6320 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
6321 gimple_return_set_retbnd.
6322 * gimple-pretty-print.c (dump_gimple_return): Remove call to
6323 gimple_return_retbnd and adjust.
6324 * tree-inline.h (struct copy_body_data): Remove retbnd field.
6325 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
6326 Explicitly return NULL in a couple more cases. Move assertion
6327 on debug statements and remove unreachable code.
6328 (reset_debug_binding): Do not test id->retbnd.
6329 (expand_call_inline): Do not set it.
6331 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6333 * configure.ac: Add --disable-gcov option.
6334 * configure: Regenerate.
6335 * Makefile.in: Honour @enable_gcov@.
6336 * doc/install.texi: Document --disable-gcov.
6338 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6340 * config/arm/arm-cpus.in (cortex-a76): New entry.
6341 (cortex-a76.cortex-a55): Likewise.
6342 * config/arm/arm-tables.opt: Regenerate.
6343 * config/arm/arm-tune.md: Likewise.
6344 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
6345 * doc/invoke.texi (ARM Options): Document cortex-a76 and
6346 cortex-a76.cortex-a55.
6348 2018-06-27 Tamar Christina <tamar.christina@arm.com>
6351 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
6353 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
6355 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
6357 (EPILOGUE_USES): Likewise.
6359 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
6361 * tree-inline.c (remap_location): New function extracted from...
6362 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
6363 (copy_phis_for_bb): ...here. Call remap_location.
6364 (copy_cfg_body): Adjust call to copy_edges_for_bb.
6366 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
6368 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
6369 unaligned vsx for 16B memset.
6371 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
6374 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
6375 ieee128_float_type_node to long_double_type_node unless
6376 TARGET_LONG_DOUBLE_128 is set.
6378 2018-06-26 David Malcolm <dmalcolm@redhat.com>
6380 * cfgloop.c (get_loop_location): Convert return type from
6381 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
6382 by implicit construction from rtx_insn *, and using
6383 dump_user_location_t::from_function_decl for the fallback case.
6384 * cfgloop.h (get_loop_location): Convert return type from
6385 location_t to dump_user_location_t.
6386 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
6387 dump_printf_loc to pass in a dump_location_t rather than a
6388 location_t, via the gimple stmt.
6389 * coverage.c (get_coverage_counts): Update calls to
6390 dump_printf_loc to pass in dump_location_t rather than a
6392 * doc/optinfo.texi (Dump types): Convert example of
6393 dump_printf_loc from taking "locus" to taking "insn". Update
6394 description of the "_loc" calls to cover dump_location_t.
6395 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
6397 (dump_user_location_t::dump_user_location_t): New constructors,
6398 from gimple *stmt and rtx_insn *.
6399 (dump_user_location_t::from_function_decl): New function.
6400 (dump_loc): Make static.
6401 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
6402 const dump_location_t &.
6403 (dump_generic_expr_loc): Delete.
6404 (dump_printf_loc): Convert param "loc" from location_t to
6405 const dump_location_t &.
6406 (selftest::test_impl_location): New function.
6407 (selftest::dumpfile_c_tests): New function.
6408 * dumpfile.h: Include "profile-count.h".
6409 (class dump_user_location_t): New class.
6410 (struct dump_impl_location_t): New struct.
6411 (class dump_location_t): New class.
6412 (dump_printf_loc): Convert 2nd param from source_location to
6413 const dump_location_t &.
6414 (dump_generic_expr_loc): Delete.
6415 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
6416 const dump_location_t &.
6417 * gimple-fold.c (fold_gimple_assign): Update call to
6418 dump_printf_loc to pass in a dump_location_t rather than a
6419 location_t, via the gimple stmt.
6420 (gimple_fold_call): Likewise.
6421 * gimple-loop-interchange.cc
6422 (loop_cand::analyze_iloop_reduction_var): Update for change to
6423 check_reduction_path.
6424 (tree_loop_interchange::interchange): Update for change to
6426 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
6427 change in return-type of find_loop_location.
6428 (graphite_regenerate_ast_isl): Likewise.
6429 * graphite-optimize-isl.c (optimize_isl): Likewise.
6430 * graphite.c (graphite_transform_loops): Likewise.
6431 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
6432 pass in a dump_location_t rather than a location_t, via the
6434 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6435 * ipa.c (walk_polymorphic_call_targets): Likewise.
6436 * loop-unroll.c (report_unroll): Convert "locus" param from
6437 location_t to dump_location_t.
6438 (decide_unrolling): Update for change to get_loop_location's
6440 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
6441 location_t to dump_user_location_t.
6442 (grid_find_single_omp_among_assignments_1): Updates calls to
6443 dump_printf_loc to pass in a dump_location_t rather than a
6444 location_t, via the gimple stmt.
6445 (grid_parallel_clauses_gridifiable): Convert "tloc" from
6446 location_t to dump_location_t. Updates calls to dump_printf_loc
6447 to pass in a dump_location_t rather than a location_t, via the
6449 (grid_inner_loop_gridifiable_p): Likewise.
6450 (grid_dist_follows_simple_pattern): Likewise.
6451 (grid_gfor_follows_tiling_pattern): Likewise.
6452 (grid_target_follows_gridifiable_pattern): Likewise.
6453 (grid_attempt_target_gridification): Convert initialization
6454 of local "grid" from memset to zero-initialization; FIXME: does
6455 this require C++11? Update call to dump_printf_loc to pass in a
6456 optinfo_location rather than a location_t, via the gimple stmt.
6457 * profile.c (read_profile_edge_counts): Updates call to
6458 dump_printf_loc to pass in a dump_location_t rather than a
6460 (compute_branch_probabilities): Likewise.
6461 * selftest-run-tests.c (selftest::run_tests): Call
6463 * selftest.h (dumpfile_c_tests): New decl.
6464 * tree-loop-distribution.c (pass_loop_distribution::execute):
6465 Update for change in return type of find_loop_location.
6466 * tree-parloops.c (parallelize_loops): Likewise.
6467 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
6468 "locus" from location_t to dump_user_location_t.
6469 (canonicalize_loop_induction_variables): Likewise.
6470 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
6471 for change in return type of find_loop_location.
6472 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
6473 to dump_printf_loc to pass in a dump_location_t rather than a
6474 location_t, via the stmt.
6475 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
6477 * tree-vect-loop-manip.c (find_loop_location): Convert return
6478 type from source_location to dump_user_location_t.
6479 (vect_do_peeling): Update for above change.
6480 (vect_loop_versioning): Update for change in type of
6482 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
6483 from location_t to dump_user_location_t.
6484 (vect_estimate_min_profitable_iters): Update for change in type
6486 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
6487 location_t to dump_location_t.
6488 (vect_slp_bb): Update for change in type of vect_location.
6489 * tree-vectorizer.c (vect_location): Convert from source_location
6490 to dump_user_location_t.
6491 (try_vectorize_loop_1): Update for change in vect_location's type.
6492 (vectorize_loops): Likewise.
6493 (increase_alignment): Likewise.
6494 * tree-vectorizer.h (vect_location): Convert from source_location
6495 to dump_user_location_t.
6496 (find_loop_location): Convert return type from source_location to
6497 dump_user_location_t.
6498 (check_reduction_path): Convert 1st param from location_t to
6499 dump_user_location_t.
6500 * value-prof.c (check_counter): Update call to dump_printf_loc to
6501 pass in a dump_user_location_t rather than a location_t; update
6502 call to error_at for change in type of "locus".
6503 (check_ic_target): Update call to dump_printf_loc to
6504 pass in a dump_user_location_t rather than a location_t, via the
6507 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
6509 * config/s390/s390.h (enum processor_flags): Do not use
6510 default tune parameter when -march was specified.
6512 2018-06-26 Jakub Jelinek <jakub@redhat.com>
6515 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
6516 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
6519 2018-06-26 Richard Biener <rguenther@suse.de>
6521 PR tree-optimization/86287
6523 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
6524 (vect_analyze_loop): Initialize n_stmts.
6526 2018-06-26 Richard Biener <rguenther@suse.de>
6529 * fold-const.c (fold_convertible_p): Pointer extension
6532 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
6535 * dwarf2out.c (loc_list_has_views): Adjust comments.
6536 (dw_loc_list): Split single cross-partition range with
6539 2018-06-25 Jeff Law <law@redhat.com>
6541 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
6542 on -mbig-switch by default.
6544 * config/v850/predicates.md (const_float_1_operand): Fix match_code
6546 (const_float_0_operand): Remove unused predicate.
6547 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
6548 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
6549 (recipsf2): New expander. Original pattern now called
6551 (recipdf2, recipdf2_insn): Similarly.
6552 (rsqrtsf2, rsqrtsf2_insn): Similarly
6553 (rsqrtdf2, rsqrtdf2_insn): Similarly
6555 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
6557 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
6558 Simplify logic for FreeBSD (twice).
6560 2018-06-25 Martin Sebor <msebor@redhat.com>
6562 PR tree-optimization/86204
6563 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
6564 a strnlen result if it's less than the length of the string.
6566 2018-06-25 Martin Sebor <msebor@redhat.com>
6568 PR tree-optimization/85700
6569 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
6570 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
6571 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
6573 2018-06-25 Martin Sebor <msebor@redhat.com>
6575 * doc/extend.texi (Zero-length arrays): Update and clarify.
6577 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
6579 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
6580 added IEEE/IBM long double multilib support on PowerPC little
6581 endian Linux systems.
6582 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
6583 (MULTILIB_DEFAULTS): Likewise.
6584 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6586 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
6587 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
6588 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
6590 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
6593 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
6594 (REORDER_45): Likewise.
6596 2018-06-25 Jeff Law <law@redhat.com>
6598 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
6599 dividend to 32 bits. Adjust length.
6600 (udivmodhi4): Cleanup output template. Fix length.
6602 2018-06-25 Carl Love <cel@us.ibm.com>
6604 * config/rs6000/vsx.md: Change word selector to prefered location.
6606 2018-06-25 Richard Biener <rguenther@suse.de>
6608 PR tree-optimization/86304
6609 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
6610 epilogue-if-converted loops as well.
6612 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
6614 * lto-section-out.c (lto_begin_section): Do not print section
6615 name for noaddr and unnumbered dumps.
6617 2018-06-25 Richard Biener <rguenther@suse.de>
6619 * tree-vectorizer.h (struct vec_info_shared): New structure
6620 with parts split out from struct vec_info and loop_nest from
6621 struct _loop_vec_info.
6622 (struct vec_info): Adjust accordingly.
6623 (struct _loop_vec_info): Likewise.
6624 (LOOP_VINFO_LOOP_NEST): Adjust.
6625 (LOOP_VINFO_DATAREFS): Likewise.
6626 (LOOP_VINFO_DDRS): Likewise.
6627 (struct _bb_vec_info): Likewise.
6628 (BB_VINFO_DATAREFS): Likewise.
6629 (BB_VINFO_DDRS): Likewise.
6630 (struct _stmt_vec_info): Add dr_aux member.
6631 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
6632 (DR_MISALIGNMENT_UNINITIALIZED): New.
6633 (set_dr_misalignment): Adjust.
6634 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
6635 (vect_analyze_loop): Adjust prototype.
6636 (vect_analyze_loop_form): Likewise.
6637 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6638 Compute dependences lazily.
6639 (vect_record_base_alignments): Use shared datarefs/ddrs.
6640 (vect_verify_datarefs_alignment): Likewise.
6641 (vect_analyze_data_refs_alignment): Likewise.
6642 (vect_analyze_data_ref_accesses): Likewise.
6643 (vect_analyze_data_refs): Likewise.
6644 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
6645 constructor parameter for shared part.
6646 (vect_analyze_loop_form): Pass in shared part and adjust.
6647 (vect_analyze_loop_2): Pass in storage for the number of
6648 stmts. Move loop nest finding to the caller. Compute
6650 (vect_analyze_loop): Pass in shared part.
6651 (vect_transform_loop): Verify shared datarefs are unchanged.
6652 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
6653 constructor parameter for shared part.
6654 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
6655 (vect_slp_bb): Verify shared datarefs are unchanged before
6657 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
6659 (new_stmt_vec_info): Initialize DR_AUX misalignment to
6660 DR_MISALIGNMENT_UNINITIALIZED.
6661 * tree-vectorizer.c (vec_info::vec_info): Add constructor
6662 parameter for shared part.
6663 (vec_info::~vec_info): Adjust.
6664 (vec_info_shared::vec_info_shared): New.
6665 (vec_info_shared::~vec_info_shared): Likewise.
6666 (vec_info_shared::save_datarefs): Likewise.
6667 (vec_info_shared::check_datarefs): Likewise.
6668 (try_vectorize_loop_1): Construct shared part live for analyses
6669 of a single loop for multiple vector sizes.
6670 * tree-parloops.c (gather_scalar_reductions): Adjust.
6672 2018-06-25 Richard Biener <rguenther@suse.de>
6674 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
6675 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
6676 (vect_analyze_data_refs): Remove similar code from here and
6677 simplify accordingly.
6679 2018-06-25 Richard Biener <rguenther@suse.de>
6681 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
6682 for reverse storage order accesses rather than asserting
6683 they cannot happen here.
6685 2018-06-25 Tom de Vries <tdevries@suse.de>
6688 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
6689 Use data16 instead of .byte for insn prefix.
6691 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
6694 * parser.c (make_char_string_pack): Pass this literal chars
6695 through cpp_interpret_string.
6696 (cp_parser_userdef_numeric_literal): Check the result of
6697 make_char_string_pack.
6699 2018-06-24 Maya Rashish <coypu@sdf.org>
6701 * ginclude/stddef.h: Simplify conditions around avoiding
6702 re-definition of __size_t.
6704 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6706 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
6707 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6709 2018-06-22 Maya Rashish <coypu@sdf.org>
6711 * doc/invoke.texi (mno-fancy-math-387): Update for changes
6712 made to OpenBSD and NetBSD through the years.
6714 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6716 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6717 behavior of vec_pack (vector double, vector double) to match
6718 behavior of vec_float2 (vector double, vector double).
6720 2018-06-22 Olivier Hainque <hainque@adacore.com>
6722 * gimplify.c (gimplify_function_tree): Prevent creation
6723 of a trampoline for the address of the current function
6724 passed to entry/exit instrumentation hooks.
6726 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
6729 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
6732 2018-06-22 Martin Liska <mliska@suse.cz>
6734 PR tree-optimization/86263
6735 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6736 Bail out if is_enabled is false.
6737 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
6739 (jump_table_cluster::is_enabled): New function.
6741 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6743 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6744 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
6745 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
6746 (lto_input_ts_binfo_tree_pointers): Likewise.
6747 * tree-streamer-out.c (streamer_write_tree_bitfields,
6748 write_ts_binfo_tree_pointers): Likewise.
6749 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
6751 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6753 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
6755 2018-06-22 Martin Liska <mliska@suse.cz>
6757 * symbol-summary.h (get): Make it pure and inline move
6758 functionality from ::get function.
6759 (get): Remove and inline into ::get and ::get_create.
6760 (get_create): Move code from ::get function.
6762 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6765 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
6766 -x assembler-with-cpp.
6768 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6770 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
6771 _FILE_OFFSET_BITS=64 for C++.
6773 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
6775 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
6776 conversion insn that shows up when pr85657-3.c is compiled using
6777 IEEE 128-bit long double.
6778 (neg<mode>2_internal): Use the correct mode to check whether the
6779 mode is IBM extended.
6780 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
6781 multiply and divide external functions from being created more
6784 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6786 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
6788 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
6789 the edge can be forwarded.
6790 (cfg_layout_merge_blocks): Likewise.
6792 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6794 * except.c (finish_eh_generation): Commit edge insertions only after
6795 the EH edges have been redirected from post-landing to landing pads.
6797 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6799 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
6800 create_tmp_var_for to create the FRAME decl.
6801 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
6803 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6805 * tree-inline.c (copy_edges_for_bb): Minor tweak.
6806 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
6807 debug statement when resetting its value.
6808 (expand_call_inline): Copy the locus of the call onto the assignment
6809 of the return value, if any. Use local variable in more cases.
6811 2018-06-21 Martin Liska <mliska@suse.cz>
6813 * ipa-pure-const.c (propagate_nothrow): Use
6814 funct_state_summaries->get.
6815 (dump_malloc_lattice): Likewise.
6816 (propagate_malloc): Likewise.
6818 2018-06-21 Richard Biener <rguenther@suse.de>
6820 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
6821 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
6822 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
6824 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
6825 BLOCK_ABSTRACT_ORIGIN unconditionally.
6827 2018-06-21 David Malcolm <dmalcolm@redhat.com>
6829 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
6831 * ipa-reference.c (ipa_reference_c_finalize): Delete
6832 ipa_ref_opt_sum_summaries and set it to NULL.
6834 2018-06-21 Tom de Vries <tdevries@suse.de>
6836 PR tree-optimization/85859
6837 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
6838 test with comment from bb_no_side_effects_p.
6840 2018-06-21 Richard Biener <rguenther@suse.de>
6842 PR tree-optimization/86232
6843 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
6844 max for constant niter.
6846 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6848 * config/aarch64/aarch64-simd.md
6849 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
6851 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6853 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
6854 Make opernads of the unspec commutative.
6856 2018-06-21 Richard Biener <rguenther@suse.de>
6858 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
6859 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
6860 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6861 (vect_analyze_data_ref_dependence): Re-order checks to deal with
6863 (vect_record_base_alignments): Do not record base alignment
6864 for gathers or scatters.
6865 (vect_compute_data_ref_alignment): Drop return value that is always
6866 true. Bail out early for gathers or scatters.
6867 (vect_enhance_data_refs_alignment): Bail out early for gathers
6869 (vect_find_same_alignment_drs): Likewise.
6870 (vect_analyze_data_refs_alignment): Remove dead code.
6871 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6872 (vect_analyze_data_refs): For possible gathers or scatters do
6873 not create an alternate DR, just check their possible validity
6874 and mark them. Adjust DECL_NONALIASED handling to not rely
6876 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
6877 update inits of gathers or scatters.
6878 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
6879 Also copy gather/scatter flag to pattern vinfo.
6881 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
6883 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6884 behavior of vec_packsu (vector unsigned long long, vector unsigned
6885 long long) to match behavior of vec_packs with same signature.
6887 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
6888 Thomas Schwinge <thomas@codesourcery.com>
6889 Cesar Philippidis <cesar@codesourcery.com>
6891 * gimplify.c (gimplify_scan_omp_clauses): Add support for
6892 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6893 (gimplify_adjust_omp_clauses): Likewise.
6894 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
6895 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
6896 (gimplify_omp_target_update): Update handling of acc update and
6898 * omp-low.c (install_var_field): Remove unused parameter
6899 base_pointers_restrict.
6900 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
6901 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
6903 (omp_target_base_pointers_restrict_p): Delete.
6904 (scan_omp_target): Update call to scan_sharing_clauses.
6905 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
6907 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
6908 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6909 (convert_local_omp_clauses): Likewise.
6910 * tree-pretty-print.c (dump_omp_clause): Likewise.
6911 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
6913 (omp_clause_code_name): Likewise.
6915 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6918 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
6921 PR tree-optimization/86231
6922 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
6923 anti-range don't overwrite *vr0min before using it to compute *vr0max.
6925 2018-06-20 Tom de Vries <tdevries@suse.de>
6927 PR tree-optimization/86097
6928 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
6929 iv type if signedness of iv type is not the same as that of *nit.
6931 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6933 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
6934 EDGE_EH edges, verify they are all EDGE_EH.
6936 2018-06-20 Maya Rashish <coypu@sdf.org>
6938 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
6940 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6942 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
6943 * config/aarch64/aarch64.c (xgene1_tunings): Add
6944 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
6945 (aarch64_mode_valid_for_sched_fusion_p):
6946 Allow 16-byte modes.
6947 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
6948 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
6950 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
6952 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6953 * config/aarch64/iterators.md (VQ2): New mode iterator.
6955 2018-06-20 Martin Liska <mliska@suse.cz>
6957 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6958 Change default ratio from 10 to 8.
6960 2018-06-20 Martin Liska <mliska@suse.cz>
6962 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6964 (bit_test_cluster::find_bit_tests): Likewise.
6965 (switch_decision_tree::analyze_switch_statement): Find clusters.
6966 * tree-switch-conversion.h (struct jump_table_cluster): Document
6969 2018-06-20 Martin Liska <mliska@suse.cz>
6971 * tree-switch-conversion.c (switch_conversion::collect):
6972 Record m_uniq property.
6973 (switch_conversion::expand): Bail out for special conditions.
6974 (group_cluster::~group_cluster): New.
6975 (group_cluster::group_cluster): Likewise.
6976 (group_cluster::dump): Likewise.
6977 (jump_table_cluster::emit): New.
6978 (switch_decision_tree::fix_phi_operands_for_edges): New.
6979 (struct case_node): Remove struct.
6980 (jump_table_cluster::can_be_handled): New.
6981 (case_values_threshold): Moved to header.
6982 (reset_out_edges_aux): Likewise.
6983 (jump_table_cluster::is_beneficial): New.
6984 (bit_test_cluster::can_be_handled): Likewise.
6985 (add_case_node): Remove.
6986 (bit_test_cluster::is_beneficial): New.
6987 (case_bit_test::cmp): New.
6988 (bit_test_cluster::emit): New.
6989 (expand_switch_as_decision_tree_p): Remove.
6990 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6991 (fix_phi_operands_for_edge): Likewise.
6992 (switch_decision_tree::analyze_switch_statement): New.
6993 (compute_cases_per_edge): Move ...
6994 (switch_decision_tree::compute_cases_per_edge): ... here.
6995 (try_switch_expansion): Likewise.
6996 (switch_decision_tree::try_switch_expansion): Likewise.
6997 (record_phi_operand_mapping): Likewise.
6998 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6999 (emit_case_decision_tree): Likewise.
7000 (switch_decision_tree::emit): Likewise.
7001 (balance_case_nodes): Likewise.
7002 (switch_decision_tree::balance_case_nodes): Likewise.
7003 (dump_case_nodes): Likewise.
7004 (switch_decision_tree::dump_case_nodes): Likewise.
7005 (emit_jump): Likewise.
7006 (switch_decision_tree::emit_jump): Likewise.
7007 (emit_cmp_and_jump_insns): Likewise.
7008 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7009 (emit_case_nodes): Likewise.
7010 (switch_decision_tree::emit_case_nodes): Likewise.
7011 (conditional_probability): Remove.
7012 * tree-switch-conversion.h (enum cluster_type): New.
7014 (struct cluster): Likewise.
7015 (cluster::cluster): Likewise.
7016 (struct simple_cluster): Likewise.
7017 (simple_cluster::simple_cluster): Likewise.
7018 (struct group_cluster): Likewise.
7019 (struct jump_table_cluster): Likewise.
7020 (struct bit_test_cluster): Likewise.
7021 (struct min_cluster_item): Likewise.
7022 (struct case_tree_node): Likewise.
7023 (case_tree_node::case_tree_node): Likewise.
7024 (jump_table_cluster::case_values_threshold): Likewise.
7025 (struct case_bit_test): Likewise.
7026 (struct switch_decision_tree): Likewise.
7027 (struct switch_conversion): Likewise.
7028 (switch_decision_tree::reset_out_edges_aux): Likewise.
7030 2018-06-20 Martin Liska <mliska@suse.cz>
7032 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7033 (hoist_edge_and_branch_if_true): Likewise.
7034 (expand_switch_using_bit_tests_p): Likewise.
7035 (struct case_bit_test): Likewise.
7036 (case_bit_test_cmp): Likewise.
7037 (emit_case_bit_tests): Likewise.
7038 (switch_conversion::switch_conversion): New class.
7039 (struct switch_conv_info): Remove old struct.
7040 (collect_switch_conv_info): More to ...
7041 (switch_conversion::collect): ... this.
7042 (check_range): Likewise.
7043 (switch_conversion::check_range): Likewise.
7044 (check_all_empty_except_final): Likewise.
7045 (switch_conversion::check_all_empty_except_final): Likewise.
7046 (check_final_bb): Likewise.
7047 (switch_conversion::check_final_bb): Likewise.
7048 (create_temp_arrays): Likewise.
7049 (switch_conversion::create_temp_arrays): Likewise.
7050 (free_temp_arrays): Likewise.
7051 (gather_default_values): Likewise.
7052 (switch_conversion::gather_default_values): Likewise.
7053 (build_constructors): Likewise.
7054 (switch_conversion::build_constructors): Likewise.
7055 (constructor_contains_same_values_p): Likewise.
7056 (switch_conversion::contains_same_values_p): Likewise.
7057 (array_value_type): Likewise.
7058 (switch_conversion::array_value_type): Likewise.
7059 (build_one_array): Likewise.
7060 (switch_conversion::build_one_array): Likewise.
7061 (build_arrays): Likewise.
7062 (switch_conversion::build_arrays): Likewise.
7063 (gen_def_assigns): Likewise.
7064 (switch_conversion::gen_def_assigns): Likewise.
7065 (prune_bbs): Likewise.
7066 (switch_conversion::prune_bbs): Likewise.
7067 (fix_phi_nodes): Likewise.
7068 (switch_conversion::fix_phi_nodes): Likewise.
7069 (gen_inbound_check): Likewise.
7070 (switch_conversion::gen_inbound_check): Likewise.
7071 (process_switch): Use the newly created class.
7072 (switch_conversion::expand): New.
7073 (switch_conversion::~switch_conversion): New.
7074 * tree-switch-conversion.h: New file.
7076 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7078 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7079 tree-vect-patterns.c.
7080 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7081 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7082 (vect_recog_sad_pattern): Likewise.
7083 (vect_recog_widen_sum_pattern): Likewise.
7084 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7085 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7086 (vect_recog_rotate_pattern): Likewise.
7087 (vect_recog_mult_pattern): Likewise.
7088 (vect_recog_vector_vector_shift_pattern): Likewise.
7089 (vect_recog_divmod_pattern): Likewise.
7090 (vect_recog_mixed_size_cond_pattern): Likewise.
7091 (vect_recog_bool_pattern): Likewise.
7092 (vect_recog_mask_conversion_pattern): Likewise.
7093 (vect_try_gather_scatter_pattern): Likewise.
7094 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7095 (vect_recog_over_widening_pattern): Likewise.
7096 (vect_recog_gather_scatter_pattern): Likewise.
7097 (vect_recog_func_ptr): Move from tree-vectorizer.h
7098 (vect_vect_recog_func_ptrs): Move further down the file.
7099 (vect_recog_func): Likewise. Remove the third argument.
7100 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7101 (vect_pattern_recog_1): Expect the pattern function to do any
7102 necessary target tests. Also expect it to provide a vector type.
7103 Remove the type_in handling.
7105 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7107 * tree-vect-patterns.c (vect_pattern_detected): New function.
7108 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7109 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7110 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7111 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7112 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7113 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7114 (vect_recog_mask_conversion_pattern)
7115 (vect_try_gather_scatter_pattern): Likewise.
7117 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7119 * tree-vect-patterns.c (vect_get_internal_def): New function.
7120 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7121 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7122 (search_type_for_mask_1): Use it.
7124 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7126 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7127 redundant WIDEN_SUM_EXPR handling.
7128 (vect_recog_sad_pattern): Likewise.
7130 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7132 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7133 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7134 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7135 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7136 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7138 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7140 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7141 use the stmt_vec_info of the original bb statement for the
7142 new zero assignment, even if the call is part of a pattern.
7144 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7146 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7147 that the sequence is attached to the original statement rather
7148 than the pattern statement.
7149 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7150 PATTERN_DEF_SEQ from the original statement rather than
7151 the main pattern statement.
7152 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7153 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7154 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7156 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7158 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7159 definition statements before the early exit for statements that aren't
7161 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7163 (vect_transform_loop): ...here. Process pattern definition
7164 statements without first checking whether the main pattern
7165 statement is live or relevant.
7167 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7169 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7170 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7172 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7174 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7175 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7176 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7177 (expand_block_compare): Change select_block_compare_mode call.
7178 (expand_strncmp_align_check): Use new functions, fix comment.
7179 (emit_final_str_compare_gpr): New function.
7180 (expand_strn_compare): Refactor and clean up code.
7181 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7183 2018-06-19 Tony Reix <tony.reix@atos.com>
7184 Damien Bergamini <damien.bergamini@atos.com>
7185 David Edelsohn <dje.gcc@gmail.com>
7187 * collect2.c (static_obj): New variable.
7188 (static_libs): New variable.
7189 (is_in_list): Uncomment declaration.
7190 (main): Track AIX libraries linked statically.
7191 (is_in_list): Uncomment definition.
7192 (scan_prog_file): Don't add AIX shared libraries initializer
7193 to constructor list if linking statically.
7195 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
7197 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
7199 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
7201 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
7203 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
7206 2018-06-19 Martin Liska <mliska@suse.cz>
7208 * config/i386/i386.c (ix86_can_inline_p): Do not use
7209 ipa_fn_summaries::get_create.
7210 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
7212 (devirtualization_time_bonus): Likewise.
7213 (ipcp_propagate_stage): Likewise.
7214 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7215 (edge_set_predicate): Likewise.
7216 (evaluate_conditions_for_known_args): Likewise.
7217 (evaluate_properties_for_edge): Likewise.
7218 (ipa_call_summary::reset): Tranform to ...
7219 (ipa_call_summary::~ipa_call_summary): ... this.
7220 (ipa_fn_summary::reset): Transform to ...
7221 (ipa_fn_summary::~ipa_fn_summary): ... this.
7222 (ipa_fn_summary_t::remove): Rename to ...
7223 (ipa_fn_summary_t::remove_callees): ... this.
7224 (ipa_fn_summary_t::duplicate): Use placement new
7225 instead of memory copy.
7226 (ipa_call_summary_t::duplicate): Likewise.
7227 (ipa_call_summary_t::remove): Remove.
7228 (dump_ipa_call_summary): Change get_create to get.
7229 (ipa_dump_fn_summary): Dump only when summary exists.
7230 (analyze_function_body): Use symbol_summary::get instead
7232 (compute_fn_summary): Likewise.
7233 (estimate_edge_devirt_benefit): Likewise.
7234 (estimate_edge_size_and_time): Likewise.
7235 (inline_update_callee_summaries): Likewise.
7236 (remap_edge_change_prob): Likewise.
7237 (remap_edge_summaries): Likewise.
7238 (ipa_merge_fn_summary_after_inlining): Likewise.
7239 (write_ipa_call_summary): Likewise.
7240 (ipa_fn_summary_write): Likewise.
7241 (ipa_free_fn_summary): Likewise.
7242 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
7243 (struct ipa_call_summary): Likewise.
7244 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
7246 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
7247 (estimate_size_after_inlining): Likewise.
7248 (estimate_growth): Likewise.
7249 (growth_likely_positive): Likewise.
7250 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7251 (inline_call): Likewise.
7252 * ipa-inline.c (caller_growth_limits): Likewise.
7253 (can_inline_edge_p): Likewise.
7254 (can_inline_edge_by_limits_p): Likewise.
7255 (compute_uninlined_call_time): Likewise.
7256 (compute_inlined_call_time): Likewise.
7257 (want_inline_small_function_p): Likewise.
7258 (edge_badness): Likewise.
7259 (update_caller_keys): Likewise.
7260 (update_callee_keys): Likewise.
7261 (inline_small_functions): Likewise.
7262 (inline_to_all_callers_1): Likewise.
7263 (dump_overall_stats): Likewise.
7264 (early_inline_small_functions): Likewise.
7265 (early_inliner): Likewise.
7266 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7267 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7268 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7269 * ipa-split.c (execute_split_functions): Likewise.
7270 * symbol-summary.h: Likewise.
7271 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7273 2018-06-19 Richard Biener <rguenther@suse.de>
7275 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
7276 (vectorize_loops): ... here. Fix dbgcnt handling.
7277 (try_vectorize_loop): Wrap try_vectorize_loop_1.
7279 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
7282 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
7283 ieee128 argument takes up only one (vector) register, not two (floating
7286 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7288 * gimplify.c (gimplify_init_constructor): Really never clear for an
7289 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
7291 2018-06-19 Richard Biener <rguenther@suse.de>
7293 PR tree-optimization/86179
7294 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
7295 after failed recognition.
7297 2018-06-18 Martin Sebor <msebor@redhat.com>
7300 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
7301 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
7302 Handle integer subtraction.
7303 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
7304 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
7306 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7308 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
7309 param from rtx to rtx_insn *.
7310 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
7312 (frv_ifcvt_modify_insn): Likwise.
7313 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
7314 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
7315 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
7316 as_a <rtx_insn *> cast to local "unprotected_region" once
7317 it's been established that it's not NULL or pc_rtx.
7318 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
7319 param "sethi" from rtx to rtx_insn *.
7320 (nds32_group_float_insns): Likewise for param "insn".
7321 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
7323 (vax_output_int_subtract): Likewise.
7324 * config/vax/vax.c (vax_output_int_add): Likewise for param
7326 (vax_output_int_subtract): Likewise.
7327 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
7328 (emit_pattern_after): Likewise for param "after".
7329 (emit_insn_after): Likewise.
7330 (emit_jump_insn_after): Likewise.
7331 (emit_call_insn_after): Likewise.
7332 (emit_debug_insn_after): Likewise.
7333 (emit_pattern_before): Likewise for param "before".
7334 (emit_insn_before): Likewise.
7335 (emit_jump_insn_before): Likewise.
7336 * final.c (get_insn_template): Likewise for param "insn", removing
7338 * output.h (get_insn_template): Likewise for 2nd param.
7339 * rtl.h (emit_insn_before): Likewise.
7340 (emit_jump_insn_before): Likewise.
7341 (emit_debug_insn_before_noloc): Likewise.
7342 (emit_insn_after): Likewise.
7343 (emit_jump_insn_after): Likewise.
7344 (emit_call_insn_after): Likewise.
7345 (emit_debug_insn_after): Likewise.
7346 (set_insn_deleted): Likewise for param.
7348 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
7351 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
7352 floating point modes, so that IFmode is numerically greater than
7353 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
7354 to declare the ordering. This prevents IFmode from being
7355 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
7356 machine. Include rs6000-modes.h to share the fractional values
7357 between genmodes* and the rest of the compiler.
7361 * config/rs6000/rs6000-modes.h: New file.
7362 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
7363 meaning of rs6000_long_double_size so that 126..128 selects an
7364 appropriate 128-bit floating point type.
7365 (rs6000_option_override_internal): Likewise.
7366 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
7367 (TARGET_LONG_DOUBLE_128): Change the meaning of
7368 rs6000_long_double_size so that 126..128 selects an appropriate
7369 128-bit floating point type.
7370 (LONG_DOUBLE_TYPE_SIZE): Update comment.
7371 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
7372 source and destination to match the standard usage.
7373 (truncifkf2): Likewise.
7374 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
7375 ISA 2.07 to use an explicit clobber, instead of passing in a
7377 (copysign<mode>3_soft): Likewise.
7379 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7381 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7382 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
7383 (vect_slp_analyze_instance_dependence): Likewise.
7384 (vect_enhance_data_refs_alignment): Likewise.
7385 (vect_analyze_data_refs_alignment): Likewise.
7386 (vect_slp_analyze_and_verify_instance_alignment
7387 (vect_analyze_data_ref_accesses): Likewise.
7388 (vect_prune_runtime_alias_test_list): Likewise.
7389 (vect_analyze_data_refs): Likewise.
7390 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
7391 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7392 (vect_analyze_scalar_cycles_1): Likewise.
7393 (vect_get_loop_niters): Likewise.
7394 (vect_analyze_loop_form_1): Likewise.
7395 (vect_update_vf_for_slp): Likewise.
7396 (vect_analyze_loop_operations): Likewise.
7397 (vect_analyze_loop): Likewise.
7398 (vectorizable_induction): Likewise.
7399 (vect_transform_loop): Likewise.
7400 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
7401 * tree-vect-slp.c (vect_analyze_slp): Likewise.
7402 (vect_make_slp_decision): Likewise.
7403 (vect_detect_hybrid_slp): Likewise.
7404 (vect_slp_analyze_operations): Likewise.
7405 (vect_slp_bb): Likewise.
7406 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
7407 (vectorizable_bswap): Likewise.
7408 (vectorizable_call): Likewise.
7409 (vectorizable_simd_clone_call): Likewise.
7410 (vectorizable_conversion): Likewise.
7411 (vectorizable_assignment): Likewise.
7412 (vectorizable_shift): Likewise.
7413 (vectorizable_operation): Likewise.
7414 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
7416 2018-06-18 Martin Sebor <msebor@redhat.com>
7418 PR tree-optimization/81384
7419 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
7420 * builtins.c (expand_builtin_strnlen): New function.
7421 (expand_builtin): Call it.
7422 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
7423 * builtins.def (BUILT_IN_STRNLEN): New.
7424 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
7425 Warn for bounds in excess of maximum object size.
7426 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
7427 single-value ranges. Handle strnlen.
7428 (handle_builtin_strlen): Handle strnlen.
7429 (strlen_check_and_optimize_stmt): Same.
7430 * doc/extend.texi (Other Builtins): Document strnlen.
7432 2018-06-18 Maya Rashish <coypu@sdf.org>
7434 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
7435 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
7436 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
7438 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
7440 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
7442 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7444 * tree.c (escaped_string::escape): Replace cast to char * by
7445 const_cast<char *> (unescaped).
7447 2018-06-18 Nick Clifton <nickc@redhat.com>
7450 * tree.c (escaped_string): New class. Converts an unescaped
7451 string into its escaped equivalent.
7452 (warn_deprecated_use): Use the new class to convert the
7453 deprecation message, if present.
7454 (test_escaped_strings): New self test.
7455 (test_c_tests): Add test_escaped_strings.
7456 * doc/extend.texi (deprecated): Add a note that the
7457 deprecation message is affected by the -fmessage-length
7458 option, and that control characters will be escaped.
7459 (#pragma GCC error): Document this pragma.
7460 (#pragma GCC warning): Likewise.
7461 * doc/invoke.texi (-fmessage-length): Document this option's
7462 effect on the #warning and #error preprocessor directives and
7463 the deprecated attribute.
7465 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
7467 * tree.c (decl_value_expr_lookup): Revert latest change.
7468 (decl_value_expr_insert): Likewise.
7470 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
7472 * gimplify.c (nonlocal_vlas): Delete.
7473 (nonlocal_vla_vars): Likewise.
7474 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
7476 (gimplify_body): Do not create and destroy nonlocal_vlas.
7477 * tree-nested.c: Include diagnostic.h.
7478 (use_pointer_in_frame): Tweak.
7479 (lookup_field_for_decl): Add assertion and declare the transformation.
7480 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
7481 internal error when the reference is in a wrong context. Do not
7482 create a debug decl by default.
7483 (note_nonlocal_block_vlas): Delete.
7484 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
7485 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
7486 create a debug decl by default.
7487 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
7488 call is in a wrong context.
7489 (fixup_vla_decls): New function.
7490 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
7491 debug variables were created.
7492 * tree.c (decl_value_expr_lookup): Add checking assertion.
7493 (decl_value_expr_insert): Likewise.
7495 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7498 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
7499 * tree-scalar-evolution.c (interpret_expr): Likewise.
7500 (expression_expensive_p): Likewise.
7501 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
7502 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
7503 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
7504 (ssa_defined_by_minus_one_stmt_p): New.
7506 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7509 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
7510 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
7511 * dojump.c (do_jump): Likewise.
7512 * expr.c (expand_expr_real_2): Check operand type's sign.
7513 * fold-const.c (const_unop): Handle ABSU_EXPR.
7514 (fold_abs_const): Likewise.
7515 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
7516 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
7517 (strip_sign_op_1): Likesise.
7518 * match.pd: Add new pattern to generate ABSU_EXPR.
7519 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
7520 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7521 * tree-eh.c (operation_could_trap_helper_p): Likewise.
7522 * tree-inline.c (estimate_operator_cost): Likewise.
7523 * tree-pretty-print.c (dump_generic_node): Likewise.
7524 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
7525 * tree.def (ABSU_EXPR): New.
7527 2018-06-16 Jakub Jelinek <jakub@redhat.com>
7530 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
7531 documented as preserved for backward compatibility only.
7532 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
7534 PR rtl-optimization/86108
7535 * bb-reorder.c (create_forwarder_block): Renamed to ...
7536 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
7537 jump from new landing pad to the second part.
7538 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
7541 2018-06-15 Jakub Jelinek <jakub@redhat.com>
7544 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
7545 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
7546 Only call store_expr for halves if the mode is the same.
7549 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
7550 Fix up comment formatting.
7552 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
7554 * typed-splay-tree.h (typed_splay_tree::remove): New function.
7555 (typed_splay_tree::closure,
7556 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
7557 (typed_splay_tree::typed_splay_tree,
7558 typed_splay_tree::operator =): Declared private.
7559 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
7560 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
7561 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
7562 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
7563 typed_splay_tree::splay_tree_splay,
7564 typed_splay_tree::splay_tree_foreach_helper,
7565 typed_splay_tree::splay_tree_insert,
7566 typed_splay_tree::splay_tree_remove,
7567 typed_splay_tree::splay_tree_lookup,
7568 typed_splay_tree::splay_tree_predecessor,
7569 typed_splay_tree::splay_tree_successor,
7570 typed_splay_tree::splay_tree_min,
7571 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
7572 (typed_splay_tree::root, typed_splay_tree::comp,
7573 typed_splay_tree::delete_key,
7574 typed_splay_tree::delete_value): New data members.
7575 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
7576 typed_splay_tree::remove.
7578 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
7580 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
7581 -mginv and -mno-ginv to the assembler.
7582 * config/mips/mips.opt (-mcrc): New option.
7584 * doc/invoke.text (-mcrc): Document.
7587 2018-06-15 Nick Clifton <nickc@redhat.com>
7590 * tree.c (escaped_string): New class. Converts an unescaped
7591 string into its escaped equivalent.
7592 (warn_deprecated_use): Use the new class to convert the
7593 deprecation message, if present.
7594 (test_escaped_strings): New self test.
7595 (test_c_tests): Add test_escaped_strings.
7596 * doc/extend.texi (deprecated): Add a note that the
7597 deprecation message is affected by the -fmessage-length
7598 option, and that control characters will be escaped.
7599 (#pragma GCC error): Document this pragma.
7600 (#pragma GCC warning): Likewise.
7601 * doc/invoke.texi (-fmessage-length): Document this option's
7602 effect on the #warning and #error preprocessor directives and
7603 the deprecated attribute.
7605 2018-06-15 Richard Biener <rguenther@suse.de>
7607 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
7608 here, also noting vector size used.
7609 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
7610 size used in MSG_OPTIMIZED_LOCATIONS dump.
7611 (pass_slp_vectorize::execute): Adjust.
7613 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
7616 * config/arc/arc.c (arc_return_address_register): Fix
7619 2018-06-15 Richard Biener <rguenther@suse.de>
7622 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
7623 leave useless conversion stripping to force_gimple_operand_gsi.
7624 (gimplify_build2): Likewise.
7625 (gimplify_build1): Likewise.
7627 2018-06-15 Richard Biener <rguenther@suse.de>
7630 * tree-cfg.c (move_stmt_op): unshare invariant addresses
7631 before adjusting their block.
7633 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
7635 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
7636 multilibs for *-*-rtems*.
7637 * config/riscv/t-rtems: New file.
7639 2018-06-14 Jakub Jelinek <jakub@redhat.com>
7642 * match.pd ((A +- CST1) +- CST2): Punt if last resort
7643 unsigned_type_for returns NULL.
7646 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
7647 subregs of multi-word pseudos unless the float mode has word size.
7649 2018-06-14 Richard Biener <rguenther@suse.de>
7652 * tree-vect-generic.c (build_word_mode_vector_type): Remove
7653 duplicate and harmful type_hash_canon.
7654 * tree.c (type_hash_canon): Assert we didn't find ourselves.
7656 2018-06-14 Richard Biener <rguenther@suse.de>
7659 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
7662 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
7664 * config/rtems.h (STDINT_LONG32): Define.
7666 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
7667 Prachi Godbole <prachi.godbole@imgtec.com>
7669 * config/mips/mips-cpus.def: Define P6600.
7670 * config/mips/mips-tables.opt: Regenerate.
7671 * config/mips/mips.c (mips_ucbranch_type): New enum.
7672 (mips_rtx_cost_data): Add support for P6600.
7673 (mips_issue_rate): Likewise.
7674 (mips_multipass_dfa_lookahead): Likewise.
7675 (mips_avoid_hazard): Likewise.
7676 (mips_reorg_process_insns): Likewise.
7677 (mips_classify_branch_p6600): New function.
7678 * config/mips/mips.h (TUNE_P6600): New define.
7679 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
7680 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
7681 * config/mips/mips.md: Include p6600.md.
7682 (processor): Add p6600.
7683 * config/mips/p6600.md: New file.
7684 * doc/invoke.texi: Add p6600 to supported architectures.
7686 2018-06-13 Martin Sebor <msebor@redhat.com>
7688 PR tree-optimization/86114
7689 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
7691 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
7693 2018-06-13 Richard Biener <rguenther@suse.de>
7695 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
7696 Properly set vector type of the intermediate stmt.
7697 * tree-vect-stmts.c (vectorizable_operation): The destination
7698 var always has vectype_out type.
7700 2018-06-13 Jeff Law <law@redhat.com>
7702 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
7703 integer 0 for argument to print_rtl_with_bb.
7704 (rl78_reorg): Likewise.
7706 2018-06-13 David Malcolm <dmalcolm@redhat.com>
7708 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
7709 from rtx to rtx_insn *.
7710 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
7712 (add_sched_insns_for_speculation): Likewise for local "target",
7713 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7714 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
7715 from rtx_insn ** to rtx_code_label **.
7716 (reorg_emit_nops): Likewise.
7717 (c6x_reorg): Likewise for local "call_labels".
7718 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
7720 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
7721 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
7722 the loops over LABEL_REFS.
7723 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
7725 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7726 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
7727 (split_branches): Strengthen local "olabel" from rtx to
7728 rtx_insn *, adding a safe_as_a cast.
7729 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
7731 (add_insn_after): Likewise for first two params.
7732 (add_insn_before): Likewise.
7733 (remove_insn): Likewise for param.
7734 (emit_pattern_before_noloc): Likewise for second and third params.
7735 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
7736 (emit_call_insn_before_noloc): Likewise.
7737 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
7739 (emit_barrier_before): Likewise.
7740 (emit_label_before): Strengthen "label" param from "rtx" to
7741 "rtx_code_label *". Strengthen "before" param from "rtx" to
7743 (emit_insn_after_1): Strengthen "after" param from "rtx" to
7745 (emit_pattern_after_noloc): Likewise.
7746 (emit_insn_after_noloc): Likewise.
7747 (emit_jump_insn_after_noloc): Likewise.
7748 (emit_call_insn_after_noloc): Likewise.
7749 (emit_debug_insn_after_noloc): Likewise.
7750 (emit_barrier_after): Likewise.
7751 (emit_label_after): Likewise for both params.
7752 (emit_pattern_after_setloc): Likewise for "after" param. Convert
7753 "loc" param from "int" to "location_t".
7754 (emit_insn_after_setloc): Likewise.
7755 (emit_jump_insn_after_setloc): Likewise.
7756 (emit_call_insn_after_setloc): Likewise.
7757 (emit_debug_insn_after_setloc): Likewise.
7758 (emit_pattern_before_setloc): Likewise for "before" param. Convert
7759 "loc" param from "int" to "location_t".
7760 (emit_pattern_before): Convert NULL_RTX to NULL.
7761 (emit_insn_before_setloc): Convert "loc" param from "int" to
7763 (emit_jump_insn_before_setloc): Likewise.
7764 (emit_call_insn_before_setloc): Likewise.
7765 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
7766 rtx_insn *. Convert "loc" param from "int" to "location_t".
7767 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
7768 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
7769 Convert 3rd param from "int" to "location_t".
7770 (emit_barrier_before, emit_barrier_after, next_real_insn):
7771 Strengthen param from rtx to rtx_insn *.
7772 (emit_label_before): Strengthen 1st param from "rtx" to
7773 "rtx_code_label *". Strengthen 2nd param from "rtx" to
7775 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
7776 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
7777 Strengthen 2nd param from "rtx" to "rtx_insn *".
7778 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
7779 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
7780 Likewise. Convert 3rd param from "int" to "location_t".
7781 (emit_label_after): Strengthen 1st param from "rtx" to
7783 (next_real_insn, remove_insn): Strengthen param from "rtx" to
7785 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
7786 from "rtx" to "rtx_insn *".
7788 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
7790 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
7791 bodies streamed in with -Q.
7792 * dumpfile.c (dump_files): Add lto-stream-out dump file.
7793 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7794 * gimple-streamer-out.c: Include gimple-pretty-print.h
7795 (output_bb): Dump stmts streamed.
7796 * lto-section-out.c: Include print-tree.h
7797 (lto_begin_section): Dump sections created.
7798 (lto_output_decl_index): Dump decl encoded.
7799 * lto-streamer-out.c: Include print-tree.h
7800 (create_output_block): Dump output block created.
7801 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
7802 (output_function): Dump function output.
7803 (output_constructor): Dump constructor streamed.
7804 (write_global_stream): Output indexes encoded.
7805 (produce_asm_for_decls): Dump streams encoded.
7806 * lto-streamer.c (streamer_dump_file): New global var.
7807 * lto-streamer.h (streamer_dump_file): Declare.
7808 * passes.c (ipa_write_summaries): Initialize streamer dump.
7809 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
7812 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
7815 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
7816 offsets for register save directives. Emit a second batch of save
7817 directives, if need be, when the function accesses prior frames.
7819 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7821 * config/arc/fpu.md (fmasf4): Force operand to register.
7822 (fnmasf4): Likewise.
7824 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7826 * config/arc/arc-protos.h (arc_pad_return): Remove.
7827 * config/arc/arc.c (machine_function): Remove force_short_suffix
7829 (arc_print_operand): Adjust printing of '&'.
7830 (arc_verify_short): Remove conditional printing of short suffix.
7831 (arc_final_prescan_insn): Remove reference to size_reason.
7832 (pad_return): New function.
7833 (arc_reorg): Call pad_return.
7834 (arc_pad_return): Remove.
7835 (arc_init_machine_status): Remove reference to force_short_suffix.
7836 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
7837 (attr length): When attribute iscompact is true force to 2
7838 regardless; in the case of maybe check if we want to force the
7839 instruction to have 4 bytes length.
7840 (nopv): Change it to generate 4 byte long nop as well.
7841 (blockage): New pattern.
7842 (simple_return): Remove call to arc_pad_return.
7843 (p_return_i): Likewise.
7845 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7847 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
7849 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7851 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
7854 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7856 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
7857 for ARC700 and ARCv2.
7859 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
7862 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
7863 operands[2] instead of operands[1].
7866 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7868 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
7869 case, check whether the outer register overlaps an unallocatable
7870 register, not just whether it fits the required class.
7872 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7874 * poly-int.h (can_div_trunc_p): Add new overload in which all values
7876 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
7877 (memrefs_conflict_p): Likewise.
7878 (init_alias_analysis): Likewise.
7879 * cfgexpand.c (expand_debug_expr): Likewise.
7880 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
7881 * cse.c (fold_rtx): Likewise.
7882 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
7883 * expr.c (emit_block_move_hints): Likewise.
7884 (clear_storage_hints, push_block, emit_push_insn): Likewise.
7885 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
7886 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
7887 (emit_group_store): Likewise.
7888 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
7889 to read the PRE/POST_MODIFY increment.
7890 * calls.c (store_one_arg): Use strip_offset.
7891 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
7893 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
7895 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
7896 (simplify_binary_operation_1): Extend CONST_INT handling to
7898 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
7899 than a HOST_WIDE_INT.
7900 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
7902 (adjust_mems, add_stores): Update accodingly.
7903 (vt_canonicalize_addr): Track polynomial offsets.
7904 (emit_note_insn_var_location): Likewise.
7905 (vt_add_function_parameter): Likewise.
7906 (vt_initialize): Likewise.
7908 2018-06-12 Jeff Law <law@redhat.com>
7910 * config.gcc (alpha*-*-freebsd*): Remove.
7911 * config/alpha/freebsd.h: Remove.
7913 2018-06-12 David Malcolm <dmalcolm@redhat.com>
7916 * spellcheck-tree.c (levenshtein_distance): Rename to...
7917 (get_edit_distance): ...this, and update for underlying renaming.
7918 * spellcheck-tree.h (levenshtein_distance): Rename to...
7919 (get_edit_distance): ...this.
7920 * spellcheck.c (levenshtein_distance): Rename to...
7921 (get_edit_distance): ...this. Convert from Levenshtein distance
7922 to Damerau-Levenshtein distance by supporting transpositions of
7923 adjacent characters. Rename "v1" to "v_next" and "v0" to
7925 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
7926 (selftest::test_edit_distance_unit_test_oneway): ...this, and
7927 update for underlying renaming.
7928 (selftest::levenshtein_distance_unit_test): Rename to...
7929 (selftest::test_get_edit_distance_unit): ...this, and update for
7930 underlying renaming.
7931 (selftest::test_find_closest_string): Add example from PR 69968
7932 where transposition helps
7933 (selftest::test_metric_conditions): Update for renaming.
7934 (selftest::test_metric_conditions): Likewise.
7935 (selftest::spellcheck_c_tests): Likewise.
7936 * spellcheck.h (levenshtein_distance): Rename both overloads to...
7937 (get_edit_distance): ...this.
7938 (best_match::consider): Update for renaming.
7940 2018-06-12 Martin Sebor <msebor@redhat.com>
7942 PR tree-optimization/85259
7943 * builtins.c (compute_objsize): Handle constant offsets.
7944 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
7945 true iff a warning has been issued.
7946 * gimple.h (gimple_nonartificial_location): New function.
7947 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
7948 gimple_nonartificial_location and handle -Wno-system-headers.
7949 (handle_builtin_stxncpy): Same.
7951 2018-06-12 Martin Sebor <msebor@redhat.com>
7954 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7956 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7958 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7959 BUILTIN_VEC_XST entries for pointer to double and long long.
7961 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7964 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7966 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7969 2018-06-12 Martin Liska <mliska@suse.cz>
7971 * doc/options.texi: Document IntegerRange.
7973 2018-06-12 Martin Liska <mliska@suse.cz>
7975 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7976 * opt-functions.awk: Handle Deprecated flag.
7977 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7979 (read_cmdline_option): Report warning for a deprecated option.
7980 * opts.h (struct cl_option): Add new field cl_deprecated.
7981 (CL_ERR_DEPRECATED): New.
7983 2018-06-12 Martin Liska <mliska@suse.cz>
7985 * doc/options.texi: Document Deprecated option flag.
7987 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7989 * config/arc/arc-arch.h (arc_extras): New enum.
7990 (arc_cpu_t):Add field extra.
7991 (arc_cpu_types): Consider the extras.
7992 * config/arc/arc-cpus.def: Add extras info.
7993 * config/arc/arc-opts.h (processor_type): Consider extra field.
7994 * config/arc/arc.c (arc_override_options): Handle extra field.
7996 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7998 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7999 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8001 * config/arc/arc.c (arc_init): Update pic warning.
8002 (irq_range): Update irq range parsing warnings.
8003 (arc_override_options): Update various warning messages.
8004 (arc_handle_aux_attribute): Likewise.
8006 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8008 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8010 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8012 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8013 compiler messages directives.
8015 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8017 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8018 * config/mips/mips-tables.opt: Regenerate.
8019 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8021 * doc/invoke.texi: Document -march=i6500.
8023 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8025 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8026 (i6400_gpmul): Add cpu_unit.
8027 (i6400_gpdiv): Likewise.
8028 (i6400_msa_add_d): Update reservations.
8029 (i6400_msa_int_add) Likewise.
8030 (i6400_msa_short_logic3) Likewise.
8031 (i6400_msa_short_logic2) Likewise.
8032 (i6400_msa_short_logic) Likewise.
8033 (i6400_msa_move) Likewise.
8034 (i6400_msa_cmp) Likewise.
8035 (i6400_msa_short_float2) Likewise.
8036 (i6400_msa_div_d) Likewise.
8037 (i6400_msa_long_logic1) Likewise.
8038 (i6400_msa_long_logic2) Likewise.
8039 (i6400_msa_mult) Likewise.
8040 (i6400_msa_long_float2) Likewise.
8041 (i6400_msa_long_float4) Likewise.
8042 (i6400_msa_long_float5) Likewise.
8043 (i6400_msa_long_float8) Likewise.
8044 (i6400_fpu_fadd): Include frint type.
8045 (i6400_fpu_store): New define_insn_reservation.
8046 (i6400_fpu_load): Likewise.
8047 (i6400_fpu_move): Likewise.
8048 (i6400_fpu_fcmp): Likewise.
8049 (i6400_fpu_fmadd): Likewise.
8050 (i6400_int_mult): Include imul3nc type and update reservation.
8051 (i6400_int_div): Include idiv3 type and update reservation.
8052 (i6400_int_load): Update to check type not move_type.
8053 (i6400_int_store): Likewise.
8054 (i6400_int_prefetch): Set zero latency.
8056 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8058 * gcc.c: Document new %@{...} sequence.
8059 (LINK_COMMAND_SPEC): Use it for the -L switches.
8060 (cpp_unique_options): Use it for the -I switches.
8061 (at_file_argbuf): New global variable.
8062 (in_at_file): Likewise.
8063 (alloc_args): Create at_file_argbuf.
8064 (clear_args): Truncate at_file_argbuf.
8065 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8066 (open_at_file): New function.
8067 (close_at_file): Likewise.
8068 (create_at_file): Delete.
8069 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8072 (validate_switches_from_spec): Deal with %@{...} sequence.
8073 (validate_switches): Likewise.
8074 (driver::finalize): Call clear_args.
8076 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8078 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8080 2018-06-11 Martin Sebor <msebor@redhat.com>
8082 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8083 (Wno-class-memaccess): Add @opindex.
8084 (Wno-templates, Wno-multiple-inheritance): Same.
8085 (Wno-virtual-inheritance, Wno-namespaces): Same.
8086 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8087 (Wno-format-overflow, Wno-format-truncation): Same.
8088 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8089 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8090 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8091 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8092 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8093 (Wno-misspelled-isr): Same.
8095 2018-06-11 Martin Sebor <msebor@redhat.com>
8097 * PR tree-optimization/86083
8098 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8100 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8102 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8104 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8107 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8108 on the correct operand.
8109 (*movdi_internal64): Ditto.
8111 2018-06-11 Martin Liska <mliska@suse.cz>
8113 PR tree-optimization/86089
8114 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8116 2018-06-11 Julia Koval <julia.koval@intel.com>
8118 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8119 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8120 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8122 2018-06-11 Olivier Hainque <hainque@adacore.com>
8124 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8125 for Ada with strict dwarf2.
8127 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8130 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8133 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8135 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8137 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8139 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8142 2018-06-08 Martin Liska <mliska@suse.cz>
8144 * tree-cfg.h (debug_function): Fix argument type to match
8147 2018-06-08 Martin Liska <mliska@suse.cz>
8149 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8150 Remove usage of MPX-related (and removed) fields.
8151 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8153 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8155 * cfg.c (debug): Use TDF_NONE rather than 0.
8156 * cfghooks.c (debug): Likewise.
8157 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8158 (struct dump_option_value_info): Convert to...
8159 (struct kv_pair): ...this template type.
8160 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8162 (optinfo_verbosity_options): Likewise.
8163 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8165 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8166 than int for "optgroup_flags" param.
8167 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8169 (dump_dec): Likewise.
8170 (dump_finish): Use TDF_NONE rather than 0.
8171 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8172 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8173 than 0. Update for change to option_ptr.
8174 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8175 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8176 0. Update for changes to optinfo_verbosity_options and
8178 (opt_info_switch_p): Convert optgroup_flags from int to
8180 (dump_basic_block): Use dump_flags_t rather than int
8181 for "dump_kind" param.
8182 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8183 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8184 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8185 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8186 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8187 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8188 TDF_NONE): Convert from macros to...
8189 (enum dump_flag): ...this new enum.
8190 (dump_flags_t): Update to use enum.
8191 (operator|, operator&, operator~, operator|=, operator&=):
8192 Implement for dump_flags_t.
8193 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
8194 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
8195 Convert from macros to...
8196 (enum optgroup_flag): ...this new enum.
8197 (optgroup_flags_t): New typedef.
8198 (operator|, operator|=): Implement for optgroup_flags_t.
8199 (struct dump_file_info): Convert field "alt_flags" to
8200 dump_flags_t. Convert field "optgroup_flags" to
8202 (dump_basic_block): Use dump_flags_t rather than int for param.
8203 (dump_generic_expr_loc): Likewise.
8204 (dump_dec): Likewise.
8205 (dump_register): Convert param "optgroup_flags" to
8207 (opt_info_enable_passes): Likewise.
8208 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
8210 * gimple-pretty-print.c (debug): Likewise.
8211 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
8212 (merged_store_group::apply_stores): Likewise.
8213 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
8214 * gimple.c (verify_gimple_pp): Likewise.
8215 * graphite-poly.c (print_pbb_body): Likewise.
8216 * passes.c (pass_manager::register_one_dump_file): Convert
8217 local "optgroup_flags" to optgroup_flags_t.
8218 * print-tree.c (print_node): Use TDF_NONE rather than 0.
8220 (debug_body): Likewise.
8221 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
8222 to optgroup_flags_t.
8223 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
8225 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
8226 (convert_mult_to_fma): Likewise.
8227 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
8228 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
8229 * tree-vect-data-refs.c (dump_lower_bound): Convert param
8230 "dump_kind" to dump_flags_t.
8232 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8234 * config/rs6000/rs6000.c (min, max): Delete.
8236 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8238 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
8239 -mabi=spe and -mabi=no-spe.
8241 2018-06-08 Martin Liska <mliska@suse.cz>
8243 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
8244 where we expect an existing summary.
8246 2018-06-08 Martin Liska <mliska@suse.cz>
8248 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
8249 * ipa-inline.h (estimate_edge_growth): Likewise.
8251 2018-06-08 Martin Liska <mliska@suse.cz>
8253 * cgraph.c (function_version_hasher::hash): Use
8254 cgraph_node::get_uid ().
8255 (function_version_hasher::equal):
8256 * cgraph.h (cgraph_node::get_uid): New method.
8257 * ipa-inline.c (update_caller_keys): Use
8258 cgraph_node::get_uid ().
8259 (update_callee_keys): Likewise.
8260 * ipa-utils.c (searchc): Likewise.
8261 (ipa_reduced_postorder): Likewise.
8262 * lto-cgraph.c (input_node): Likewise.
8263 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
8264 * symbol-summary.h (symtab_insertion): Likewise.
8265 (symtab_removal): Likewise.
8266 (symtab_duplication): Likewise.
8267 * tree-pretty-print.c (dump_function_header): Likewise.
8268 * tree-sra.c (convert_callers_for_node): Likewise.
8270 2018-06-08 Martin Liska <mliska@suse.cz>
8272 * cgraph.c (symbol_table::create_edge): Always assign a new
8274 (symbol_table::free_edge): Do not recycle numbers.
8275 * cgraph.h (cgraph_edge::get): New method.
8276 * symbol-summary.h (symtab_removal): Use it.
8277 (symtab_duplication): Likewise.
8278 (call_summary::hashable_uid): Remove.
8280 2018-06-08 Martin Liska <mliska@suse.cz>
8282 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
8283 (initialize_growth_caches): Remove.
8284 (free_growth_caches): Likewise.
8285 (do_estimate_edge_time): Use edge_growth_cache.
8286 (do_estimate_edge_size): Likewise.
8287 (do_estimate_edge_hints): Likewise.
8288 * ipa-inline.c (reset_edge_caches): Likewise.
8289 (recursive_inlining): Likewise.
8290 (inline_small_functions): Likewise.
8291 * ipa-inline.h (initialize_growth_caches): Remove.
8292 (estimate_edge_size): Likewise.
8293 (estimate_edge_time): Likewise.
8294 (estimate_edge_hints): Likewise.
8295 (reset_edge_growth_cache): Likewise.
8296 * symbol-summary.h (call_summary::remove): New method.
8298 2018-06-08 Martin Liska <mliska@suse.cz>
8300 * ipa-cp.c (class edge_clone_summary): New summary.
8301 (grow_edge_clone_vectors): Remove.
8302 (ipcp_edge_duplication_hook): Remove.
8303 (class edge_clone_summary_t): New call_summary class.
8304 (ipcp_edge_removal_hook): Remove.
8305 (edge_clone_summary_t::duplicate): New function.
8306 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
8307 (create_specialized_node): Likewise.
8308 (ipcp_driver): Initialize edge_clone_summaries and do not
8311 2018-06-08 Martin Liska <mliska@suse.cz>
8313 * symbol-summary.h (get): New function.
8314 (call_summary::m_initialize_when_cloning): New class member.
8316 2018-06-08 Martin Liska <mliska@suse.cz>
8318 * cgraph.c (cgraph_node::remove): Do not recycle uid.
8319 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
8320 (symbol_table::allocate_cgraph_symbol): Do not set uid.
8321 * passes.c (uid_hash_t): Record removed_nodes by their uids.
8322 (remove_cgraph_node_from_order): Use the removed_nodes set.
8323 (do_per_function_toporder): Likwise.
8324 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
8325 instead of summary_uid.
8326 (symtab_removal): Likewise.
8327 (symtab_duplication): Likewise.
8329 2018-06-08 Martin Liska <mliska@suse.cz>
8331 * ipa-cp.c (ipcp_store_bits_results): Use
8332 ipcp_transformation_sum.
8333 (ipcp_store_vr_results): Likewise.
8334 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
8336 (ipcp_transformation_initialize): ... this.
8337 (ipa_set_node_agg_value_chain):
8338 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
8339 (write_ipcp_transformation_info): Likewise.
8340 (read_ipcp_transformation_info): Likewise.
8341 (ipcp_update_bits): Likewise.
8342 (ipcp_update_vr): Likewise.
8343 (ipcp_transform_function): Likewise.
8344 * ipa-prop.h: Rename ipcp_transformation_summary to
8345 ipcp_transformation.
8346 (class ipcp_transformation_t): New function summary.
8347 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
8348 (ipa_get_agg_replacements_for_node): Likewise.
8350 2018-06-08 Martin Liska <mliska@suse.cz>
8352 * ipa-pure-const.c (struct funct_state_d): Do it class instead
8354 (class funct_state_summary_t): New function_summary class.
8355 (has_function_state): Remove.
8356 (get_function_state): Likewise.
8357 (set_function_state): Likewise.
8358 (add_new_function): Likewise.
8359 (funct_state_summary_t::insert): New function.
8360 (duplicate_node_data): Remove.
8361 (remove_node_data): Remove.
8362 (funct_state_summary_t::duplicate): New function.
8363 (register_hooks): Create new funct_state_summaries.
8364 (pure_const_generate_summary): Use it.
8365 (pure_const_write_summary): Likewise.
8366 (pure_const_read_summary): Likewise.
8367 (propagate_pure_const): Likewise.
8368 (propagate_nothrow): Likewise.
8369 (dump_malloc_lattice): Likewise.
8370 (propagate_malloc): Likewise.
8371 (execute): Do not register hooks, just remove summary
8373 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
8376 2018-06-08 Martin Liska <mliska@suse.cz>
8378 * ipa-reference.c (remove_node_data): Remove.
8379 (duplicate_node_data): Likewise.
8380 (class ipa_ref_var_info_summary_t): New class.
8381 (class ipa_ref_opt_summary_t): Likewise.
8382 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
8383 (get_reference_optimization_summary): Use
8384 ipa_ref_opt_sum_summaries.
8385 (set_reference_vars_info): Remove.
8386 (set_reference_optimization_summary): Likewise.
8387 (ipa_init): Create summaries.
8388 (init_function_info): Use function summary.
8389 (ipa_ref_opt_summary_t::duplicate): New function.
8390 (ipa_ref_opt_summary_t::remove): New function.
8391 (get_read_write_all_from_node): Fix GNU coding style.
8392 (propagate): Use function summary.
8393 (write_node_summary_p): Fix GNU coding style.
8394 (stream_out_bitmap): Likewise.
8395 (ipa_reference_read_optimization_summary): Use function summary.
8396 (ipa_reference_c_finalize): Do not release hooks.
8398 2018-06-08 Martin Liska <mliska@suse.cz>
8400 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
8401 (analyze_function_body): Extract multiple calls of get_create.
8402 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8403 * ipa-inline.c (recursive_inlining): Use ::get method.
8404 * ipa-inline.h (estimate_edge_growth): Likewise.
8406 2018-06-08 Martin Liska <mliska@suse.cz>
8408 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
8410 (hsa_function_summary::hsa_function_summary): Use the new enum
8412 (hsa_gpu_implementation_p): Use hsa_summaries::get.
8413 * hsa-gen.c (hsa_get_host_function): Likewise.
8414 (get_brig_function_name): Likewise.
8415 * ipa-hsa.c (process_hsa_functions): Likewise.
8416 (ipa_hsa_write_summary): Likewise.
8417 * symbol-summary.h (symtab_duplication): Use ::get function/
8418 (get): New function.
8420 2018-06-08 Martin Liska <mliska@suse.cz>
8422 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
8424 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
8425 (hsa_register_kernel): Likewise.
8426 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
8427 * hsa-gen.c (hsa_get_host_function): Likewise.
8428 (get_brig_function_name): Likewise.
8429 (generate_hsa): Likewise.
8430 (pass_gen_hsail::execute): Likewise.
8431 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
8432 (devirtualization_time_bonus): Likewise.
8433 (ipcp_propagate_stage): Likewise.
8434 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8435 (edge_set_predicate): Likewise.
8436 (evaluate_conditions_for_known_args): Likewise.
8437 (evaluate_properties_for_edge): Likewise.
8438 (ipa_fn_summary::reset): Likewise.
8439 (ipa_fn_summary_t::duplicate): Likewise.
8440 (dump_ipa_call_summary): Likewise.
8441 (ipa_dump_fn_summary): Likewise.
8442 (analyze_function_body): Likewise.
8443 (compute_fn_summary): Likewise.
8444 (estimate_edge_devirt_benefit): Likewise.
8445 (estimate_edge_size_and_time): Likewise.
8446 (estimate_calls_size_and_time): Likewise.
8447 (estimate_node_size_and_time): Likewise.
8448 (inline_update_callee_summaries): Likewise.
8449 (remap_edge_change_prob): Likewise.
8450 (remap_edge_summaries): Likewise.
8451 (ipa_merge_fn_summary_after_inlining): Likewise.
8452 (ipa_update_overall_fn_summary): Likewise.
8453 (read_ipa_call_summary): Likewise.
8454 (inline_read_section): Likewise.
8455 (write_ipa_call_summary): Likewise.
8456 (ipa_fn_summary_write): Likewise.
8457 (ipa_free_fn_summary): Likewise.
8458 * ipa-hsa.c (process_hsa_functions): Likewise.
8459 (ipa_hsa_write_summary): Likewise.
8460 (ipa_hsa_read_section): Likewise.
8461 * ipa-icf.c (sem_function::merge): Likewise.
8462 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8463 (do_estimate_edge_time): Likewise.
8464 (estimate_size_after_inlining): Likewise.
8465 (estimate_growth): Likewise.
8466 (growth_likely_positive): Likewise.
8467 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8468 (inline_call): Likewise.
8469 * ipa-inline.c (caller_growth_limits): Likewise.
8470 (can_inline_edge_p): Likewise.
8471 (can_inline_edge_by_limits_p): Likewise.
8472 (compute_uninlined_call_time): Likewise.
8473 (compute_inlined_call_time): Likewise.
8474 (want_inline_small_function_p): Likewise.
8475 (edge_badness): Likewise.
8476 (update_caller_keys): Likewise.
8477 (update_callee_keys): Likewise.
8478 (recursive_inlining): Likewise.
8479 (inline_small_functions): Likewise.
8480 (inline_to_all_callers_1): Likewise.
8481 (dump_overall_stats): Likewise.
8482 (early_inline_small_functions): Likewise.
8483 (early_inliner): Likewise.
8484 * ipa-inline.h (estimate_edge_growth): Likewise.
8485 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8486 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8487 * ipa-prop.h (IPA_NODE_REF): Likewise.
8488 (IPA_EDGE_REF): Likewise.
8489 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8490 (propagate_malloc): Likewise.
8491 * ipa-split.c (execute_split_functions): Likewise.
8492 * symbol-summary.h: Rename get to get_create.
8494 (get_create): Likewise.
8495 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8497 2018-06-08 Martin Liska <mliska@suse.cz>
8499 * symbol-summary.h (release): Move definition out of class
8501 (symtab_removal): Likewise.
8502 (symtab_duplication): Likewise.
8504 2018-06-08 Martin Liska <mliska@suse.cz>
8506 * symbol-summary.h (function_summary): Move constructor
8507 implementation out of class declaration.
8508 (release): Likewise.
8509 (symtab_insertion): Likewise.
8510 (symtab_removal): Likewise.
8511 (symtab_duplication): Likewise.
8514 2018-06-08 Martin Liska <mliska@suse.cz>
8516 * Makefile.in: Remove support for MPX (macros, related functions,
8517 fields in cgraph_node, ...).
8518 * builtin-types.def (BT_BND): Likewise.
8519 (BT_FN_BND_CONST_PTR): Likewise.
8520 (BT_FN_CONST_PTR_BND): Likewise.
8521 (BT_FN_VOID_PTR_BND): Likewise.
8522 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
8523 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
8524 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
8525 (expand_builtin_mempcpy_with_bounds): Likewise.
8526 (expand_builtin_memset_with_bounds): Likewise.
8527 (expand_builtin_memset_args): Likewise.
8528 (std_expand_builtin_va_start): Likewise.
8529 (expand_builtin): Likewise.
8530 (expand_builtin_with_bounds): Likewise.
8531 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
8532 (DEF_LIB_BUILTIN_CHKP): Likewise.
8533 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
8534 (DEF_CHKP_BUILTIN): Likewise.
8535 (BUILT_IN_MEMCPY): Likewise.
8536 (BUILT_IN_MEMMOVE): Likewise.
8537 (BUILT_IN_MEMPCPY): Likewise.
8538 (BUILT_IN_MEMSET): Likewise.
8539 (BUILT_IN_STPCPY): Likewise.
8540 (BUILT_IN_STRCAT): Likewise.
8541 (BUILT_IN_STRCHR): Likewise.
8542 (BUILT_IN_STRCPY): Likewise.
8543 (BUILT_IN_STRLEN): Likewise.
8544 (BUILT_IN_MEMCPY_CHK): Likewise.
8545 (BUILT_IN_MEMMOVE_CHK): Likewise.
8546 (BUILT_IN_MEMPCPY_CHK): Likewise.
8547 (BUILT_IN_MEMSET_CHK): Likewise.
8548 (BUILT_IN_STPCPY_CHK): Likewise.
8549 (BUILT_IN_STRCAT_CHK): Likewise.
8550 (BUILT_IN_STRCPY_CHK): Likewise.
8551 * calls.c (store_bounds): Likewise.
8552 (emit_call_1): Likewise.
8553 (special_function_p): Likewise.
8554 (maybe_warn_nonstring_arg): Likewise.
8555 (initialize_argument_information): Likewise.
8556 (finalize_must_preallocate): Likewise.
8557 (compute_argument_addresses): Likewise.
8558 (expand_call): Likewise.
8559 * cfgexpand.c (expand_call_stmt): Likewise.
8560 (expand_return): Likewise.
8561 (expand_gimple_stmt_1): Likewise.
8562 (pass_expand::execute): Likewise.
8563 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
8564 (cgraph_node::remove): Likewise.
8565 (cgraph_node::dump): Likewise.
8566 (cgraph_node::verify_node): Likewise.
8567 * cgraph.h (chkp_function_instrumented_p): Likewise.
8568 (symtab_node::get_alias_target): Likewise.
8569 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
8570 (cgraph_local_p): Likewise.
8571 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
8572 (cgraph_edge::rebuild_references): Likewise.
8573 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
8574 (walk_polymorphic_call_targets): Likewise.
8575 (cgraph_node::expand_thunk): Likewise.
8576 (symbol_table::output_weakrefs): Likewise.
8577 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
8578 (ix86_handle_option): Likewise.
8579 * config/i386/constraints.md: Likewise.
8580 * config/i386/i386-builtin-types.def (BND): Likewise.
8584 * config/i386/i386-builtin.def (BDESC_END): Likewise.
8585 (BDESC_FIRST): Likewise.
8587 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
8588 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
8589 * config/i386/i386.c (enum reg_class): Likewise.
8590 (ix86_target_string): Likewise.
8591 (ix86_option_override_internal): Likewise.
8592 (ix86_conditional_register_usage): Likewise.
8593 (ix86_valid_target_attribute_inner_p): Likewise.
8594 (ix86_set_indirect_branch_type): Likewise.
8595 (ix86_set_current_function): Likewise.
8596 (ix86_function_arg_regno_p): Likewise.
8597 (init_cumulative_args): Likewise.
8598 (ix86_function_arg_advance): Likewise.
8599 (ix86_function_arg): Likewise.
8600 (ix86_pass_by_reference): Likewise.
8601 (ix86_function_value_regno_p): Likewise.
8602 (ix86_function_value_1): Likewise.
8603 (ix86_function_value_bounds): Likewise.
8604 (ix86_return_in_memory): Likewise.
8605 (ix86_setup_incoming_vararg_bounds): Likewise.
8606 (ix86_va_start): Likewise.
8607 (indirect_thunk_need_prefix): Likewise.
8608 (print_reg): Likewise.
8609 (ix86_print_operand): Likewise.
8610 (ix86_expand_call): Likewise.
8611 (ix86_output_function_return): Likewise.
8612 (reg_encoded_number): Likewise.
8613 (BDESC_VERIFYS): Likewise.
8614 (ix86_init_mpx_builtins): Likewise.
8615 (ix86_init_builtins): Likewise.
8616 (ix86_emit_cmove): Likewise.
8617 (ix86_emit_move_max): Likewise.
8618 (ix86_expand_builtin): Likewise.
8619 (ix86_builtin_mpx_function): Likewise.
8620 (ix86_get_arg_address_for_bt): Likewise.
8621 (ix86_load_bounds): Likewise.
8622 (ix86_store_bounds): Likewise.
8623 (ix86_load_returned_bounds): Likewise.
8624 (ix86_store_returned_bounds): Likewise.
8625 (ix86_class_likely_spilled_p): Likewise.
8626 (ix86_hard_regno_mode_ok): Likewise.
8627 (x86_order_regs_for_local_alloc): Likewise.
8628 (ix86_mitigate_rop): Likewise.
8629 (ix86_bnd_prefixed_insn_p): Likewise.
8630 (ix86_mpx_bound_mode): Likewise.
8631 (ix86_make_bounds_constant): Likewise.
8632 (ix86_initialize_bounds): Likewise.
8633 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
8634 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
8635 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
8636 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
8637 (TARGET_CHKP_BOUND_MODE): Likewise.
8638 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
8639 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
8640 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
8641 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
8642 * config/i386/i386.h (TARGET_MPX): Likewise.
8643 (TARGET_MPX_P): Likewise.
8644 (VALID_BND_REG_MODE): Likewise.
8645 (FIRST_BND_REG): Likewise.
8646 (LAST_BND_REG): Likewise.
8647 (enum reg_class): Likewise.
8648 (BND_REG_P): Likewise.
8649 (BND_REGNO_P): Likewise.
8650 (BNDmode): Likewise.
8651 (ADJUST_INSN_LENGTH): Likewise.
8652 * config/i386/i386.md: Likewise.
8653 * config/i386/i386.opt: Likewise.
8654 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
8655 (defined): Likewise.
8656 (LINK_MPX): Likewise.
8657 (MPX_SPEC): Likewise.
8658 (LIBMPX_SPEC): Likewise.
8659 (LIBMPXWRAPPERS_SPEC): Likewise.
8660 (CHKP_SPEC): Likewise.
8661 * config/i386/predicates.md: Likewise.
8662 * dbxout.c (dbxout_type): Likewise.
8663 * doc/extend.texi: Likewise.
8664 * doc/invoke.texi: Likewise.
8665 * doc/md.texi: Likewise.
8666 * doc/tm.texi: Likewise.
8667 * doc/tm.texi.in: Likewise.
8668 * dwarf2out.c (is_base_type): Likewise.
8669 (gen_formal_types_die): Likewise.
8670 (gen_subprogram_die): Likewise.
8671 (gen_type_die_with_usage): Likewise.
8672 (gen_decl_die): Likewise.
8673 (dwarf2out_late_global_decl): Likewise.
8674 * expr.c (expand_assignment): Likewise.
8675 (emit_storent_insn): Likewise.
8676 (store_expr_with_bounds): Likewise.
8677 (store_expr): Likewise.
8678 (expand_expr_real_1): Likewise.
8679 * expr.h (store_expr_with_bounds): Likewise.
8680 * function.c (use_register_for_decl): Likewise.
8681 (struct bounds_parm_data): Likewise.
8682 (assign_parms_augmented_arg_list): Likewise.
8683 (assign_parm_find_entry_rtl): Likewise.
8684 (assign_parm_is_stack_parm): Likewise.
8685 (assign_parm_load_bounds): Likewise.
8686 (assign_bounds): Likewise.
8687 (assign_parms): Likewise.
8688 (expand_function_start): Likewise.
8689 * gcc.c (CHKP_SPEC): Likewise.
8690 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
8691 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
8692 (wrestrict_dom_walker::check_call): Likewise.
8693 * gimple.c (gimple_build_call_from_tree): Likewise.
8694 * gimple.h (enum gf_mask): Likewise.
8695 (gimple_call_with_bounds_p): Likewise.
8696 (gimple_call_set_with_bounds): Likewise.
8697 * gimplify.c (gimplify_init_constructor): Likewise.
8698 * ipa-cp.c (initialize_node_lattices): Likewise.
8699 (propagate_constants_across_call): Likewise.
8700 (find_more_scalar_values_for_callers_subset): Likewise.
8701 * ipa-hsa.c (process_hsa_functions): Likewise.
8702 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
8703 * ipa-icf.c (sem_function::merge): Likewise.
8704 * ipa-inline.c (early_inliner): Likewise.
8705 * ipa-pure-const.c (warn_function_noreturn): Likewise.
8706 (warn_function_cold): Likewise.
8707 (propagate_pure_const): Likewise.
8708 * ipa-ref.h (enum GTY): Likewise.
8709 * ipa-split.c (find_retbnd): Likewise.
8710 (consider_split): Likewise.
8711 (split_function): Likewise.
8712 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8713 * ipa.c (walk_polymorphic_call_targets): Likewise.
8714 (symbol_table::remove_unreachable_nodes): Likewise.
8715 (process_references): Likewise.
8716 (cgraph_build_static_cdtor_1): Likewise.
8717 * lto-cgraph.c (lto_output_node): Likewise.
8718 (output_refs): Likewise.
8719 (compute_ltrans_boundary): Likewise.
8720 (input_overwrite_node): Likewise.
8721 (input_node): Likewise.
8722 (input_cgraph_1): Likewise.
8723 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
8724 * passes.c (pass_manager::execute_early_local_passes): Likewise.
8725 (class pass_chkp_instrumentation_passes): Likewise.
8726 (make_pass_chkp_instrumentation_passes): Likewise.
8727 * passes.def: Likewise.
8728 * rtl.h (struct GTY): Likewise.
8729 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
8730 * stor-layout.c (layout_type): Likewise.
8731 * symtab.c: Likewise.
8732 * target.def: Likewise.
8733 * targhooks.c (default_chkp_bound_type): Likewise.
8734 (default_chkp_bound_mode): Likewise.
8735 (default_builtin_chkp_function): Likewise.
8736 (default_chkp_function_value_bounds): Likewise.
8737 (default_chkp_make_bounds_constant): Likewise.
8738 (default_chkp_initialize_bounds): Likewise.
8739 * targhooks.h (default_chkp_bound_type): Likewise.
8740 (default_chkp_bound_mode): Likewise.
8741 (default_builtin_chkp_function): Likewise.
8742 (default_chkp_function_value_bounds): Likewise.
8743 (default_chkp_make_bounds_constant): Likewise.
8744 (default_chkp_initialize_bounds): Likewise.
8745 * toplev.c (compile_file): Likewise.
8746 (process_options): Likewise.
8747 * tree-core.h (DEF_BUILTIN): Likewise.
8748 (DEF_BUILTIN_CHKP): Likewise.
8749 * tree-inline.c (declare_return_variable): Likewise.
8750 (remap_gimple_stmt): Likewise.
8751 (copy_bb): Likewise.
8752 (initialize_inlined_parameters): Likewise.
8753 (expand_call_inline): Likewise.
8754 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
8755 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
8756 (make_pass_ipa_chkp_produce_thunks): Likewise.
8757 (make_pass_chkp): Likewise.
8758 (make_pass_chkp_opt): Likewise.
8759 (make_pass_chkp_instrumentation_passes): Likewise.
8760 * tree-pretty-print.c (dump_generic_node): Likewise.
8761 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
8762 * tree-ssa-dce.c (propagate_necessity): Likewise.
8763 (eliminate_unnecessary_stmts): Likewise.
8764 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
8765 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
8766 * tree-ssa-sccvn.h: Likewise.
8767 * tree-ssa-strlen.c (get_string_length): Likewise.
8768 (valid_builtin_call): Likewise.
8769 (adjust_last_stmt): Likewise.
8770 (handle_builtin_strchr): Likewise.
8771 (handle_builtin_strcpy): Likewise.
8772 (handle_builtin_stxncpy): Likewise.
8773 (handle_builtin_memcpy): Likewise.
8774 (handle_builtin_strcat): Likewise.
8775 (strlen_check_and_optimize_stmt): Likewise.
8776 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
8777 * tree-streamer-in.c: Likewise.
8778 * tree-streamer.c (record_common_node): Likewise.
8779 * tree.c (tree_code_size): Likewise.
8780 (wide_int_to_tree_1): Likewise.
8781 (type_contains_placeholder_1): Likewise.
8782 (build_common_tree_nodes): Likewise.
8783 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
8784 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
8785 (POINTER_BOUNDS_P): Likewise.
8786 (BOUNDED_TYPE_P): Likewise.
8787 (BOUNDED_P): Likewise.
8788 (CALL_WITH_BOUNDS_P): Likewise.
8789 (pointer_bounds_type_node): Likewise.
8790 * value-prof.c (gimple_ic): Likewise.
8791 * var-tracking.c (vt_add_function_parameters): Likewise.
8792 * varasm.c (make_decl_rtl): Likewise.
8793 (assemble_start_function): Likewise.
8794 (output_constant): Likewise.
8795 (maybe_assemble_visibility): Likewise.
8796 * varpool.c (ctor_for_folding): Likewise.
8797 * chkp-builtins.def: Remove.
8798 * ipa-chkp.c: Remove.
8799 * ipa-chkp.h: Remove.
8800 * rtl-chkp.c: Remove.
8801 * rtl-chkp.h: Remove.
8802 * tree-chkp-opt.c: Remove.
8803 * tree-chkp.c: Remove.
8804 * tree-chkp.h: Remove.
8806 2018-06-07 Carl Love <cel@us.ibm.com>
8808 * config/rs6000/vsx.md (vextract_fp_from_shorth,
8809 vextract_fp_from_shortl): Add BE support.
8811 2018-06-07 Paul Koning <ni1d@arrl.net>
8813 * compare-elim.c (try_merge_compare): Don't merge compare if
8814 address contains a side effect.
8815 (try_eliminate_compare): Likewise.
8817 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
8819 * config.gcc: Support "tremont".
8820 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
8821 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8823 * config/i386/i386.c (m_TREMONT): Define.
8824 (processor_target_table): Add "tremont".
8825 (PTA_TREMONT): Define.
8826 (ix86_lea_outperforms): Add TARGET_TREMONT.
8827 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
8828 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
8829 and M_INTEL_GOLDMONT_PLUS.
8830 (fold_builtin_cpu): Add "tremont".
8831 (ix86_add_stmt_cost): Add TARGET_TREMONT.
8832 (ix86_option_override_internal): Add "tremont".
8833 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
8834 (processor_type): Add PROCESSOR_TREMONT.
8835 * config/i386/x86-tune.def: Add m_TREMONT.
8836 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
8838 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8840 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
8841 symbol defined for msp430i* devices to be lower case.
8843 2018-06-07 Richard Biener <rguenther@suse.de>
8845 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
8846 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
8847 Properly wrap signed arithmetic if overflow wraps.
8849 2018-06-07 Jakub Jelinek <jakub@redhat.com>
8851 PR tree-optimization/69615
8852 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
8853 of a cast from a same precision integral SSA_NAME in a bb dominated
8854 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
8855 cast to utype if rhs2 has already a compatible type.
8857 2018-06-07 Richard Biener <rguenther@suse.de>
8859 PR tree-optimization/85935
8860 * graphite-scop-detection.c (find_params_in_bb): Analyze
8861 condition operands with respect to the correct loop. Assert
8862 the analysis doesn't fail.
8864 2018-06-04 Carl Love <cel@us.ibm.com>
8866 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
8867 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
8868 as it is slightly cheaper.
8869 (first_match_or_eos_index_<mode>):
8870 Calculate index using natural element order.
8871 (first_match_index_<mode>):
8872 Calculate index using natural element order.
8873 (first_match_or_eos_index_<mode>):
8874 Calculate index using natural order.
8875 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
8876 for BE and LE modes.
8877 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
8878 P9V_BUILTIN_VCLZLSBB_V16QI.
8879 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
8882 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8884 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
8885 indentation and line wrap for many prototypes. Add missing
8886 @smallexample directives around block of prototypes for vec_xl and
8889 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
8891 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
8892 track if we pass or return IEEE 128-bit floating point.
8893 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
8894 C++ mangling that is compatible with GCC 8.1.
8895 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8896 (init_cumulative_args): Note if we pass or return IEEE 128-bit
8897 floating point types.
8898 (rs6000_function_arg_advance_1): Likewise.
8899 (rs6000_mangle_type): Optionally generate mangled names that match
8900 what GCC 8.1 generated for IEEE 128-bit floating point types.
8901 (rs6000_globalize_decl_name): If we have an external function that
8902 passes or returns IEEE 128-bit types, generate a weak reference
8903 from the mangled name used in GCC 8.1 to the current mangled
8905 (rs6000_init_builtins): Make __ibm128 use the long double type if
8906 long double is IBM extended double. Make __float128 use the long
8907 double type if long double is IEEE 128-bit.
8910 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
8911 macro for __ibm128 built-in functions.
8912 (PACK_IF): Add __ibm128 pack/unpack functions.
8913 (UNPACK_IF): Likewise.
8914 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
8915 enable long double built-in functions if long double is IEEE
8916 128-bit floating point.
8917 (rs6000_invalid_builtin): Update long double built-in function
8919 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
8920 functions, adjust the built-in function to use the long double
8921 built-in function if __ibm128 and long double are the same type.
8922 * doc/extend.texi (PowerPC builtins): Update documention for
8923 __builtin_{,un}pack_longdouble. Add documentation for
8924 __builtin_{,un}pack_ibm128.
8926 2018-06-06 Jim Wilson <jimw@sifive.com>
8928 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
8929 (struct machine_function): New field interrupt_mode.
8930 (riscv_handle_type_attribute): New function. Add forward declaration.
8931 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
8932 (riscv_expand_epilogue): Check interrupt_mode field.
8933 (riscv_set_current_function): Check interrupt attribute args and
8934 set interrupt_mode field.
8935 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
8936 (riscv_sret, riscv_uret): New.
8937 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
8938 new arguments to interrupt attribute.
8940 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
8943 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
8944 Don't handle -mcpu=power8 if -mpower9-vector is also used.
8946 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8948 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
8949 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
8950 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8951 several redundant entries.
8953 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8955 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8956 type from "rtx" to "rtx_insn *".
8957 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8958 for local "call_insn", removing cast.
8959 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8961 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8963 PR tree-optimization/86066
8964 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8965 for BIT_INSERT_EXPR stores.
8967 2018-06-06 Richard Biener <rguenther@suse.de>
8969 PR tree-optimization/86062
8970 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8971 component refs ontop
8972 of to be offsetted base.
8974 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8976 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8977 to be static and remove check on interrupt attribute name.
8979 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8981 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8982 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8984 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8987 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8989 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8990 Remove second argument, change how error is called.
8991 (aarch64_layout_arg): Remove second argument from
8992 aarch64_err_no_fpadvsimd call.
8993 (aarch64_init_cumulative_args): Ditto.
8994 (aarch64_gimplify_va_arg_expr): Ditto.
8995 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8997 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8999 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9000 (*simple_return_indirect_internal<mode>): Rename from
9001 simple_return_indirect_internal. Use W mode iterator.
9002 (rstorssp): New expander.
9003 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9004 (clrssbsy): New expander.
9005 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9007 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9009 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9011 (cmse_check_pointed_object): Likewise.
9013 2018-06-05 Martin Liska <mliska@suse.cz>
9015 PR gcov-profile/47618
9016 * doc/invoke.texi: Document how -fprofile-dir format
9019 2018-06-05 Richard Biener <rguenther@suse.de>
9021 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9022 removal pretend DOM info isn't available so we do not update
9023 it and only remove edges, not dominated blocks. Actually free
9024 DOM info in case we removed something. Remove unreachable blocks.
9025 (mfb_keep_latches): Work with either DOM info or marked backedges.
9026 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9027 first. Mark backedges if DOM info isn't available.
9028 (Re-)compute DOM info after cleanup_control_flow_pre.
9030 2018-06-05 Richard Biener <rguenther@suse.de>
9032 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9033 (locus_discrim_hasher::hash): Adjust.
9034 (locus_discrim_hasher::equal): Likewise.
9035 (next_discriminator_for_locus): Work on line directly.
9036 (same_line_p): Pass in expanded locus1 as well.
9037 (assign_discriminators): Avoid redundant location expansions.
9039 2018-06-05 Richard Biener <rguenther@suse.de>
9041 PR tree-optimization/86046
9042 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9043 if required after clearing TREE_ADDRESSABLE.
9045 2018-06-05 Richard Biener <rguenther@suse.de>
9047 PR tree-optimization/86047
9048 * tree-ssa-loop.c (for_each_index): Glob handling of all
9049 decls and constants and really handle all of them.
9051 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9054 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9055 qualifier_void_pointer and qualifier_const_void_pointer.
9056 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9057 (arm_init_builtins): Handle the above.
9058 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9059 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9060 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9063 2018-06-05 Martin Liska <mliska@suse.cz>
9065 * auto-profile.c (read_autofdo_file): Do not use
9066 gcov_ctr_summary struct.
9067 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9068 * coverage.c (struct counts_entry): Likewise.
9069 (read_counts_file): Read just single summary entry.
9070 (get_coverage_counts): Use gcov_summary struct.
9071 * coverage.h (get_coverage_counts): Likewise.
9072 * gcov-dump.c (dump_working_sets): Likewise.
9073 (tag_summary): Dump just single summary.
9074 * gcov-io.c (gcov_write_summary): Write just histogram
9076 (gcov_read_summary): Read just single summary.
9077 (compute_working_sets): Use gcov_summary struct.
9078 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9079 of GCOV_COUNTERS_SUMMABLE.
9080 (GCOV_COUNTERS_SUMMABLE): Remove.
9081 (GCOV_FIRST_VALUE_COUNTER): Replace with
9082 GCOV_COUNTER_V_INTERVAL.
9083 (struct gcov_ctr_summary): Remove.
9084 (struct gcov_summary): Directly use fields of former
9086 (compute_working_sets): Use gcov_summary struct.
9087 * gcov.c (read_count_file): Do not use ctrs fields.
9088 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9090 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9092 * profile.c: Likewise.
9093 * profile.h: Likewise.
9095 2018-06-05 Martin Liska <mliska@suse.cz>
9097 PR gcov-profile/84846
9098 * gcov.c (output_lines): Print working directory only
9099 in intermediate format.
9101 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9103 * config/s390/s390-builtin-types.def: Add void function type.
9104 * config/s390/s390-builtins.def: Use the function type for the
9107 2018-06-04 Jim Wilson <jimw@sifive.com>
9109 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9111 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9112 and maybe_eh_return. Change regno to unsigned int. Use new args to
9113 handle EH_RETURN_DATA_REGNO registers properly.
9114 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9115 (riscv_expand_epilogue): Update comment. Change argument name and
9116 type. Update code to use new name and type. Pass new args to
9117 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9119 * config/riscv/riscv.md (NORMAL_RETURN): New.
9120 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9121 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9122 (eh_return): Call gen_eh_return_internal and emit barrier.
9123 (eh_return_internal): Call riscv_expand_epilogue.
9125 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9127 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9128 bit_insertion field and declare can_be_merged_into method.
9129 (merged_store_group::can_be_merged_into): New method.
9130 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9131 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9132 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9134 2018-06-04 Richard Biener <rguenther@suse.de>
9136 PR tree-optimization/85955
9137 * builtins.c (fold_builtin_sincos): Convert pointers to
9138 destination to appropriate type before dereferencing.
9140 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9142 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9144 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9146 * expr.c (expand_expr_real_1): Force the operand into memory if
9147 its TYPE_MODE is BLKmode and if there is no integer mode for
9148 the number of bits being extracted.
9150 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9154 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9155 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9157 2018-06-04 Richard Biener <rguenther@suse.de>
9159 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9160 (cleanup_tree_cfg_noloop): ... single caller. Do
9161 start_recording_case_labels later.
9163 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9165 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9166 to _IMMINTRIN_H_INCLUDED.
9167 * config/i386/pconfigintrin.h: Ditto.
9168 * config/i386/waitpkgintrin.h: Ditto.
9169 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9170 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9171 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9172 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9173 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9174 waitpkgintrin.h and cldemoteintrin.h.
9176 2018-06-04 Richard Biener <rguenther@suse.de>
9178 PR tree-optimization/86038
9179 * tracer.c (find_best_successor): Check probability for
9180 being initialized, bail out if not.
9182 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9185 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9186 of bits to ignore when comparing architectures.
9188 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9190 PR tree-optimization/69615
9191 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
9192 maximum or minimum of the type, try to merge it also as if
9193 range1 is + [-, x - 1] or + [x + 1, -].
9196 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
9198 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
9200 PR tree-optimization/86034
9201 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
9202 the unsigned bitfield type in a bit insertion sequence if it does not
9203 have a larger precision than the bitfield size.
9204 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
9206 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
9208 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
9210 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9212 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
9213 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9214 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
9215 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
9217 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9219 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
9220 Disable -fdelete-null-pointer-checks for ELF toolchain.
9222 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9223 Kito Cheng <kito.cheng@gmail.com>
9225 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
9226 (nds32le-*-*, nds32be-*-*): Integrate checking process.
9227 (nds32*-*-*): Add glibc and uclibc conditions.
9228 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
9229 (TARGET_EXCEPT_UNWIND_INFO): Define.
9230 * config/nds32/elf.h: New file.
9231 * config/nds32/linux.h: New file.
9232 * config/nds32/nds32-elf.opt: New file.
9233 * config/nds32/nds32-linux.opt: New file.
9234 * config/nds32/nds32-fp-as-gp.c
9235 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
9236 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
9238 (nds32_asm_file_end): Ditto.
9239 (nds32_print_operand): Ditto.
9240 (nds32_insert_attributes): Ditto.
9241 (nds32_init_libfuncs): New function.
9242 (TARGET_HAVE_TLS): Define.
9243 (TARGET_INIT_LIBFUNCS): Define.
9244 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
9246 (TARGET_ELF): Apply different mcmodel setting.
9247 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
9248 been migrated into elf.h and linux.h files.
9249 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
9250 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
9251 (mcmodel): The content has been migrated into nds32-elf.opt and
9252 nds32-linux.opt files.
9253 * config/nds32/t-elf: New file.
9254 * config/nds32/t-linux: New file.
9256 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9257 Shiva Chen <shiva0217@gmail.com>
9259 * config/nds32/constants.md (unspec_volatile_element): Add
9260 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
9261 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
9263 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9264 (make_pass_nds32_fp_as_gp): Declare.
9265 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
9267 (nds32_asm_function_end_prologue): Remove unused asm output.
9268 (nds32_asm_function_begin_epilogue): Remove unused asm output.
9269 (nds32_asm_file_start): Output necessary fp_as_gp information.
9270 (nds32_option_override): Adjust register usage.
9271 (nds32_expand_prologue): Consider fp_as_gp situation.
9272 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
9273 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
9276 (simple_return): Ditto.
9277 (omit_fp_begin): Output special directive for fp_as_gp.
9278 (omit_fp_end): Output special directive for fp_as_gp.
9279 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
9280 mforbid-fp-as-gp): New options.
9282 2018-06-01 Mark Wielaard <mark@klomp.org>
9284 * dwarf2out.c (dwarf2out_finish): Remove generation of
9285 DW_AT_loclists_base.
9287 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
9289 * gimple-ssa-store-merging.c: Include gimple-fold.h.
9290 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
9291 (struct merged_store_group): Add bit_insertion field.
9292 (dump_char_array): Use standard hexadecimal format.
9293 (merged_store_group::merged_store_group): Set bit_insertion to false.
9294 (merged_store_group::apply_stores): Use optimal buffer size. Deal
9295 with BIT_INSERT_EXPR stores. Move up code updating the mask and
9296 also print the mask in the dump file.
9297 (pass_store_merging::gate): Minor tweak.
9298 (imm_store_chain_info::coalesce_immediate): Fix wrong association
9299 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
9300 stores with INTEGER_CST stores.
9301 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
9302 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
9303 and use it throughout. Generate bit insertion sequences if need be.
9304 (pass_store_merging::process_store): Remove redundant condition.
9305 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
9307 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
9309 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
9310 the 128-bit floating point types. Fix function comment.
9312 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9314 * config/aarch64/aarch64-simd.md
9315 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
9317 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
9320 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9322 PR tree-optimization/85989
9323 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
9325 (backprop::intersect_uses): Check it when deciding whether this
9326 is a backedge reference.
9327 (backprop::process_block): Add each phi to m_visited_phis
9328 after visiting it, then clear it at the end.
9330 2018-06-01 Richard Biener <rguenther@suse.de>
9332 * tree-vectorizer.h (vect_dr_stmt): New function.
9333 (vect_get_load_cost): Adjust.
9334 (vect_get_store_cost): Likewise.
9335 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
9336 Use vect_dr_stmt instead of DR_SMTT.
9337 (vect_record_base_alignments): Likewise.
9338 (vect_calculate_target_alignment): Likewise.
9339 (vect_compute_data_ref_alignment): Likewise and make static.
9340 (vect_update_misalignment_for_peel): Likewise.
9341 (vect_verify_datarefs_alignment): Likewise.
9342 (vector_alignment_reachable_p): Likewise.
9343 (vect_get_data_access_cost): Likewise. Pass down
9344 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
9345 (vect_get_peeling_costs_all_drs): Likewise.
9346 (vect_peeling_hash_get_lowest_cost): Likewise.
9347 (vect_enhance_data_refs_alignment): Likewise.
9348 (vect_find_same_alignment_drs): Likewise.
9349 (vect_analyze_data_refs_alignment): Likewise.
9350 (vect_analyze_group_access_1): Likewise.
9351 (vect_analyze_group_access): Likewise.
9352 (vect_analyze_data_ref_access): Likewise.
9353 (vect_analyze_data_ref_accesses): Likewise.
9354 (vect_vfa_segment_size): Likewise.
9355 (vect_small_gap_p): Likewise.
9356 (vectorizable_with_step_bound_p): Likewise.
9357 (vect_prune_runtime_alias_test_list): Likewise.
9358 (vect_analyze_data_refs): Likewise.
9359 (vect_supportable_dr_alignment): Likewise.
9360 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9361 (vect_gen_prolog_loop_niters): Likewise.
9362 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9363 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
9365 (vect_recog_mask_conversion_pattern): Likewise.
9366 (vect_try_gather_scatter_pattern): Likewise.
9367 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
9368 to vect_get_store_cost.
9369 (vect_get_store_cost): Get stmt_info instead of DR.
9370 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
9371 (vect_get_load_cost): Get stmt_info instead of DR.
9373 2018-06-01 Richard Biener <rguenther@suse.de>
9376 * gimple-fold.c (var_decl_component_p): Also allow offsetted
9377 vars wrapped in MEM_REFs.
9379 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9381 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9382 Fix subreg tests so that we only return a choice between
9383 GENERAL_REGS and FP_REGS if the original classes included both.
9385 2018-06-01 Richard Biener <rguenther@suse.de>
9388 * tree-ssa-structalias.c (get_function_part_constraint):
9389 Handle NULL fi->decl.
9390 (find_func_aliases_for_call): Properly handle indirect
9391 fi from direct call.
9392 (find_func_clobbers): Likewise.
9393 (ipa_pta_execute): Likewise.
9394 (create_variable_info_for): For functions that are ifunc_resolver
9395 resolve to a varinfo that contains the result of the resolver call.
9396 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
9399 2018-05-31 Michael Collison <michael.collison@arm.com>
9401 * config/aarch64/aarch64.md:
9402 (*fix_to_zero_extenddfdi2): New pattern.
9403 * gcc.target/aarch64/fix_extend1.c: New testcase.
9405 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
9409 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
9410 and BUILT_IN_STRNCMP_EQ.
9411 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
9412 BUILT_IN_STRNCMP_EQ.
9413 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
9414 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9415 (gimple_fold_builtin): Likewise.
9416 * tree-ssa-strlen.c (compute_string_length): New function.
9417 (determine_min_obsize): New function.
9418 (handle_builtin_string_cmp): New function to handle calls to
9419 string compare functions.
9420 (strlen_optimize_stmt): Add handling to builtin string compare
9422 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
9423 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9424 * tree.c (build_common_builtin_nodes): Add new defines of
9425 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
9427 2018-05-31 Jakub Jelinek <jakub@redhat.com>
9430 * bb-reorder.c (pass_partition_blocks::gate): Return false for
9431 functions with naked attribute.
9433 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
9435 * config/i386/sse.md (avx_vec_concat<mode>):
9436 Substitute concat_tg_mode mode attribute with xtg_mode.
9437 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
9438 (concat_tg_mode): Remove mode attribute.
9440 2018-05-31 Martin Sebor <msebor@redhat.com>
9443 * calls.c (alloc_max_size): Correct a logic error/typo.
9444 Treat excessive arguments as infinite. Warn for invalid arguments.
9445 * doc/invoke.texi (-Walloc-size-larger-than): Update.
9447 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
9450 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
9451 and movx for Haswell.
9453 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
9454 Cesar Philippidis <cesar@codesourcery.com>
9457 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
9458 when emitting error on private/firstprivate reductions.
9459 * omp-low.c (lower_omp_target): Avoid reference-type processing
9460 on pointers for firstprivate clause.
9462 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9464 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
9467 * config/aarch64/aarch64-simd.md
9468 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
9469 (aarch64_ld1_x3_<mode>): Likewise
9470 (aarch64_st1x2<VALLDIF:mode>): Likewise
9471 (aarch64_st1_x2_<mode>): Likewise
9472 (aarch64_st1x3<VALLDIF:mode>): Likewise
9473 (aarch64_st1_x3_<mode>): Likewise
9474 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
9475 (vld1_s8_x3): Likewise.
9476 (vld1_u16_x3): Likewise.
9477 (vld1_s16_x3): Likewise.
9478 (vld1_u32_x3): Likewise.
9479 (vld1_s32_x3): Likewise.
9480 (vld1_u64_x3): Likewise.
9481 (vld1_s64_x3): Likewise.
9482 (vld1_f16_x3): Likewise.
9483 (vld1_f32_x3): Likewise.
9484 (vld1_f64_x3): Likewise.
9485 (vld1_p8_x3): Likewise.
9486 (vld1_p16_x3): Likewise.
9487 (vld1_p64_x3): Likewise.
9488 (vld1q_u8_x3): Likewise.
9489 (vld1q_s8_x3): Likewise.
9490 (vld1q_u16_x3): Likewise.
9491 (vld1q_s16_x3): Likewise.
9492 (vld1q_u32_x3): Likewise.
9493 (vld1q_s32_x3): Likewise.
9494 (vld1q_u64_x3): Likewise.
9495 (vld1q_s64_x3): Likewise.
9496 (vld1q_f16_x3): Likewise.
9497 (vld1q_f32_x3): Likewise.
9498 (vld1q_f64_x3): Likewise.
9499 (vld1q_p8_x3): Likewise.
9500 (vld1q_p16_x3): Likewise.
9501 (vld1q_p64_x3): Likewise.
9502 (vst1_s64_x2): Likewise.
9503 (vst1_u64_x2): Likewise.
9504 (vst1_f64_x2): Likewise.
9505 (vst1_s8_x2): Likewise.
9506 (vst1_p8_x2): Likewise.
9507 (vst1_s16_x2): Likewise.
9508 (vst1_p16_x2): Likewise.
9509 (vst1_s32_x2): Likewise.
9510 (vst1_u8_x2): Likewise.
9511 (vst1_u16_x2): Likewise.
9512 (vst1_u32_x2): Likewise.
9513 (vst1_f16_x2): Likewise.
9514 (vst1_f32_x2): Likewise.
9515 (vst1_p64_x2): Likewise.
9516 (vst1q_s8_x2): Likewise.
9517 (vst1q_p8_x2): Likewise.
9518 (vst1q_s16_x2): Likewise.
9519 (vst1q_p16_x2): Likewise.
9520 (vst1q_s32_x2): Likewise.
9521 (vst1q_s64_x2): Likewise.
9522 (vst1q_u8_x2): Likewise.
9523 (vst1q_u16_x2): Likewise.
9524 (vst1q_u32_x2): Likewise.
9525 (vst1q_u64_x2): Likewise.
9526 (vst1q_f16_x2): Likewise.
9527 (vst1q_f32_x2): Likewise.
9528 (vst1q_f64_x2): Likewise.
9529 (vst1q_p64_x2): Likewise.
9530 (vst1_s64_x3): Likewise.
9531 (vst1_u64_x3): Likewise.
9532 (vst1_f64_x3): Likewise.
9533 (vst1_s8_x3): Likewise.
9534 (vst1_p8_x3): Likewise.
9535 (vst1_s16_x3): Likewise.
9536 (vst1_p16_x3): Likewise.
9537 (vst1_s32_x3): Likewise.
9538 (vst1_u8_x3): Likewise.
9539 (vst1_u16_x3): Likewise.
9540 (vst1_u32_x3): Likewise.
9541 (vst1_f16_x3): Likewise.
9542 (vst1_f32_x3): Likewise.
9543 (vst1_p64_x3): Likewise.
9544 (vst1q_s8_x3): Likewise.
9545 (vst1q_p8_x3): Likewise.
9546 (vst1q_s16_x3): Likewise.
9547 (vst1q_p16_x3): Likewise.
9548 (vst1q_s32_x3): Likewise.
9549 (vst1q_s64_x3): Likewise.
9550 (vst1q_u8_x3): Likewise.
9551 (vst1q_u16_x3): Likewise.
9552 (vst1q_u32_x3): Likewise.
9553 (vst1q_u64_x3): Likewise.
9554 (vst1q_f16_x3): Likewise.
9555 (vst1q_f32_x3): Likewise.
9556 (vst1q_f64_x3): Likewise.
9557 (vst1q_p64_x3): Likewise.
9559 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9561 * config/msp430/msp430.c (msp430_output_labelref): Prepend
9562 user_label_prefix to name.
9564 * tree-core.h: Update comment about the format of NAME string
9565 passed to handler in attribute_spec.
9567 * config/msp430/msp430.md: Remove erroneous subreg expression from
9568 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
9569 zero_extend{q,h}isi2.
9571 2018-05-30 Borislav Petkov <bp@suse.de>
9573 * doc/extend.texi: Document some architecture specific
9574 constraints and sort entries.
9576 2018-05-30 Martin Sebor <msebor@redhat.com>
9579 * builtins.c (expand_builtin_stpcpy_1): New function.
9580 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
9581 only if the former succeeds.
9583 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9585 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
9588 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9590 * doc/invoke.texi (-flinker-output): Document
9592 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9594 * passes.c (ipa_write_summaries): Only modify statements if body
9596 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
9597 incrementally linking.
9598 (ipa_passes): Likewise.
9599 * lto-cgraph.c (lto_output_node): When incrementally linking do not
9600 pass down resolution info.
9601 * common.opt (flag_incremental_link): Update info.
9602 * gcc.c (plugin specs): Turn flinker-output=* to
9603 -plugin-opt=-linker-output-known
9604 * toplev.c (compile_file): Also cut compilation when doing incremental
9606 * flag-types. (enum lto_partition_model): Add
9607 LTO_LINKER_OUTPUT_NOLTOREL.
9608 (invoke.texi): Add -flinker-output docs.
9609 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
9610 link same way as WPA; do not stream in dead initializers.
9612 * dwarf2out.c (dwarf2out_die_ref_for_decl,
9613 darf2out_register_external_decl): Support incremental link.
9615 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9617 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
9619 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9621 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
9622 it down to simple_object_copy_lto_debug_sections.
9623 (run_gcc): Determine incremental LTO link time and configure
9624 lto1 into non-wpa mode, disable renaming of debug sections.
9626 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
9628 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9629 descriptions of various incorrectly documented functions.
9631 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
9634 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9635 address check not strict.
9637 2018-05-30 Richard Biener <rguenther@suse.de>
9639 PR tree-optimization/85964
9640 * tracer.c (better_p): Drop initialized count check, we only
9641 call the function with initialized counts now.
9642 (find_best_successor): Do find a best edge if one
9643 has uninitialized count.
9644 (find_best_predecessor): Likewise. Do BB frequency check only
9645 if count is initialized.
9647 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
9649 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
9650 (aarch64_ldrstr_offset_compare): New.
9651 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
9652 load/store orderings.
9653 (aarch64_gen_adjusted_ldpstp): Likewise.
9655 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
9657 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9658 Check for subset of GENERAL_REGS and FP_REGS.
9659 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
9662 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
9664 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
9665 and wi::to_poly_offset. Add the current offset and then check
9666 whether the sum fits, rather than using an unchecked addition of
9667 a checked term. Check for a shwi rather than a uhwi.
9668 * expr.c (get_bit_range): Use tree_to_poly_uint64.
9669 (store_constructor): Use poly_int_tree_p.
9670 (expand_expr_real_1): Likewise.
9671 * function.c (assign_temp): Likewise.
9672 * fold-const.c (const_binop): Use poly_int_tree_p and
9674 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
9676 * ipa-icf-gimple.c (func_checker::compare_operand): Use
9677 to_poly_offset for MEM offsets.
9678 * ipa-icf.c (sem_variable::equals): Likewise.
9679 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
9680 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
9681 wi::to_poly_offset for BIT_FIELD_REF offsets.
9682 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
9684 * var-tracking.c (emit_note_insn_var_location): Use
9685 tree_to_poly_uint64.
9687 2018-05-29 Jim Wilson <jimw@sifive.com>
9689 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
9691 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
9694 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
9695 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
9697 (sse4_1_round<mode>2): Use nonimmediate_operand
9698 for operand 1 predicate.
9700 2018-05-29 Martin Sebor <msebor@redhat.com>
9701 Richard Biener <rguenther@suse.de>
9704 * calls.c (get_size_range): Call determine_value_range instead
9705 of get_value_range..
9706 * tree-vrp.h (determine_value_range): Declared new function.
9707 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
9709 2018-05-29 Richard Biener <rguenther@suse.de>
9711 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
9712 sure to use non-pattern stmts for get_earlier_stmt arguments.
9713 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
9714 called on pattern stmts.
9715 (get_later_stmt): Likewise.
9717 2018-05-29 Martin Liska <mliska@suse.cz>
9719 PR gcov-profile/85759
9720 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
9723 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9725 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
9727 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
9731 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9732 VEC_PACK_FLOAT_EXPR): New tree codes.
9733 * tree-pretty-print.c (op_code_prio): Handle
9734 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9735 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9736 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9737 * tree-inline.c (estimate_operator_cost): Likewise.
9738 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
9739 * fold-const.c (const_binop): Likewise.
9740 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
9741 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9742 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9743 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
9744 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9745 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9746 * expr.c (expand_expr_real_2): Likewise.
9747 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
9748 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
9749 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
9751 * optabs.c (expand_widen_pattern_expr): For
9752 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
9753 sign from result type rather than operand's type.
9754 (expand_binop_directly): For vec_packu_float_optab and
9755 vec_packs_float_optab allow result type to be different from operand's
9757 * optabs-tree.c (optab_for_tree_code): Handle
9758 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9759 VEC_PACK_FLOAT_EXPR. Formatting fixes.
9760 * tree-vect-generic.c (expand_vector_operations_1): Handle
9761 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9762 VEC_PACK_FLOAT_EXPR.
9763 * tree-vect-stmts.c (supportable_widening_operation): Handle
9765 (supportable_narrowing_operation): Handle FLOAT_EXPR.
9766 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
9767 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
9768 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
9769 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
9771 (vec_pack<floatprefix>_float_<mode>): New expander.
9772 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
9774 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
9775 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
9776 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
9777 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
9778 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
9780 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
9781 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
9782 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9783 VEC_PACK_FLOAT_EXPR): Document.
9785 2018-05-29 Richard Biener <rguenther@suse.de>
9787 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
9789 (stmt_vec_info_vec): Make pointer.
9790 (init_stmt_vec_info_vec): Remove.
9791 (free_stmt_vec_info_vec): Likewise.
9792 (set_stmt_vec_info_vec): New function.
9793 (free_stmt_vec_infos): Likewise.
9794 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
9795 (set_vinfo_for_stmt): Likewise.
9796 (get_earlier_stmt): Likewise.
9797 (get_later_stmt): Likewise.
9798 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
9799 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
9800 (vec_info::~vec_info): Free stmt_vec_infos.
9801 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
9802 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
9803 (pass_slp_vectorize::execute): Likewise.
9804 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
9805 (free_stmt_vec_info_vec): Likewise.
9806 (set_stmt_vec_info_vec): New function.
9807 (free_stmt_vec_infos): Likewise.
9808 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
9809 the global stmt_vec_info_vec.
9810 * tree-parloops.c (gather_scalar_reductions): Use
9811 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
9814 2018-05-29 Richard Biener <rguenther@suse.de>
9816 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
9818 2018-05-29 Martin Liska <mliska@suse.cz>
9819 David Malcolm <dmalcolm@redhat.com>
9821 * vec.c (test_reverse): New.
9822 (vec_c_tests): Add new test.
9823 * vec.h (vl_ptr>::reverse): New function.
9825 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
9827 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
9829 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
9832 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
9834 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
9836 2018-05-28 Richard Biener <rguenther@suse.de>
9838 PR tree-optimization/85933
9839 * tree-vect-data-refs.c (vect_record_base_alignments): Only
9840 look at stmts marked as vectorizable.
9842 2018-05-28 Richard Biener <rguenther@suse.de>
9844 PR tree-optimization/85934
9845 * tree-vect-generic.c (expand_vector_operations_1): Hoist
9846 vector boolean check before scalar optimization.
9848 2018-05-28 Jakub Jelinek <jakub@redhat.com>
9850 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
9853 2018-05-28 Mark Wielaard <mark@klomp.org>
9855 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9856 if it is an expression containing a minus sign.
9858 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
9860 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
9862 2018-05-27 Paul Koning <ni1d@arrl.net>
9864 * config/pdp11/pdp11.md (truncsihi2): Remove.
9866 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
9867 Chung-Ju Wu <jasonwucj@gmail.com>
9869 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
9871 (unaligned_store_dw): Ditto.
9872 * config/nds32/nds32-memory-manipulation.c
9873 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
9874 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
9875 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
9876 (emit_setmem_word_loop): Rename to ...
9877 (emit_setmem_doubleword_loop): ... this.
9878 (nds32_gen_dup_4_byte_to_word_value): New function.
9879 (nds32_gen_dup_8_byte_to_double_word_value): New function.
9880 (nds32_expand_setmem_loop): Refine implementation.
9881 (nds32_expand_setmem_loop_v3m): Ditto.
9882 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
9885 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9887 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
9889 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9891 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
9892 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9893 machine->attr_no_prologue_p.
9894 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
9895 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
9896 (nds32_expand_epilogue): Consider attr_naked_p.
9897 (nds32_expand_epilogue_v3pop): Likewise.
9898 (nds32_can_use_return_insn): Likewise.
9899 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
9900 attr_no_prologue_p fields.
9901 * config/nds32/nds32.opt (mret-in-naked-func): New option.
9903 2018-05-27 Jakub Jelinek <jakub@redhat.com>
9906 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
9908 * config/i386/sse.md
9909 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
9911 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
9913 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
9915 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
9917 (*<floatsuffix>floatv2div2sf2): Rename to ...
9918 (*float<floatunssuffix>v2div2sf2): ... this.
9919 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
9920 (float<floatunssuffix>v2div2sf2_mask): ... this.
9921 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
9922 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
9923 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
9925 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
9927 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9929 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9931 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9933 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9935 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
9936 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
9937 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
9938 gen_ufix_truncv8dfv8si2.
9939 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
9940 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
9941 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
9942 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
9943 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
9944 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
9945 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
9946 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
9948 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9952 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9954 2018-05-25 Jim Wilson <jimw@sifive.com>
9956 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9957 * config/riscv/riscv.c (struct machine_function): Add
9958 interrupt_handler_p and attribute_checked_p fields.
9959 (riscv_attribute_table): Add interrupt.
9960 (riscv_interrupt_type_p): New.
9961 (riscv_save_reg_p): Save extra regs for interrupt handler.
9962 (riscv_use_save_libcall): Return false for interrupt handler.
9963 (riscv_first_stack_step): Add forward declaration.
9964 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9965 for interrupt handler with large frame. Use it for saved reg list.
9966 (riscv_expand_prologue): Move flag_stack_usage_info support to
9967 eliminate duplication.
9968 (riscv_expand_epilogue): Generate mret for interrupt handler.
9969 (riscv_epilogue_uses): New.
9970 (riscv_can_use_return_insn): Return false for interrupt handler.
9971 (riscv_function_ok_for_sibcall): Likewise.
9972 (riscv_set_current_function): Add interrupt handler support.
9973 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9974 * config/riscv/riscv.md (UNSPECV_MRET): New.
9976 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9978 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9980 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9982 PR tree-optimization/85712
9983 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9984 this candidate has already been replaced in-situ by a copy.
9986 2018-05-25 Jason Merrill <jason@redhat.com>
9988 PR c++/80485 - inline function non-zero address.
9989 * symtab.c (nonzero_address): Check DECL_COMDAT.
9991 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9994 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9995 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9996 extension of SImode operation. Use const123_operand predicate.
9997 (*saddsi_1): Remove.
9998 (*saddl_se_1): Ditto.
10000 (*ssubl_se_1): Ditto.
10001 * config/alpha/predicates.md (const123_operand): New predicate.
10002 * config/alpha/constraints.md (P): Use IN_RANGE.
10004 2018-05-25 Richard Biener <rguenther@suse.de>
10006 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10008 (ref_maybe_used_by_stmt_p): Likewise.
10009 (stmt_may_clobber_ref_p): Likewise.
10010 (stmt_may_clobber_ref_p_1): Likewise.
10011 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10013 (ref_maybe_used_by_stmt_p): Likewise.
10014 (stmt_may_clobber_ref_p): Likewise.
10015 (stmt_may_clobber_ref_p_1): Likewise.
10016 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10017 the alias oracle to disambiguate DRs with stmts DR analysis
10019 (vect_analyze_data_refs): Do not give up on not analyzable
10020 DRs for BB vectorization. Remove code truncating the dataref
10023 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10026 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10027 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10028 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10030 2018-05-25 Richard Biener <rguenther@suse.de>
10032 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10033 function, combining stmt data ref gathering and fatal analysis
10035 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10036 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10037 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10038 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10039 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10041 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10043 PR tree-optimization/85720
10044 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10045 SCC if all partitions are builtins.
10046 (version_loop_by_alias_check): New parameter. Generate cancelable
10047 runtime alias check if all partitions are builtins.
10048 (distribute_loop): Update call to above function.
10050 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10052 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10053 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10054 (parm_default_def_partition_arg): Ditto.
10055 (set_parm_default_def_partition): Ditto.
10056 (get_parm_default_def_partitions): Ditto and make it static.
10057 (get_undefined_value_partitions): Ditto and make it static.
10058 (remove_ssa_form): Refactor call to init_var_map here.
10059 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10060 computation for loop region.
10061 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10062 (register_default_def): Delete.
10063 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10064 (parm_default_def_partition_arg): Ditto.
10065 (set_parm_default_def_partition): Ditto.
10066 (get_parm_default_def_partitions): Ditto and make it static.
10067 (get_undefined_value_partitions): Ditto and make it static.
10068 (coalesce_with_default, coalesce_with_default): Update comment.
10069 (create_coalesce_list_for_region): New func factored out from
10070 create_outofssa_var_map.
10071 (populate_coalesce_list_for_outofssa): New func factored out from
10072 create_outofssa_var_map and coalesce_ssa_name.
10073 (create_outofssa_var_map): Delete.
10074 (coalesce_ssa_name): Refactor to support live range computation.
10075 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10076 (get_parm_default_def_partitions): Delete.
10077 (get_undefined_value_partitions): Ditto.
10078 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10079 computation for loop region.
10080 (new_tree_live_info, loe_visit_block): Ditto.
10081 (live_worklist, set_var_live_on_entry): Ditto.
10082 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10083 * tree-ssa-live.h (struct _var_map): New fields.
10084 (init_var_map): Change decl.
10085 (region_contains_p): New.
10087 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10089 * tree-ssa-live.h (live_merge_and_clear): Delete.
10091 2018-05-25 Richard Biener <rguenther@suse.de>
10094 * tree-dump.c (dequeue_and_dump): Remove access to removed
10095 operand 2 of a SWITCH_EXPR.
10097 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10099 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10100 multiplication and division.
10101 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10102 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10103 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10104 (cond_udiv_optab, cond_umod_optab): New optabs.
10105 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10106 (IFN_COND_RDIV): New internal functions.
10107 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10108 TRUNC_MOD_EXPR and RDIV_EXPR.
10109 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10110 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10112 (SVE_INT_BINARY): Include mult.
10113 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10114 (optab, sve_int_op): Handle mult.
10115 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10117 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10118 for SVE_INT_BINARY_SD.
10120 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10122 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10123 (optab, sve_int_op): Handle div and udiv.
10124 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10125 for SVE_INT_BINARY_SD.
10126 (*<optab><mode>3): New insn for the same.
10128 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10130 * tree-vect-patterns.c: Include predict.h.
10131 (vect_recog_divmod_pattern): Restrict check for division support
10132 to when optimizing for size.
10134 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10136 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10137 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10138 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10139 (gimple_match_op::set_op): Likewise.
10140 (gimple_resimplify4): Declare.
10141 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10142 (expr::gen_transform): Likewise.
10143 (decision_tree::gen): Generate a simplification routine for 4 operands.
10144 * gimple-match-head.c (gimple_simplify): Add an overload for
10145 4 operands. In the top-level function, handle up to 4 call
10146 arguments and call gimple_resimplify4.
10147 (gimple_resimplify4): New function.
10148 (build_call_internal): Pass a fourth operand.
10149 (maybe_push_to_seq): Likewise.
10150 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10151 Fold VEC_COND_EXPRs of an operation and a default value into
10152 an IFN_COND_* function if possible.
10153 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10155 (SVE_COND_FP_BINARY): Include them.
10156 (optab, sve_fp_op): Handle them.
10157 (SVE_INT_BINARY_REV): New code iterator.
10158 (SVE_COND_FP_BINARY_REV): New int iterator.
10159 (commutative): New int attribute.
10160 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10162 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10164 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10165 (*cond_<optab><mode>): New patterns for reversed operands.
10167 2018-05-25 Richard Biener <rguenther@suse.de>
10169 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10170 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10171 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10172 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10173 * tree-vect-data-refs.c (everywhere): Adjust users.
10174 * tree-vect-loop.c (everywhere): Likewise.
10175 * tree-vect-slp.c (everywhere): Likewise.
10176 * tree-vect-stmts.c (everywhere): Likewise.
10177 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10179 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10181 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10183 (gcc_cv_as_section_exclude): ... this.
10184 Try Solaris as #exclude syntax.
10185 * configure: Regenerate.
10186 * config.in: Regenerate.
10187 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
10189 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
10190 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
10192 * varasm.c (default_elf_asm_named_section): Don't check if
10193 HAVE_GAS_SECTION_EXCLUDE is defined.
10195 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10197 * doc/md.texi: Update the documentation of the cond_* optabs
10198 to mention the new final operand. Fix GET_MODE_NUNITS call.
10199 Describe the scalar case too.
10200 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
10201 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
10203 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
10204 (get_conditional_internal_fn): Update comment.
10205 * tree-vect-loop.c (vectorizable_reduction): Pass the original
10206 accumulator value as a final argument to conditional functions.
10207 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
10208 a define_expand and add an "else" operand. Assert for now that
10209 the else operand is equal to operand 2. Use SVE_INT_BINARY and
10210 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
10211 (*cond_<optab><mode>): New patterns.
10212 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
10213 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10214 (UNSPEC_COND_EOR): Delete.
10215 (optab): Remove associated mappings.
10216 (SVE_INT_BINARY): New code iterator.
10217 (sve_int_op): Remove int attribute and add "minus" to the code
10219 (SVE_COND_INT_OP): Delete.
10220 (SVE_COND_FP_OP): Rename to...
10221 (SVE_COND_FP_BINARY): ...this.
10223 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10225 * optabs.c (can_reuse_operands_p): New function.
10226 (maybe_legitimize_operands): Try to reuse the results for
10229 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10231 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10232 Add {q} suffix to insn mnemonic.
10234 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10236 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
10237 (msp430_warn_func_return): New.
10239 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
10241 * fold-const.c (tree_nonzero_bits): New function.
10242 * fold-const.h (tree_nonzero_bits): Likewise.
10243 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
10244 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
10246 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10250 * varasm.c (assemble_alias): Check ifunc_resolver only on
10253 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10256 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
10257 when memory input operand is handled.
10259 2018-05-24 Luis Machado <luis.machado@linaro.org>
10261 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
10263 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
10265 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10267 * match.pd: Delay FMA folds until after vectorization.
10269 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
10272 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10273 address check not strict.
10275 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10277 * gimple-match.h (gimple_match_op): New class.
10278 (mprts_hook): Replace parameters with a gimple_match_op *.
10279 (maybe_build_generic_op): Likewise.
10280 (gimple_simplified_result_is_gimple_val): Replace parameters with
10281 a const gimple_match_op *.
10282 (gimple_simplify): Replace code_helper * and tree * parameters with
10283 a gimple_match_op * parameter.
10284 (gimple_resimplify1): Replace code_helper *, tree and tree *
10285 parameters with a gimple_match_op * parameter.
10286 (gimple_resimplify2): Likewise.
10287 (gimple_resimplify3): Likewise.
10288 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
10289 parameters with a gimple_match_op * parameter.
10290 * gimple-match-head.c (gimple_simplify): Change prototypes of
10291 auto-generated functions to take a gimple_match_op * instead of
10292 separate code_helper * and tree * parameters. Make the same
10293 change in the top-level overload and update calls to the
10294 gimple_resimplify routines. Update calls to the auto-generated
10295 functions and to maybe_push_res_to_seq in the publicly-facing
10296 operation-specific gimple_simplify overloads.
10297 (gimple_match_op::MAX_NUM_OPS): Define.
10298 (gimple_resimplify1): Replace rcode and ops with a single res_op
10299 parameter. Update call to gimple_simplify.
10300 (gimple_resimplify2): Likewise.
10301 (gimple_resimplify3): Likewise.
10302 (mprts_hook): Replace parameters with a gimple_match_op *.
10303 (maybe_build_generic_op): Likewise.
10304 (build_call_internal): Replace type, nargs and ops with
10305 a gimple_match_op *.
10306 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
10307 with a single gimple_match_op *. Update calls to mprts_hook,
10308 build_call_internal and gimple_simplified_result_is_gimple_val.
10309 Factor out code that is common to the tree_code and combined_fn cases.
10310 * genmatch.c (expr::gen_transform): Replace tem_code and
10311 tem_ops with a gimple_match_op called tem_op. Update calls
10312 to the gimple_resimplify functions and maybe_push_res_to_seq.
10313 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
10314 res_ops. Update call to the gimple_resimplify functions.
10315 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
10316 (decision_tree::gen): Make the functions take a gimple_match_op *
10317 called res_op instead of separate res_code and res_ops parameters.
10318 Update call accordingly.
10319 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
10320 and ops with a single res_op parameter. Update calls to
10321 maybe_build_generic_op and maybe_push_res_to_seq.
10322 (fold_stmt_1): Update calls to gimple_simplify and
10323 replace_stmt_with_simplification.
10324 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
10325 and gimple_simplified_result_is_gimple_val.
10326 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
10328 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
10329 with a gimple_match_op *.
10330 (vn_nary_build_or_lookup): Likewise. Update call to
10331 vn_nary_build_or_lookup_1.
10332 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
10333 gimple_match_op *. Update calls to the gimple_resimplify routines
10334 and to gimple_simplified_result_is_gimple_val.
10335 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
10336 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
10337 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
10338 (visit_nary_op): Likewise.
10339 (visit_reference_op_load): Likewise.
10341 2018-05-23 Luis Machado <luis.machado@linaro.org>
10343 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
10344 modifier for printing the step amount.
10346 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
10349 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
10352 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
10354 * doc/sourcebuild.texi (Endianness): New subsubsection.
10356 2018-05-23 Luis Machado <luis.machado@linaro.org>
10358 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10359 <prefetch_dynamic_strides>: New const bool field.
10360 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10361 prefetch_dynamic_strides.
10362 (exynosm1_prefetch_tune): Likewise.
10363 (thunderxt88_prefetch_tune): Likewise.
10364 (thunderx_prefetch_tune): Likewise.
10365 (thunderx2t99_prefetch_tune): Likewise.
10366 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
10368 (aarch64_override_options_internal): Update to set
10369 PARAM_PREFETCH_DYNAMIC_STRIDES.
10370 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10371 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10372 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10373 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10374 prefetch-dynamic-strides setting.
10376 2018-05-23 Luis Machado <luis.machado@linaro.org>
10378 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10379 <minimum_stride>: New const int field.
10380 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10381 minimum_stride field defaulting to -1.
10382 (exynosm1_prefetch_tune): Likewise.
10383 (thunderxt88_prefetch_tune): Likewise.
10384 (thunderx_prefetch_tune): Likewise.
10385 (thunderx2t99_prefetch_tune): Likewise.
10386 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
10387 <default_opt_level>: Set to 3.
10388 (aarch64_override_options_internal): Update to set
10389 PARAM_PREFETCH_MINIMUM_STRIDE.
10390 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10391 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10392 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10393 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10394 stride is constant and is below the minimum stride threshold.
10396 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10398 * config/arm/arm-cpus.in (mode26): Delete.
10399 (armv4): Delete mode26 reference.
10400 * config/arm/arm.c (arm_configure_build_target): Delete use of
10403 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10405 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
10407 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
10408 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
10410 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
10411 (floatunsdidf2): Ditto.
10413 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10415 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
10416 (fixuns_trunc<mode>si2_avx512f): Ditto.
10417 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
10418 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
10419 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
10421 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10423 PR rtl-optimization/79985
10424 * df-scan.c (df_insn_refs_collect): Remove special case for
10425 global registers and asm statements.
10427 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10429 * extend.texi (Global Register Variables): Rewrite the bullet list.
10430 Note that the register is available for allocation. Note that access
10431 via inline asm must use constraints. Add note about async-signal
10432 handlers. Remove paragraph about automagic register selection.
10434 2018-05-23 Richard Biener <rguenther@suse.de>
10436 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
10437 of fixed offset from memset VN.
10439 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
10441 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
10442 first_interp field.
10443 (alloc_cand_and_find_basis): Initialize first_interp field.
10444 (slsr_process_mul): Modify first_interp field.
10445 (slsr_process_add): Likewise.
10446 (slsr_process_cast): Modify first_interp field for each new
10448 (slsr_process_copy): Likewise.
10449 (dump_candidate): Dump first_interp field.
10450 (replace_mult_candidate): Process all interpretations, not just
10452 (replace_rhs_if_not_dup): Likewise.
10453 (replace_one_candidate): Likewise.
10455 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
10457 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
10459 (aarch64_needs_frame_chain): New function.
10460 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
10462 2018-05-23 Sudakshina Das <sudi.das@arm.com>
10465 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
10466 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
10467 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
10468 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
10469 as true for strict-align.
10470 (aarch64_can_inline_p): Perform checks even when callee has no
10471 attributes to check for strict alignment.
10472 * doc/extend.texi (AArch64 Function Attributes): Document
10474 * doc/invoke.texi: (AArch64 Options): Likewise.
10476 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
10478 PR tree-optimization/85853
10479 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
10480 the handling of the root of the node to...
10481 (vect_slp_analyze_node_operations_1): ...this new function,
10482 and run the whole thing with the child nodes' def types
10483 set according to their SLP node's def type.
10485 2018-05-23 Richard Biener <rguenther@suse.de>
10487 PR middle-end/85874
10488 * tree-data-ref.c (create_runtime_alias_checks): Defer
10489 and ignore overflow warnings.
10491 2018-05-23 Yury Gribov <tetra2005@gmail.com>
10493 PR tree-optimization/85822
10494 * tree-vrp.c (is_masked_range_test): Fix handling of negative
10497 2018-05-23 Richard Biener <rguenther@suse.de>
10499 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10500 memset constants via native_interpret_expr.
10502 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
10505 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
10507 (cgraph_node::create_alias): Likewise.
10508 (cgraph_node::get_availability): Check ifunc_resolver instead
10509 of looking up ifunc attribute.
10510 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10511 * varasm.c (do_assemble_alias): Likewise.
10512 (assemble_alias): Likewise.
10513 (default_binds_local_p_3): Likewise.
10514 * cgraph.h (cgraph_node): Add ifunc_resolver.
10515 (cgraph_node::only_called_directly_or_aliased_p): Return false
10516 for IFUNC resolver.
10517 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
10519 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
10520 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
10521 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
10522 instead of looking up ifunc attribute.
10524 2018-05-22 Luis Machado <luis.machado@linaro.org>
10526 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
10528 2018-05-22 Martin Sebor <msebor@redhat.com>
10530 PR middle-end/85359
10531 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
10532 only when expasion succeeds.
10533 (expand_builtin_strcmp): Same.
10534 (expand_builtin_strncmp): Same.
10536 2018-05-22 Martin Sebor <msebor@redhat.com>
10538 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
10540 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
10541 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10543 * config/aarch64/aarch64-ldpstp.md: Replace uses of
10544 aarch64_mem_pair_operand with memory_operand and delete operand swapping
10546 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10547 Add check for legitimate_address.
10548 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
10549 (aarch64_swap_ldrstr_operands): New.
10550 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
10553 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
10554 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10556 * config/aarch64/aarch64.md: New patterns to generate stp
10558 (store_pair_sw, store_pair_dw): New patterns to generate stp for
10559 single words and double words.
10560 (load_pair_sw, load_pair_dw): Likewise.
10561 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
10563 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
10565 * config/aarch64/aarch64-ldpstp.md: Modify peephole
10566 for different mode ldpstp and add peephole for merged zero stores.
10567 Likewise for loads.
10568 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10570 (aarch64_gen_store_pair): Rename calls to match new patterns.
10571 (aarch64_gen_load_pair): Rename calls to match new patterns.
10572 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
10573 (load_pair<DREG:mode><DREG2:mode>): ... This.
10574 (store_pair<mode>): Rename to...
10575 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
10576 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
10577 New mode iterators.
10578 (V_INT_EQUIV): Handle SImode.
10579 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
10582 2018-05-22 Martin Sebor <msebor@redhat.com>
10585 * calls.c (maybe_warn_nonstring_arg): Use string length to set
10586 or ajust the presumed bound on an operation to avoid unnecessary
10589 2018-05-22 Martin Sebor <msebor@redhat.com>
10591 PR tree-optimization/85826
10592 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
10593 assuming that a DECL necesarily has a constant size.
10595 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
10597 PR middle-end/85862
10598 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
10600 2018-05-22 Richard Biener <rguenther@suse.de>
10602 PR tree-optimization/85834
10603 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
10604 non-constant and non-zero memset arguments.
10606 2018-05-22 Martin Liska <mliska@suse.cz>
10609 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
10611 2018-05-22 Richard Biener <rguenther@suse.de>
10613 PR tree-optimization/85863
10614 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
10615 comparisons when vectype is specified.
10616 (vectorizable_condition): Do not specify vectype for
10617 vect_is_simple_cond when SLP vectorizing.
10619 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
10622 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
10623 define __ibm128 as long double.
10624 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
10625 as a distinct type when IEEE 128-bit support is enabled.
10626 (init_float128_ieee): Fix up conversions between IFmode and IEEE
10627 128-bit types to use the correct functions.
10628 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
10629 convert between 128-bit floating point types that have different
10630 modes but the same representation, instead of using gen_lowpart to
10632 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
10634 (IFKF_reg): New attributes to give the register constraints for
10636 (extend<mode>tf2_internal): New insns to mark an explicit
10637 conversion between 128-bit floating point types that have a
10638 different mode but share the same representation.
10640 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
10642 PR tree-optimization/85814
10643 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
10644 a null return from get_strinfo when unsharing the next
10645 strinfo in the chain.
10647 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10650 * varasm.c (weak_finish): Clean up weak_decls.
10652 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10654 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
10655 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
10656 UNSPEC_UADALP values.
10657 * config/aarch64/iterators.md (ABAL): New int iterator.
10660 (sur): Add mappings for the above.
10661 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
10663 (aarch64_<sur>abal<mode>_4): Likewise.
10664 (aarch64_<sur>adalp<mode>_3): Likewise.
10665 (<sur>sadv16qi): New define_expand.
10667 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
10669 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
10670 (*movdf_internal): Ditto.
10671 (*rcpsf2_sse): Ditto.
10672 (*rsqrtsf2_sse): Ditto.
10673 (*sqrt<mode>2_sse): Ditto.
10675 2018-05-21 Tamar Christina <tamar.christina@arm.com>
10677 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
10679 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
10680 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
10681 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
10682 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
10684 * config/aarch64/arm_neon.h: Likewise.
10685 * config/aarch64/iterators.md (VQ_I): New.
10687 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
10689 * config.gcc: Add arc/t-multilib-linux to tmake_file for
10691 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
10694 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10696 * config/nds32/constraints.md (S): New constraint.
10697 * config/nds32/nds32.md (call_internal): Use constraint S.
10698 (call_value_internal): Likewise.
10699 (sibcall_internal): Likewise.
10700 (sibcall_value_internal): Likewise.
10702 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10703 Chung-Ju Wu <jasonwucj@gmail.com>
10705 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
10706 into consideration.
10708 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10709 Chung-Ju Wu <jasonwucj@gmail.com>
10711 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
10712 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
10713 (nds32_rtx_costs_impl): Simplify.
10714 (nds32_address_cost_impl): Simplify.
10715 (nds32_init_rtx_costs): New function.
10716 (nds32_rtx_costs_speed_prefer): Likewise.
10717 (nds32_rtx_costs_size_prefer): Likewise.
10718 (nds32_address_cost_speed_prefer): Likewise.
10719 (nds32_address_cost_speed_fwprop): Likewise.
10720 (nds32_address_cost_size_prefer): Likewise.
10721 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
10722 * config/nds32/nds32.c (nds32_option_override): Use
10723 nds32_init_rtx_costs function.
10725 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10727 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
10728 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
10729 (TARGET_PIPELINE_N8): Likewise.
10730 (TARGET_PIPELINE_N10): Likewise.
10731 (TARGET_PIPELINE_N13): Likewise.
10732 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10734 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
10736 * config/nds32/nds32-fpu.md: Update copyright year.
10738 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10740 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
10742 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10744 * config/nds32/nds32.c
10745 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
10746 * config/nds32/nds32.opt (minline-asm-r15): New option.
10748 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10750 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
10752 * config/nds32/nds32.md (abssi2): New pattern.
10754 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
10756 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10757 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
10758 Merge insn pattern from sse_cvtsi2ss<round_name> and
10759 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
10760 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
10761 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
10762 using SWI48 mode iterator.
10763 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
10764 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
10765 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
10766 pattern from sse_cvttss2si<round_saeonly_name>
10767 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
10768 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
10769 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
10770 using SWI48 mode iterator.
10771 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
10772 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
10773 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
10774 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
10775 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
10776 using SWI48 mode iterator.
10777 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
10778 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
10779 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
10780 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
10781 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
10782 SWI48 mode iterator.
10783 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
10784 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
10785 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
10786 pattern from sse_cvttsd2si<round_saeonly_name>
10787 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
10789 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10791 * config/nds32/nds32-md-auxiliary.c
10792 (nds32_valid_smw_lwm_base_p): Refine.
10793 (nds32_output_smw_single_word): Refine.
10794 (nds32_output_smw_double_word): New.
10795 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
10797 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10799 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
10800 (nds32_output_stack_pop): Refine.
10801 (nds32_expand_unaligned_load): Refine.
10802 (nds32_expand_unaligned_store): Refine.
10804 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10805 Chung-Ju Wu <jasonwucj@gmail.com>
10807 * config/nds32/constants.md: Add TP_REGNUM constant.
10808 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
10809 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
10811 * config/nds32/nds32-doubleword.md: Consider flag_pic.
10812 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
10813 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
10814 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
10815 and PIC code generation.
10816 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
10818 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
10820 * config/nds32/nds32.md: Support TLS and PIC.
10821 * config/nds32/nds32.c: Support TLS and PIC.
10822 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
10823 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
10826 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10828 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
10829 mode with E_ prefix.
10831 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10832 Chung-Ju Wu <jasonwucj@gmail.com>
10834 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
10835 * config/nds32/nds32-md-auxiliary.c
10836 (symbolic_reference_mentioned_p): New.
10837 (nds32_legitimize_ict_address): New.
10838 (nds32_expand_ict_move): New.
10839 (nds32_indirect_call_referenced_p): New.
10840 (nds32_symbol_binds_local_p): Delete.
10841 (nds32_long_call_p): Modify.
10842 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
10843 * config/nds32/nds32-protos.h
10844 (symbolic_reference_mentioned_p): Declare.
10845 (nds32_legitimize_ict_address): Declare.
10846 (nds32_expand_ict_move): Declare.
10847 (nds32_indirect_call_referenced_p): Declare.
10848 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
10849 (nds32_relax_group): Use nds32_ict_const_p as condition.
10850 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
10851 (nds32_asm_file_start): Output ict_model directive in asm code.
10852 (nds32_legitimate_address_p): Consider indirect call.
10853 (nds32_print_operand): Consider indirect call.
10854 (nds32_print_operand_address): Consider indirect call.
10855 (nds32_insert_attributes): Handle "indirect_call" attribute.
10856 (TARGET_LEGITIMATE_ADDRESS_P): Define.
10857 (TARGET_LEGITIMATE_CONSTANT_P): Define.
10858 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
10859 (TARGET_DELEGITIMIZE_ADDRESS): Define.
10860 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
10861 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
10862 (TARGET_ICT_MODEL_SMALL): Define.
10863 (TARGET_ICT_MODEL_LARGE): Define.
10864 * config/nds32/nds32.md (movsi): Consider ict model.
10865 (call, call_value): Consider ict model.
10866 (sibcall, sibcall_value): Consider ict model.
10867 * config/nds32/nds32.opt (mict-model): New option.
10868 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
10871 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
10872 Monk Chiang <sh.chiang04@gmail.com>
10873 Jim Wilson <jimw@sifive.com>
10875 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
10876 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
10877 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
10878 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
10879 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
10880 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
10881 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
10882 compute save_libcall_adjustment properly.
10883 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
10884 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
10885 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
10886 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
10887 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
10888 (ABI_SPEC): Handle mabi=ilp32e.
10889 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
10890 (RVE): Add RVE mask.
10891 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
10892 <-march>: Add rv32e as an example.
10894 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
10897 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
10898 (intra_create_variable_infos): Handle C++ constructors.
10900 2018-05-18 Martin Liska <mliska@suse.cz>
10902 * passes.def: Remove a redundant pass.
10904 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
10907 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10909 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10911 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10913 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
10915 (armv2, armv2a, armv3, armv3m): Delete architectures.
10916 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
10917 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10918 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
10920 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
10921 (*mulsidi3adddi): Likewise.
10922 (mulsidi3): Likewise.
10923 (*mulsidi3_nov6): Likewise.
10924 (umulsidi3): Likewise.
10925 (umulsidi3_nov6): Likewise.
10926 (umaddsidi4): Likewise.
10927 (*umulsidi3adddi): Likewise.
10928 (smulsi3_highpart): Likewise.
10929 (*smulsi3_highpart_nov6): Likewise.
10930 (umulsi3_highpart): Likewise.
10931 (*umulsi3_highpart_nov6): Likewise.
10932 * config/arm/arm.h (arm_arch3m): Delete.
10933 * config/arm/arm.c (arm_arch3m): Delete.
10934 (arm_option_override_internal): Update armv3-related comment.
10935 (arm_configure_build_target): Delete use of isa_bit_mode32.
10936 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
10937 (arm_rtx_costs_internal): Delete check of arm_arch3m.
10938 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
10939 (mulsa3): Likewise.
10940 (mulusa3): Likewise.
10941 * config/arm/arm-protos.h (arm_arch3m): Delete.
10942 * config/arm/arm-tables.opt: Regenerate.
10943 * config/arm/arm-tune.md: Likewise.
10944 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
10945 deleted architectures.
10947 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10949 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
10950 (armv5t, armv5te): New features.
10951 (ARMv5, ARMv5e): Delete fgroups.
10952 (ARMv5t, ARMv5te): Adjust for above changes.
10953 (ARMv6m): Likewise.
10954 (armv5, armv5e): Delete arches.
10955 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10957 (*call_reg_arm): Likewise.
10958 (*call_value_reg_armv5): Likewise.
10959 (*call_value_reg_arm): Likewise.
10960 (*call_symbol): Likewise.
10961 (*call_value_symbol): Likewise.
10962 (*sibcall_insn): Likewise.
10963 (*sibcall_value_insn): Likewise.
10964 (clzsi2): Likewise.
10965 (prefetch): Likewise.
10966 (define_split and define_peephole2 dependent on arm_arch5):
10968 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10970 (TARGET_ARM_QBIT): Likewise.
10971 (TARGET_DSP_MULTIPLY): Likewise.
10972 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10973 (arm_arch5, arm_arch5e): Delete.
10974 (arm_arch5t, arm_arch5te): Declare.
10975 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10976 (arm_arch5t): Declare.
10977 (arm_option_reconfigure_globals): Update for the above.
10978 (arm_options_perform_arch_sanity_checks): Update comment, replace
10979 use of arm_arch5 with arm_arch5t.
10980 (use_return_insn): Likewise.
10981 (arm_emit_call_insn): Likewise.
10982 (output_return_instruction): Likewise.
10983 (arm_final_prescan_insn): Likewise.
10984 (arm_coproc_builtin_available): Likewise.
10985 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10986 arm_arch5e with arm_arch5t and arm_arch5te.
10987 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10988 (arm_arch5t, arm_arch5te): Declare.
10989 * config/arm/arm-tables.opt: Regenerate.
10990 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10991 * config/arm/t-multilib: Likewise.
10992 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10993 instead of arm_arch5.
10994 (*call_reg_thumb1): Likewise.
10995 (*call_value_reg_thumb1_v5): Likewise.
10996 (*call_value_reg_thumb1): Likewise.
10997 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10999 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11001 2018-05-18 Martin Liska <mliska@suse.cz>
11003 PR gcov-profile/84846
11004 * doc/gcov.texi: Document -t option of gcov tool.
11006 2018-05-18 Martin Liska <mliska@suse.cz>
11008 PR gcov-profile/84846
11009 * gcov.c (print_usage): Add new -t option.
11010 (process_args): Handle the option.
11011 (generate_results): Use stdout as output when requested by
11014 2018-05-18 Martin Liska <mliska@suse.cz>
11016 PR gcov-profile/84846
11017 * coverage.c (coverage_init): Write PWD to .gcno file.
11018 * doc/gcov.texi: Document how working directory is printed.
11019 * gcov-dump.c (dump_gcov_file): Print PWD.
11020 * gcov.c (output_intermediate_file): Likewise.
11021 (read_graph_file): Read PWD string.
11022 (output_lines): Print PWD.
11024 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11026 PR middle-end/85817
11027 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11028 for retval and return false if all args to phi are zero.
11030 2018-05-18 Richard Biener <rguenther@suse.de>
11032 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11034 (evrp_dom_walker::before_dom_children): Call it.
11036 2018-05-18 Richard Biener <rguenther@suse.de>
11038 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11039 results when processing array refs with variable index.
11041 2018-05-18 Toon Moene <toon@moene.org>
11043 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11044 directly after that of -floop-interchange. Indicate that both
11045 options are enabled by default when specifying -O3.
11047 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11049 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11050 iterator. Delete separate integer-mode vec_set<mode> expander.
11051 (aarch64_simd_vec_setv2di): Delete.
11052 (vec_setv2di): Delete.
11053 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11054 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11055 the "w, r" alternative.
11057 2018-05-18 Martin Liska <mliska@suse.cz>
11059 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11060 * tree-pass.h (make_pass_lower_switch_O0): New function.
11061 * tree-switch-conversion.c (node_has_low_bound): Remove.
11062 (node_has_high_bound): Likewise.
11063 (node_is_bounded): Likewise.
11064 (class pass_lower_switch): Make it a template type and create
11066 (pass_lower_switch::execute): Add template argument.
11067 (make_pass_lower_switch): New function.
11068 (make_pass_lower_switch_O0): New function.
11069 (do_jump_if_equal): Remove.
11070 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11071 all the hard work to tree optimization passes.
11073 2018-05-18 Martin Liska <mliska@suse.cz>
11075 * dbgcnt.c (limit_low): Renamed from limit.
11076 (limit_high): New variable.
11077 (dbg_cnt_is_enabled): Check for upper limit.
11078 (dbg_cnt): Adjust dumping.
11079 (dbg_cnt_set_limit_by_index): Add new argument for high
11081 (dbg_cnt_set_limit_by_name): Likewise.
11082 (dbg_cnt_process_single_pair): Parse new format.
11083 (dbg_cnt_process_opt): Use strtok.
11084 (dbg_cnt_list_all_counters): Remove 'value' and add
11086 * doc/invoke.texi: Document changes.
11088 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11090 * doc/sourcebuild.texi (scalar_all_fma): Document.
11091 * tree.def (FMA_EXPR): Delete.
11092 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11093 * internal-fn.c (ternary_direct): New macro.
11094 (expand_ternary_optab_fn): Likewise.
11095 (direct_ternary_optab_supported_p): Likewise.
11096 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11097 * builtins.c (fold_builtin_fma): Delete.
11098 (fold_builtin_3): Don't call it.
11099 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11100 * expr.c (expand_expr_real_2): Likewise.
11101 * fold-const.c (operand_equal_p): Likewise.
11102 (fold_ternary_loc): Likewise.
11103 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11104 * gimple.c (DEFTREECODE): Likewise.
11105 * gimplify.c (gimplify_expr): Likewise.
11106 * optabs-tree.c (optab_for_tree_code): Likewise.
11107 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11108 * tree-eh.c (operation_could_trap_p): Likewise.
11109 (stmt_could_throw_1_p): Likewise.
11110 * tree-inline.c (estimate_operator_cost): Likewise.
11111 * tree-pretty-print.c (dump_generic_node): Likewise.
11112 (op_code_prio): Likewise.
11113 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11114 * tree-ssa-operands.c (get_expr_operands): Likewise.
11115 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11116 * fold-const-call.h (fold_fma): Delete.
11117 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11118 CFN_FNMA and CFN_FNMS.
11119 (fold_fma): Delete.
11120 * genmatch.c (combined_fn): New enum.
11121 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11122 (commutative_op): New function.
11123 (commutate): Use it. Handle more than 2 operands.
11124 (dt_operand::gen_gimple_expr): Use commutative_op.
11125 (parser::parse_expr): Allow :c to be used with non-binary
11126 operators if the commutative operand is known.
11127 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11128 CFN_FMS, CFN_FNMA and CFN_FNMS.
11129 (backprop::process_assign_use): Remove FMA_EXPR handling.
11130 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11131 (gen_hsa_fma): New function.
11132 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11133 IFN_FNMA and IFN_FNMS.
11134 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11135 * gimple-fold.h (follow_all_ssa_edges): Declare.
11136 * gimple-fold.c (follow_all_ssa_edges): New function.
11137 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11138 gimple_build interface and use follow_all_ssa_edges to fold the result.
11139 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11140 instead of checking for optabs directly.
11141 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11142 rather than FMA_EXPRs.
11143 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11144 call to IFN_FMA instead of an FMA_EXPR.
11146 2018-05-17 Jim Wilson <jimw@sifive.com>
11148 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11149 sign extended promoted subreg, and the range does not have the sign bit
11150 set, then do a sign extend.
11152 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11153 test, check for sign extended subreg and/or constant operands, and
11154 do a sign extend in that case.
11156 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11158 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11159 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11161 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11162 Change logics_shift_reg to logics_shift_imm.
11163 (thunderx2t99_fp_loadpair_basic): Delete.
11164 (thunderx2t99_fp_storepair_basic): Delete.
11165 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11166 (thunderx2t99_asimd_polynomial): Delete.
11167 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11168 and neon_fp_mul_d_scalar_q.
11169 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11170 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11171 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11172 (thunderx2t99_asimd_lut): Add missing tbl types.
11173 (thunderx2t99_asimd_ext): Delete.
11174 (thunderx2t99_asimd_load1_1_mult): Delete.
11175 (thunderx2t99_asimd_load1_2_mult): Delete.
11176 (thunderx2t99_asimd_load1_ldp): New.
11177 (thunderx2t99_asimd_load1): New.
11178 (thunderx2t99_asimd_load2): Add missing *load2* types.
11179 (thunderx2t99_asimd_load3): New.
11180 (thunderx2t99_asimd_load4): New.
11181 (thunderx2t99_asimd_store1_1_mult): Delete.
11182 (thunderx2t99_asimd_store1_2_mult): Delete.
11183 (thunderx2t99_asimd_store2_mult): Delete.
11184 (thunderx2t99_asimd_store2_onelane): Delete.
11185 (thunderx2t99_asimd_store_stp): New.
11186 (thunderx2t99_asimd_store1): New.
11187 (thunderx2t99_asimd_store2): New.
11188 (thunderx2t99_asimd_store3): New.
11189 (thunderx2t99_asimd_store4): New.
11191 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
11193 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
11194 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
11196 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
11197 Segher Boessenkool <segher@kernel.crashing.org>
11200 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
11203 2018-05-17 Richard Biener <rguenther@suse.de>
11205 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11206 for pruning loop and prune defs feeding only already visited PHIs.
11208 2018-05-17 Richard Biener <rguenther@suse.de>
11210 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
11212 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11213 Richard Biener <rguenther@suse.de>
11215 PR tree-optimization/85793
11216 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
11217 for VMAT_ELEMENTWISE.
11219 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11221 * internal-fn.h (lookup_internal_fn): Declare
11222 * internal-fn.c (lookup_internal_fn): New function.
11223 * gimple.c (gimple_build_call_from_tree): Handle calls to
11224 internal functions.
11225 * gimple-pretty-print.c (dump_gimple_call): Print "." before
11226 internal function names.
11227 * tree-pretty-print.c (dump_generic_node): Likewise.
11228 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
11230 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11232 * gimple-fold.h (gimple_build): Make the function forms take
11233 combined_fn rather than built_in_function.
11234 (gimple_simplify): Likewise.
11235 * gimple-match-head.c (gimple_simplify): Likewise.
11236 * gimple-fold.c (gimple_build): Likewise.
11237 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
11238 rather than gimple_build_call_internal.
11239 (get_initial_defs_for_reduction): Likewise.
11240 (vect_create_epilog_for_reduction): Likewise.
11241 (vectorizable_live_operation): Likewise.
11243 2018-05-17 Martin Liska <mliska@suse.cz>
11245 * gimple-ssa-sprintf.c (format_directive): Do not use
11246 space in between 'G_' and '('.
11248 2018-05-17 Jakub Jelinek <jakub@redhat.com>
11251 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
11252 even if the mask is not all ones.
11255 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
11257 (ix86_gimple_fold_builtin): Likewise.
11260 * config/i386/i386.c: Include tree-vector-builder.h.
11261 (ix86_vector_shift_count): New function.
11262 (ix86_fold_builtin): Fold shift builtins by scalar count.
11263 (ix86_gimple_fold_builtin): Likewise.
11265 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
11266 _mm512_setzero): New intrinsics.
11268 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
11269 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11271 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
11272 code generation for cases where splatting a value is not useful.
11273 * simplify-rtx.c (simplify_ternary_operation): Simplify
11274 vec_merge across a vec_duplicate and a paradoxical subreg forming
11275 a vector mode to a vec_concat.
11277 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
11279 * config.gcc: Support "goldmont-plus".
11280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11283 PROCESSOR_GOLDMONT_PLUS.
11284 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
11285 (processor_target_table): Add "goldmont-plus".
11286 (PTA_GOLDMONT_PLUS): Define.
11287 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
11288 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
11289 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
11290 (fold_builtin_cpu): Add "goldmont-plus".
11291 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
11292 (ix86_option_override_internal): Add "goldmont-plus".
11293 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
11294 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
11295 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
11296 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
11298 2018-05-17 Richard Biener <rguenther@suse.de>
11300 PR tree-optimization/85757
11301 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
11302 remove defs that only feed that PHI from further processing.
11304 2018-05-16 Jim Wilson <jimw@sifive.com>
11306 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11308 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
11310 2018-05-16 Mark Wielaard <mark@klomp.org>
11312 * dwarf2out.c (count_index_strings): New function.
11313 (output_indirect_strings): Call count_index_strings and generate
11314 header for dwarf_version >= 5.
11316 2018-05-16 Mark Wielaard <mark@klomp.org>
11318 * dwarf2out.c (dwarf_FORM): New function.
11319 (set_indirect_string): Use dwarf_FORM.
11320 (reset_indirect_string): Likewise.
11321 (size_of_die): Likewise.
11322 (value_format): Likewise.
11323 (output_die): Likewise.
11324 (add_skeleton_AT_string): Likewise.
11325 (output_macinfo_op): Likewise.
11326 (index_string): Likewise.
11327 (output_index_string_offset): Likewise.
11328 (output_index_string): Likewise.
11329 (count_index_strings): Likewise.
11331 2018-05-16 Carl Love <cel@us.ibm.com>
11333 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11334 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
11336 2018-05-16 Martin Jambor <mjambor@suse.cz>
11338 * ipa-prop.c (ipa_free_all_edge_args): Remove.
11339 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
11341 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
11343 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
11344 (fnma<mode>4): Likewise.
11345 (fms<mode>4): Likewise.
11346 (fnms<mode>4): Likewise.
11347 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
11348 (aarch64_fnma<mode>4): Likewise.
11349 (aarch64_fms<mode>4): Likewise.
11350 (aarch64_fnms<mode>4): Likewise.
11351 (aarch64_fnmadd<mode>4): Likewise.
11353 2018-05-16 Jason Merrill <jason@redhat.com>
11355 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
11357 2018-05-16 Richard Biener <rguenther@suse.de>
11359 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
11360 (dump_stmt_cost): Declare.
11361 (add_stmt_cost): Dump cost we add.
11362 (add_stmt_costs): New function.
11363 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
11364 No longer exported.
11365 (vect_analyze_stmt): Adjust prototype.
11366 (vectorizable_condition): Likewise.
11367 (vectorizable_live_operation): Likewise.
11368 (vectorizable_reduction): Likewise.
11369 (vectorizable_induction): Likewise.
11370 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
11371 cost vector to pass to vectorizable_ and record afterwards.
11372 (vect_model_reduction_cost): Take cost vector argument and adjust.
11373 (vect_model_induction_cost): Likewise.
11374 (vectorizable_reduction): Likewise.
11375 (vectorizable_induction): Likewise.
11376 (vectorizable_live_operation): Likewise.
11377 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
11378 SLP_TREE_NUMBER_OF_VEC_STMTS.
11379 (vect_analyze_slp_cost_1): Remove.
11380 (vect_analyze_slp_cost): Likewise.
11381 (vect_slp_analyze_node_operations): Take visited args and
11382 a target cost vector. Avoid processing already visited stmt sets.
11383 (vect_slp_analyze_operations): Use a local cost vector to gather
11384 costs and register those of non-discarded instances.
11385 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
11386 (vect_schedule_slp_instance): Remove copying of
11387 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
11389 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
11390 adding cost. Record cost entry location.
11391 (vect_prologue_cost_for_slp_op): Function to compute cost of
11392 a constant or invariant generated for SLP vect in the prologue,
11393 split out from vect_analyze_slp_cost_1.
11394 (vect_model_simple_cost): Make static. Adjust for SLP costing.
11395 (vect_model_promotion_demotion_cost): Likewise.
11396 (vect_model_store_cost): Likewise, make static.
11397 (vect_model_load_cost): Likewise.
11398 (vectorizable_bswap): Add cost vector arg and adjust.
11399 (vectorizable_call): Likewise.
11400 (vectorizable_simd_clone_call): Likewise.
11401 (vectorizable_conversion): Likewise.
11402 (vectorizable_assignment): Likewise.
11403 (vectorizable_shift): Likewise.
11404 (vectorizable_operation): Likewise.
11405 (vectorizable_store): Likewise.
11406 (vectorizable_load): Likewise.
11407 (vectorizable_condition): Likewise.
11408 (vectorizable_comparison): Likewise.
11409 (can_vectorize_live_stmts): Likewise.
11410 (vect_analyze_stmt): Likewise.
11411 (vect_transform_stmt): Adjust calls to vectorizable_*.
11412 * tree-vectorizer.c: Include gimple-pretty-print.h.
11413 (dump_stmt_cost): New function.
11415 2018-05-16 Richard Biener <rguenther@suse.de>
11417 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
11418 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
11419 * tree-ssa-dse.c: Include tree-ssa-loop.h.
11420 (check_name): New callback.
11421 (dse_classify_store): Track cycles via a visited bitmap of PHI
11422 defs and simplify handling of in-loop and across loop dead stores
11423 and properly fail for loop-variant refs. Handle byte-tracking with
11424 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
11427 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
11429 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
11430 (vect_get_mask_type_for_stmt): Likewise.
11431 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
11433 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
11434 to determine the statement's vector type and the vector type that
11435 should be used for calculating nunits. Deal with cases in which
11436 the type has to be deferred.
11437 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
11438 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
11439 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
11440 (vect_determine_vf_for_stmt): New functions, split out from...
11441 (vect_determine_vectorization_factor): ...here.
11442 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
11443 (vect_get_mask_type_for_stmt): New functions, split out from
11444 vect_determine_vectorization_factor.
11446 2018-05-16 Richard Biener <rguenther@suse.de>
11448 * tree-cfg.c (verify_gimple_assign_ternary): Properly
11449 verify the [VEC_]COND_EXPR embedded comparison.
11451 2018-05-15 Martin Sebor <msebor@redhat.com>
11453 PR tree-optimization/85753
11454 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
11455 RECORD_TYPE in addition to ARRAY_TYPE.
11457 2018-05-15 Martin Sebor <msebor@redhat.com>
11459 PR middle-end/85643
11460 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
11462 2018-05-15 Richard Biener <rguenther@suse.de>
11464 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
11465 add by_clobber_p one. Change algorithm to collect all defs
11466 representing uses we need to walk and try reducing them to
11467 a single one before failing.
11468 (dse_dom_walker::dse_optimize_stmt): Adjust.
11470 2018-05-13 Mark Wielaard <mark@klomp.org>
11472 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
11473 (size_of_loc_descr): Likewise.
11474 (output_loc_operands): Likewise.
11475 (output_loc_operands_raw): Likewise.
11476 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
11477 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
11478 (hash_loc_operands): Likewise.
11479 (compare_loc_operands): Likewise.
11481 2018-05-14 Mark Wielaard <mark@klomp.org>
11483 * dwarf2out.c (count_index_addrs): New function.
11484 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
11486 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11488 PR tree-optimization/83648
11489 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
11490 return value as malloc candidate.
11492 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11495 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
11496 param as true in call to suggest_attribute.
11498 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
11500 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
11501 -mreadonly-in-sdata.
11503 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11505 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
11507 (aarch64_crypto_aesd_fused): Likewise.
11509 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
11511 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
11512 (movsi_aarch64): Likewise.
11513 (load_pairsi): Likewise.
11514 (load_pairdi): Likewise.
11515 (store_pairsi): Likewise.
11516 (store_pairdi): Likewise.
11517 (load_pairsf): Likewise.
11518 (load_pairdf): Likewise.
11519 (store_pairsf): Likewise.
11520 (store_pairdf): Likewise.
11521 (zero_extend): Likewise.
11522 (trunc): Swap alternatives.
11523 (fcvt_target): Add '?' to prefer w over r.
11525 2018-05-14 Jakub Jelinek <jakub@redhat.com>
11528 * config/i386/i386.md: Disallow non-commutative arithmetics in
11529 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
11530 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
11531 in the peephole2 before it.
11533 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
11535 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
11536 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
11537 (ix86_handle_option): Handle -mcldemote.
11538 * config.gcc: New header.
11539 * config/i386/cldemoteintrin.h: New file.
11540 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
11541 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11543 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11544 OPTION_MASK_ISA_CLDEMOTE.
11545 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
11546 (ix86_valid_target_attribute_inner_p): Ditto.
11547 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
11548 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
11549 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
11550 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
11551 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
11553 * config/i386/i386.opt: Add -mcldemote.
11554 * config/i386/x86intrin.h: New header.
11555 * doc/invoke.texi: Add -mcldemote.
11557 2018-05-14 Richard Biener <rguenther@suse.de>
11559 * doc/match-and-simplify.texi: Adjust :s documentation.
11561 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
11563 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
11564 intended memcpy size.
11565 (REORDER_45): Likewise.
11567 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
11569 * sort.cc: New file.
11570 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
11571 * vec.c (qsort_chk): Use gcc_qsort.
11572 * Makefile.in (OBJS-libcommon): Add sort.o.
11573 (build/sort.o): New target. Use it...
11574 (BUILD_RTL): ... here, and...
11575 (build/gencfn-macros): ... here, and...
11576 (build/genmatch): ... here.
11578 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11579 Chung-Ju Wu <jasonwucj@gmail.com>
11581 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
11582 * config/nds32/nds32-graywolf.md: New file.
11583 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
11584 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
11586 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
11587 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
11588 * config/nds32/nds32.md (pipeline_model): Add graywolf.
11589 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
11590 * config/nds32/pipelines.md: Include n15 settings.
11592 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11593 Chung-Ju Wu <jasonwucj@gmail.com>
11595 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
11596 * config/nds32/nds32-n13.md: New file.
11597 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
11598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
11600 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
11601 * config/nds32/nds32.md (pipeline_model): Add n13.
11602 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
11603 * config/nds32/pipelines.md: Include n13 settings.
11605 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11606 Chung-Ju Wu <jasonwucj@gmail.com>
11608 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
11609 * config/nds32/nds32-n10.md: New file.
11610 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
11611 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
11613 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
11614 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
11615 * config/nds32/nds32.md (pipeline_model): Add n10.
11616 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
11617 * config/nds32/pipelines.md: Include n10 settings.
11619 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
11620 Kito Cheng <kito.cheng@gmail.com>
11621 Chung-Ju Wu <jasonwucj@gmail.com>
11623 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
11624 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11625 Add enum values for DSP extension instructions.
11626 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
11628 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
11629 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
11630 New code iterators.
11631 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
11632 * config/nds32/nds32-dspext.md: New file for DSP implementation.
11633 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
11634 * config/nds32/nds32-intrinsic.md: Likewise.
11635 * config/nds32/nds32_intrinsic.h: Likewise.
11636 * config/nds32/nds32-md-auxiliary.c: Likewise.
11637 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
11638 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
11639 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
11640 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
11641 * config/nds32/nds32-protos.h: New declarations for DSP extension.
11642 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
11643 TYPE_DMAC in switch statement.
11644 * config/nds32/nds32.c: New checking and implementation for DSP
11645 extension instructions.
11646 * config/nds32/nds32.h: Likewise.
11647 * config/nds32/nds32.md: Likewise.
11648 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
11649 * config/nds32/predicates.md: Implement new predicates for DSP
11652 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11654 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
11655 Reformat alternatives and attributes so it is easier to identify
11656 which constraints/attributes go with which instruction.
11657 (mov<mode>_hardfloat32, FMOVE64): Likewise.
11658 (mov<mode>_softfloat32, FMOVE64): Likewise.
11659 (mov<mode>_hardfloat64, FMOVE64): Likewise.
11660 (mov<mode>_softfloat64, FMOVE64): Likewise.
11662 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
11664 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11666 (Basic PowerPC Built-in Functions): The new name of the
11667 subsection previously known as "PowerPC Built-in Functions".
11668 (Basic PowerPC Built-in Functions Available on all Configurations):
11670 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
11671 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
11672 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
11673 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
11675 2018-05-11 Martin Jambor <mjambor@suse.cz>
11678 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
11681 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11684 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
11686 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
11688 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
11689 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
11690 (ix86_handle_option): Handle -mwaitpkg.
11691 * config.gcc: New header.
11692 * config/i386/cpuid.h (bit_WAITPKG): New bit.
11693 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
11694 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
11696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11697 OPTION_MASK_ISA_WAITPKG.
11698 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
11699 (ix86_option_override_internal): Add PTA_WAITPKG.
11700 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
11701 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
11702 IX86_BUILTIN_TPAUSE.
11703 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
11704 __builtin_ia32_umwait and __builtin_ia32_tpause.
11705 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
11706 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
11707 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
11708 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
11709 UNSPECV_TPAUSE): New.
11710 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
11711 * config/i386/i386.opt: Add -mwaitpkg.
11712 * config/i386/waitpkgintrin.h: New file.
11713 * config/i386/x86intrin.h: New header.
11714 * doc/invoke.texi: Add -mwaitpkg.
11716 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11719 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
11721 (cortex-m0): Use armv6s-m isa.
11722 (cortex-m0plus): Likewise.
11723 (cortex-m1): Likewise.
11724 (cortex-m0.small-multiply): Likewise.
11725 (cortex-m0plus.small-multiply): Likewise.
11726 (cortex-m1.small-multiply): Likewise.
11728 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
11729 Jakub Jelinek <jakub@redhat.com>
11731 PR tree-optimization/85692
11732 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
11733 source permute as well.
11735 2018-05-11 Martin Liska <mliska@suse.cz>
11738 * doc/extend.texi: Document LLVM style format for no_sanitize
11741 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11743 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
11744 mode_supports_vsx_dform_quad to mode_supports_dq_form.
11745 (mode_supports_vsx_dform_quad): Likewise.
11746 (mode_supports_vmx_dform): Move these functions to be next to the
11747 other mode_supports functions.
11748 (mode_supports_dq_form): Likewise.
11749 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
11750 mode_supports_dq_form.
11751 (reg_offset_addressing_ok_p): Likewise.
11752 (offsettable_ok_by_alignment): Likewise.
11753 (rs6000_legitimate_offset_address_p): Likewise.
11754 (legitimate_lo_sum_address_p): Likewise.
11755 (rs6000_legitimize_address): Likewise.
11756 (rs6000_legitimize_reload_address): Likewise.
11757 (rs6000_secondary_reload_inner): Likewise.
11758 (rs6000_preferred_reload_class): Likewise.
11759 (rs6000_output_move_128bit): Likewise.
11761 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11763 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
11764 Generate SImode target register for null target.
11765 <case IX86_BUILTIN_XGETBV>: Ditto.
11766 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
11767 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
11769 2018-05-10 Carl Love <cel@us.ibm.com>
11771 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11772 dcbtt and dcbtstt if operands[2] is 0.
11774 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11777 * config/i386/sse.md (usadv64qi): New expander.
11779 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
11781 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
11782 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
11783 -maltivec=be support.
11784 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
11785 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
11786 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
11787 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
11788 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
11789 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
11790 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
11791 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
11792 altivec_vsumsws): Adjust.
11793 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
11794 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
11795 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
11797 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
11798 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
11799 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
11800 (altivec_lve<VI_char>x): Delete expand.
11801 (*altivec_lve<VI_char>x_internal): Rename to...
11802 (altivec_lve<VI_char>x): ... this.
11803 (altivec_lvxl_<mode>): Delete expand.
11804 (*altivec_lvxl_<mode>_internal): Rename to ...
11805 (altivec_lvxl_<mode>): ... this.
11806 (altivec_stvxl_<mode>): Delete expand.
11807 (*altivec_stvxl_<mode>_internal): Rename to ...
11808 (altivec_stvxl_<mode>): ... this.
11809 (altivec_stve<VI_char>x): Delete expand.
11810 (*altivec_stve<VI_char>x_internal): Rename to ...
11811 (altivec_stve<VI_char>x): ... this.
11812 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
11813 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
11814 reduc_plus_scal_<mode>): Adjust.
11815 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
11817 (rs6000_cpu_cpp_builtins): Adjust.
11818 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
11819 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
11820 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
11821 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
11822 -maltivec=be support.
11823 (rs6000_split_vec_extract_var): Adjust.
11824 (rs6000_split_v4si_init): Adjust.
11825 (swap_selector_for_mode): Delete.
11826 (altivec_expand_lvx_be, altivec_expand_stvx_be,
11827 altivec_expand_stvex_be): Delete.
11828 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
11829 -maltivec=be support.
11830 (rs6000_gimple_fold_builtin): Ditto.
11831 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
11833 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
11834 (TARGET_DIRECT_MOVE_64BIT): Adjust.
11835 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
11836 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
11837 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
11838 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
11839 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
11840 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
11841 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
11842 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
11843 anonymous split): Adjust.
11844 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
11845 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
11847 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
11849 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
11850 when --with-gxx-include-dir is also specified.
11851 * configure: Regenerate.
11853 2018-05-09 Jim Wilson <jimw@sifive.com>
11856 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11857 * config/riscv/t-withmultilib: New.
11858 * config/riscv/withmultilib.h: New.
11859 * doc/install.texi: Document RISC-V --with-multilib-list support.
11861 2018-05-09 Richard Biener <rguenther@suse.de>
11863 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
11865 (vect_bb_vectorization_profitable_p): Adjust. Compute
11866 actual scalar cost using the cost vector and the add_stmt_cost
11869 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11871 PR rtl-optimization/85645
11872 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
11873 in the REG_CFA_REGISTER note for LR, don't leave it empty.
11875 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11877 PR rtl-optimization/85645
11878 * shrink-wrap.c (spread_components): Return a boolean saying if
11879 anything was changed.
11880 (try_shrink_wrapping_separate): Iterate spread_components until
11881 nothing changes anymore.
11883 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11885 PR rtl-optimization/85645
11886 * regrename.c (build_def_use): Also kill the chains that include the
11887 destination of a REG_CFA_REGISTER note.
11889 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11891 PR rtl-optimization/85645
11892 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
11893 insn that has a REG_CFA_REGISTER note.
11895 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
11897 * cfgexpand.c (expand_clobber): New function.
11898 (expand_gimple_stmt_1): Use it.
11899 * tree-vect-stmts.c (vect_clobber_variable): New function,
11901 (vectorizable_simd_clone_call): ...here.
11902 (vectorizable_store): Emit a clobber either side of an
11903 IFN_STORE_LANES sequence.
11904 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
11906 2018-05-09 Tom de Vries <tom@codesourcery.com>
11909 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
11910 (define_insn "trap_if_false"): Add exit after trap.
11912 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
11914 PR rtl-optimization/85638
11915 * bb-reorder.c: Include common/common-target.h.
11916 (create_forwarder_block): New function extracted from...
11917 (fix_up_crossing_landing_pad): ...here. Rename into...
11918 (dw2_fix_up_crossing_landing_pad): ...this.
11919 (sjlj_fix_up_crossing_landing_pad): New function.
11920 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
11921 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
11922 from both partitions and exit the loop after one iteration.
11924 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11927 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11929 (Basic PowerPC Built-in Functions): The new name of the
11930 subsection previously known as "PowerPC Built-in Functions".
11931 (Basic PowerPC Built-in Functions Available on all Configurations):
11933 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11935 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11937 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11939 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11942 2018-05-08 Jim Wilson <jimw@sifive.com>
11944 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
11945 (LD_EMUL_SUFFIX): New.
11946 (LINK_SPEC): Use it.
11948 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11950 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11952 (Basic PowerPC Built-in Functions): The new name of the
11953 subsection previously known as "PowerPC Built-in Functions".
11954 (Basic PowerPC Built-in Functions Available on all Configurations):
11956 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11958 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11960 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11962 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11965 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11968 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11969 after cmpelim optimization.
11971 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11973 * config.gcc: Support "goldmont".
11974 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11975 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11976 PROCESSOR_GOLDMONT.
11977 * config/i386/i386.c (m_GOLDMONT): Define.
11978 (processor_target_table): Add "goldmont".
11979 (PTA_GOLDMONT): Define.
11980 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11981 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11982 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11983 (fold_builtin_cpu): Add "goldmont".
11984 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11985 (ix86_option_override_internal): Add "goldmont".
11986 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11987 (processor_type): Add PROCESSOR_GOLDMONT.
11988 * config/i386/i386.md: Add CPU "glm".
11989 * config/i386/glm.md: New file.
11990 * config/i386/x86-tune.def: Add m_GOLDMONT.
11991 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11993 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11996 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11998 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11999 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12000 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12003 * config/i386/i386.c (ix86_fold_builtin): Handle
12004 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12007 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12008 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12010 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12013 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12014 (check_arch): Likewise.
12015 (check_fpu): Return the result rather than printing it.
12016 (end arch): Fix operator precedence.
12017 (end cpu): Likewise.
12018 (END): Print the result from check_fpu.
12020 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12021 Alan Hayward <alan.hayward@arm.com>
12022 David Sherwood <david.sherwood@arm.com>
12024 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12025 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12026 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12027 (*fcmuo<mode>_and): New patterns.
12029 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12031 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12032 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12033 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12034 (cmp_op, sve_imm_con): New code attributes.
12035 (SVE_COND_INT_CMP, imm_con): Delete.
12036 (cmp_op): Remove above unspecs from int attribute.
12037 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12039 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12040 comparison-specific unspecs.
12041 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12042 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12043 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12044 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12045 (*vec_fcm<cmp_op><mode>): Rename to...
12046 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12047 (*vec_fcmuo<mode>): Rename to...
12048 (*fcmuo<mode>): ...this and adjust likewise.
12049 (*pred_fcm<cmp_op><mode>): New pattern.
12050 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12051 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12053 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12055 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12056 (aarch64_emit_sve_predicated_cond): New function.
12057 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12058 (aarch64_emit_unspec_cond_or): Replace with...
12059 (aarch64_emit_sve_or_conds): ...this new function. Use
12060 aarch64_emit_sve_ptrue_op for the individual comparisons and
12061 aarch64_emit_binop to OR them together.
12062 (aarch64_emit_inverted_unspec_cond): Replace with...
12063 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12064 aarch64_emit_sve_ptrue_op for the comparison and
12065 aarch64_emit_unop to invert the result.
12066 (aarch64_expand_sve_vec_cmp_float): Update after the above
12067 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12069 2018-05-07 Nathan Sidwell <nathan@acm.org>
12071 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12072 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12073 (Backwards Compatibility): Likewise.
12075 2018-05-07 Luis Machado <luis.machado@linaro.org>
12079 2018-05-07 Luis Machado <luis.machado@linaro.org>
12081 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12082 <prefetch_dynamic_strides>: New const bool field.
12083 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12084 prefetch_dynamic_strides.
12085 (exynosm1_prefetch_tune): Likewise.
12086 (thunderxt88_prefetch_tune): Likewise.
12087 (thunderx_prefetch_tune): Likewise.
12088 (thunderx2t99_prefetch_tune): Likewise.
12089 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12091 (aarch64_override_options_internal): Update to set
12092 PARAM_PREFETCH_DYNAMIC_STRIDES.
12093 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12094 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12095 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12096 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12097 prefetch-dynamic-strides setting.
12099 2018-05-07 Luis Machado <luis.machado@linaro.org>
12101 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12102 <minimum_stride>: New const int field.
12103 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12104 minimum_stride field.
12105 (exynosm1_prefetch_tune): Likewise.
12106 (thunderxt88_prefetch_tune): Likewise.
12107 (thunderx_prefetch_tune): Likewise.
12108 (thunderx2t99_prefetch_tune): Likewise.
12109 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12110 (aarch64_override_options_internal): Update to set
12111 PARAM_PREFETCH_MINIMUM_STRIDE.
12112 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12113 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12114 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12115 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12116 stride is constant and is below the minimum stride threshold.
12118 2018-05-07 Luis Machado <luis.machado@linaro.org>
12120 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12123 2018-05-07 Luis Machado <luis.machado@linaro.org>
12125 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12126 <prefetch_dynamic_strides>: New const bool field.
12127 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12128 prefetch_dynamic_strides.
12129 (exynosm1_prefetch_tune): Likewise.
12130 (thunderxt88_prefetch_tune): Likewise.
12131 (thunderx_prefetch_tune): Likewise.
12132 (thunderx2t99_prefetch_tune): Likewise.
12133 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12135 (aarch64_override_options_internal): Update to set
12136 PARAM_PREFETCH_DYNAMIC_STRIDES.
12137 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12138 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12139 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12141 prefetch-dynamic-strides setting.
12143 2018-05-07 Luis Machado <luis.machado@linaro.org>
12145 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12146 <minimum_stride>: New const int field.
12147 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12148 minimum_stride field.
12149 (exynosm1_prefetch_tune): Likewise.
12150 (thunderxt88_prefetch_tune): Likewise.
12151 (thunderx_prefetch_tune): Likewise.
12152 (thunderx2t99_prefetch_tune): Likewise.
12153 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12154 (aarch64_override_options_internal): Update to set
12155 PARAM_PREFETCH_MINIMUM_STRIDE.
12156 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12157 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12158 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12159 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12160 stride is constant and is below the minimum stride threshold.
12162 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12165 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12166 the type is addressable. Don't force op into register if it has
12169 2018-05-05 Roland McGrath <mcgrathr@google.com>
12172 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12173 any section for which we don't know a specific type it should have,
12174 regardless of name. Previously this was done only for the exact
12175 names ".init_array", ".fini_array", and ".preinit_array".
12176 (default_elf_asm_named_section): Add comment about
12177 relationship with default_section_type_flags and SECTION_NOTYPE.
12178 (get_section): Don't consider it a type conflict if one side has
12179 SECTION_NOTYPE and the other doesn't, as long as neither has the
12180 SECTION_BSS et al used in the default_section_type_flags logic.
12182 2018-05-05 Tom de Vries <tom@codesourcery.com>
12185 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12186 (workaround_barsyncs): New function.
12187 (nvptx_reorg): Use workaround_barsyncs.
12188 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12189 (define_expand "nvptx_membar_cta"): New define_expand.
12190 (define_insn "*nvptx_membar_cta"): New insn.
12192 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
12194 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
12195 To improve optimization opportunities.
12196 * builtin-types.def: The new needed builtin types for the above.
12198 2018-05-04 Richard Biener <rguenther@suse.de>
12200 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
12201 * gimple-ssa-store-merging.c
12202 (imm_store_chain_info::output_merged_store): Remove redundant create,
12203 release split_store vector contents on failure.
12204 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
12205 scalar stmt vector on cache hit.
12207 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
12209 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
12211 * config.gcc (powerpc-xilinx-eabi*): Remove.
12212 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
12214 (fusion_addis_mem_combo_load): Ditto.
12215 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
12217 (rs6000_cpu_cpp_builtins): Ditto.
12218 * config/rs6000/rs6000-linux.c
12219 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
12220 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
12221 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
12223 (rs6000_setup_reg_addr_masks): Ditto.
12224 (rs6000_init_hard_regno_mode_ok): Ditto.
12225 (rs6000_option_override_internal): Ditto.
12226 (legitimate_lo_sum_address_p): Ditto.
12227 (rs6000_legitimize_address): Ditto.
12228 (rs6000_legitimize_reload_address): Ditto.
12229 (rs6000_legitimate_address_p): Ditto.
12230 (abi_v4_pass_in_fpr): Ditto.
12231 (setup_incoming_varargs): Ditto.
12232 (rs6000_gimplify_va_arg): Ditto.
12233 (rs6000_split_multireg_move): Ditto.
12234 (rs6000_savres_strategy): Ditto.
12235 (rs6000_emit_prologue_components): Ditto.
12236 (rs6000_emit_epilogue_components): Ditto.
12237 (rs6000_emit_prologue): Ditto.
12238 (rs6000_emit_epilogue): Ditto.
12239 (rs6000_elf_file_end): Ditto.
12240 (rs6000_function_value): Ditto.
12241 (rs6000_libcall_value): Ditto.
12242 * config/rs6000/rs6000.h: Ditto.
12243 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
12244 (TARGET_MINMAX): ... this. New.
12245 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
12246 * config/rs6000/rs6000.md: Remove Xilinx FP support.
12247 (*movsi_internal1_single): Delete.
12248 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
12249 mfpu=, mxilinx-fpu): Delete.
12250 * config/rs6000/singlefp.h: Delete.
12251 * config/rs6000/sysv4.h: Remove Xilinx FP support.
12252 * config/rs6000/t-rs6000: Ditto.
12253 * config/rs6000/t-xilinx: Delete.
12254 * config/rs6000/titan.md: Adjust for fp_type removal.
12255 * config/rs6000/vsx.md: Remove Xilinx FP support.
12256 (VStype_simple): Delete.
12257 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
12258 * config/rs6000/xfpu.h: Delete.
12259 * config/rs6000/xfpu.md: Delete.
12260 * config/rs6000/xilinx.h: Delete.
12261 * config/rs6000/xilinx.opt: Delete.
12262 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
12263 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
12265 2018-05-04 Tom de Vries <tom@codesourcery.com>
12268 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
12271 2018-05-04 Richard Biener <rguenther@suse.de>
12273 PR middle-end/85627
12274 * tree-complex.c (update_complex_assignment): We are always in SSA form.
12275 (expand_complex_div_wide): Likewise.
12276 (expand_complex_operations_1): Likewise.
12277 (expand_complex_libcall): Preserve EH info of the original stmt.
12278 (tree_lower_complex): Handle removed blocks.
12279 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
12280 on complex multiplication and division libcall builtins.
12282 2018-05-04 Richard Biener <rguenther@suse.de>
12284 PR middle-end/85574
12285 * fold-const.c (negate_expr_p): Restrict negation of operand
12286 zero of a division to when we know that can happen without
12288 (fold_negate_expr_1): Likewise.
12290 2018-05-04 Jakub Jelinek <jakub@redhat.com>
12293 * real.h (real_nextafter): Declare.
12294 * real.c (real_nextafter): New function.
12295 * fold-const-call.c (fold_const_nextafter): New function.
12296 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
12297 CASE_CFN_NEXTTOWARD.
12298 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
12299 even when arg1_mode is different from arg0_mode.
12301 2018-05-03 Nathan Sidwell <nathan@acm.org>
12303 * doc/extend.texi (Deprecated Features): Remove
12304 -ffriend-injection.
12305 (Backwards Compatibility): Likewise.
12306 * doc/invoke.texi (C++ Language Options): Likewise.
12307 (C++ Dialect Options): Likewise.
12309 2018-05-03 Jakub Jelinek <jakub@redhat.com>
12312 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
12313 _mm512_mask_mullox_epi64): New intrinsics.
12315 2018-05-03 Tom de Vries <tom@codesourcery.com>
12318 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12319 dump files): Add offload-tree.
12321 2018-05-03 Richard Biener <rguenther@suse.de>
12323 PR tree-optimization/85615
12324 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
12325 to loops not nested in BBs loop father to avoid creating multi-entry
12328 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12330 PR tree-optimization/70291
12331 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
12332 arguments. Change return type to tree. Emit libcall as a new
12333 statement rather than replacing existing one when inplace_p is true.
12334 (expand_complex_multiplication_components): New function.
12335 (expand_complex_multiplication): Expand floating-point complex
12336 multiplication using the above.
12337 (expand_complex_division): Rename inner_type parameter to type.
12338 Update expand_complex_libcall call-site.
12339 (expand_complex_operations_1): Update expand_complex_multiplication
12340 and expand_complex_division call-sites.
12342 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12345 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12346 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12347 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
12348 the highest significant bit of the shift count mask is clear. In
12349 check whether and[sq]i3 is needed verify that all significant bits
12350 of the shift count other than the highest are set.
12352 2018-05-02 Tom de Vries <tom@codesourcery.com>
12355 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
12356 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
12357 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
12358 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
12359 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
12360 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
12361 __builtin_goacc_parlevel_size.
12363 2018-05-02 Richard Biener <rguenther@suse.de>
12365 PR tree-optimization/85597
12366 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
12367 do not use split vect_get_vec_defs call but call vect_get_slp_defs
12370 2018-05-02 Tom de Vries <tom@codesourcery.com>
12373 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12374 dump files): Add ltrans-tree.
12376 2018-05-02 Tom de Vries <tom@codesourcery.com>
12379 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12380 dump files): Add wpa-ipa.
12382 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
12384 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
12385 powerpc*-*-linux*paired* target.
12386 * config/rs6000/750cl.h: Delete.
12387 * config/rs6000/paired.h: Delete.
12388 * config/rs6000/paired.md: Delete.
12389 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
12391 * config/rs6000/rs6000-builtin.def: Remove paired float support.
12392 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
12393 comment. Remove paired float support.
12394 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
12395 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
12397 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
12398 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
12400 * config/rs6000/rs6000.c: Remove paired float support.
12401 (paired_expand_vector_init, paired_expand_vector_move,
12402 paired_emit_vector_compare, paired_emit_vector_cond_expr,
12403 (paired_expand_lv_builtin, paired_expand_stv_builtin,
12404 paired_expand_builtin, paired_expand_predicate_builtin,
12405 paired_init_builtins): Delete.
12406 * config/rs6000/rs6000.h: Remove paired float support.
12407 * config/rs6000/rs6000.md: Remove paired float support.
12408 (move_from_CR_ov_bit): Delete.
12409 * config/rs6000/rs6000.opt (mpaired): Delete.
12410 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
12411 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
12413 2018-05-02 Richard Biener <rguenther@suse.de>
12415 PR middle-end/85567
12416 * gimplify.c (gimplify_save_expr): When in SSA form allow
12417 SAVE_EXPRs to compute to SSA vars.
12419 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12422 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12423 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12424 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
12425 clobber operands[2], instead use a new pseudo. Formatting fixes.
12427 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
12429 PR tree-optimization/85586
12430 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
12431 exit early for statements in the same group if the accesses are
12434 2018-05-02 Tom de Vries <tom@codesourcery.com>
12437 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
12440 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
12442 PR tree-optimization/85143
12443 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
12445 2018-05-01 Tom de Vries <tom@codesourcery.com>
12448 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
12449 not found" error message.
12451 2018-05-01 Tom de Vries <tom@codesourcery.com>
12454 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
12455 * vec.c (test_ordered_remove_if): New function.
12456 (vec_c_tests): Call test_ordered_remove_if.
12457 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
12458 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
12459 * tree-vect-patterns.c (vect_pattern_recog_1): Use
12460 VEC_ORDERED_REMOVE_IF.
12462 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12464 PR tree-optimization/82665
12465 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
12466 pointer subtraction where arguments come from a memchr call.
12468 2018-05-01 Jakub Jelinek <jakub@redhat.com>
12470 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
12471 --push-state --as-needed and --pop-state instead of --as-needed and
12472 --no-as-needed if ld supports it.
12473 * configure: Regenerated.
12476 * doc/install.texi2html: Replace _002d with - and _002a with * in
12477 generated html files using sed.
12479 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12482 * gcc-rich-location.c (blank_line_before_p): New function.
12483 (use_new_line): New function.
12484 (gcc_rich_location::add_fixit_insert_formatted): New function.
12485 * gcc-rich-location.h
12486 (gcc_rich_location::add_fixit_insert_formatted): New function.
12488 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12490 * selftest.c (assert_streq): Rename "expected" and "actual" to
12491 "val1" and "val2". Extend NULL-handling to cover both inputs
12492 symmetrically, while still requiring both to be non-NULL for a pass.
12493 * selftest.h (assert_streq): Rename "expected" and "actual" to
12495 (ASSERT_EQ): Likewise.
12496 (ASSERT_EQ_AT): Likewise.
12497 (ASSERT_KNOWN_EQ): Likewise.
12498 (ASSERT_KNOWN_EQ_AT): Likewise.
12499 (ASSERT_NE): Likewise.
12500 (ASSERT_MAYBE_NE): Likewise.
12501 (ASSERT_MAYBE_NE_AT): Likewise.
12502 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
12503 the assertion to pass.
12504 (ASSERT_STREQ_AT): Likewise.
12506 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12508 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
12509 interaction with -pie.
12511 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12513 * selftest.h: Fix alphabetization of per-source-file selftest
12516 2018-04-30 Jason Merrill <jason@redhat.com>
12518 PR c++/61982 - dead stores to destroyed objects.
12519 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
12522 2018-04-30 Jason Merrill <jason@redhat.com>
12524 * tree.c (build_clobber): New.
12525 * tree.h: Declare it.
12526 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
12528 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12530 * diagnostic-show-locus.c (layout::layout): Update for
12531 location_get_source_line returning a char_span.
12532 (struct char_span): Move to input.h.
12533 (struct correction): Update for fields in char_span becoming
12535 (struct source_line): Update for location_get_source_line
12536 returning a char_span.
12537 (layout::print_line): Likewise.
12538 * edit-context.c (edited_file::print_content): Likewise.
12539 (edited_file::print_diff_hunk): Likewise.
12540 (edited_file::print_run_of_changed_lines): Likewise.
12541 (edited_file::get_num_lines): Likewise.
12542 (edited_line::edited_line): Likewise.
12543 * final.c (asm_show_source): Likewise.
12544 * input.c (location_get_source_line): Convert return type
12545 from const char * to char_span, losing the final "line_len"
12547 (dump_location_info): Update for the above.
12548 (get_substring_ranges_for_loc): Likewise. Use a char_span
12549 when handling the literal within the line.
12550 (test_reading_source_line): Update for location_get_source_line
12551 returning a char_span.
12552 * input.h (class char_span): Move here from
12553 diagnostic-show-locus.c, converting from a struct to a class.
12554 Make data members private.
12555 (char_span::operator bool): New.
12556 (char_span::length): New.
12557 (char_span::get_buffer): New.
12558 (char_span::operator[]): New.
12559 (char_span::subspan): Make const.
12560 (char_span::xstrdup): New.
12561 (location_get_source_line): Convert return type from const char *
12562 to char_span, losing the final "line_size" param.
12564 2018-04-30 Jan Hubicka <jh@suse.cz>
12566 * lto-wrapper.c (ltrans_priorities): New static var.
12567 (cmp_priority): New.
12568 (run_gcc): Read priorities and if doing parallel build order
12569 the Makefile by them.
12571 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12573 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
12575 2018-04-30 Richard Biener <rguenther@suse.de>
12577 * tree-cfg.c (verify_address): Remove base argument, add
12578 flag whether to check TREE_ADDRESSABLE and do that.
12579 (verify_expr): Remove.
12580 (verify_types_in_gimple_reference): Add pieces from verify_expr.
12581 (verify_gimple_assign_single): Likewise.
12582 (verify_gimple_switch): Likewise.
12583 (verify_expr_location_1): Dereference tp once. Add (disabled)
12584 piece from verify_expr.
12585 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
12587 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12589 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
12591 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12593 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
12594 (small_data_pattern): Likewise.
12595 (arc_rewrite_small_data): Likewise.
12596 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
12597 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
12598 (get_symbol_alignment): New function.
12599 (legitimate_small_data_address_p): Likewise.
12600 (legitimate_scaled_address): Update, call
12601 legitimate_small_data_address_p.
12602 (output_sdata): New static variable.
12603 (arc_print_operand): Update how we handle small data operands.
12604 (arc_print_operand_address): Likewise.
12605 (arc_legitimate_address_p): Update, use
12606 legitimate_small_data_address_p.
12607 (arc_rewrite_small_data_p): Remove.
12608 (arc_rewrite_small_data_1): Likewise.
12609 (arc_rewrite_small_data): Likewise.
12610 (small_data_pattern): Likewise.
12611 (compact_sda_memory_operand): Update to use
12612 legitimate_small_data_address_p and get_symbol_alignment.
12613 (prepare_move_operands): Don't rewite sdata pattern.
12614 (prepare_extend_operands): Remove.
12615 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
12617 (zero_extendqisi2): Likewise.
12618 (zero_extendhisi2): Likewise.
12619 (extendqihi2): Likewise.
12620 (extendqisi2): Likewise.
12621 (extendhisi2): Likewise.
12622 (addsi3): Likewise.
12623 (subsi3): Likewise.
12624 (andsi3): Likewise.
12625 * config/arc/constraints.md (Usd): Change it to memory constraint.
12627 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12629 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
12630 as source of std instructions.
12631 * config/arc/arc.md (movsi_insn): Update pattern predicate to
12632 allow 6-bit constants as source for store instructions.
12633 (movdi_insn): Update instruction pattern to allow 6-bit constants
12634 as source for store instructions.
12636 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12638 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
12640 2018-04-30 Nathan Sidwell <nathan@acm.org>
12641 Sandra Loosemore <sandra@codesourcery.com>
12643 * dumpfile.c (dump_open): Allow '-' for stdout.
12644 * doc/invoke.texi (Developer Options): Document dump filename
12645 determination early. Document stdin/stdout selection.
12647 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
12649 Microblaze Target: PIC data text relative
12651 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
12652 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
12654 * config/microblaze/microblaze.h (microblaze_constant_address_p):
12655 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
12656 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
12657 New addressing mode for data-text relative position indepenedent code.
12658 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
12659 'ADDRESS_SYMBOLIC_TXT_REL'.
12660 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
12661 (microblaze_legitimate_pic_operand): Exclude function calls from
12662 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
12663 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
12665 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12666 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12667 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
12668 for 'address + offset'.
12669 (microblaze_expand_prologue): Add new function prologue call for
12671 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
12672 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
12673 table in case of TARGET_PIC_DATA_TEXT_REL.
12674 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
12675 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
12676 Add new macros 'UNSPEC_TEXT',
12677 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
12678 + exclude function calls from 'UNSPEC_PLT' in case of data text
12680 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12681 new target hook for generating address diff vector tables in case of
12683 * doc/tm.texi : Regenerate.
12684 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
12685 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
12686 of addr diff vector generation.
12687 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12688 target hook definition.
12689 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
12690 Add default function for generate_pic_addr_diff_vec -> flag_pic.
12691 * doc/invoke.texi (Add new pic option): Add new microblaze pic
12692 option for data text relative.
12694 2018-04-30 Richard Biener <rguenther@suse.de>
12696 * tree-chrec.h (evolution_function_is_constant_p): Remove
12698 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
12700 2018-04-30 Richard Biener <rguenther@suse.de>
12703 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12705 2018-04-30 Richard Biener <rguenther@suse.de>
12707 PR tree-optimization/28364
12708 PR tree-optimization/85275
12709 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
12710 copying first exit test.
12712 2018-04-28 Mark Wielaard <mark@klomp.org>
12714 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
12715 dwarf_version >= 5.
12716 (dwarf_AT): Handle DW_AT_addr_base.
12717 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
12719 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
12722 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
12723 (*ashl<dwi>3_doubleword_mask_1): Ditto.
12724 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
12725 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
12727 2018-04-28 Richard Biener <rguenther@suse.de>
12729 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
12730 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
12731 to reflect use. Only add interesting stmts.
12733 2018-04-27 Martin Jambor <mjambor@suse.cz>
12736 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
12737 the jump function allows for passing through aggregate values.
12739 2018-04-27 David Malcolm <dmalcolm@redhat.com>
12741 * input.h (in_system_header_at): Convert from macro to inline
12743 (from_macro_expansion_at): Likewise.
12744 (from_macro_definition_at): Likewise.
12746 2018-04-27 Jeff Law <law@redhat.com>
12748 * config.gcc: Mark tile* targets as deprecated/obsolete.
12750 2018-04-27 Richard Biener <rguenther@suse.de>
12752 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
12755 2018-04-27 Richard Biener <rguenther@suse.de>
12757 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
12759 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
12761 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
12762 with Yd constraint. Set "preferred_for_speed" attribute from
12763 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
12764 with Yd constraint.
12765 (*movdi_internal): Ditto.
12766 (movti_interunit splitters): Remove
12767 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
12768 (movdi_interunit splitters): Ditto.
12769 * config/i386/constraints.md (Ye): Remove.
12770 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
12772 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12775 * config/aarch64/constraints.md (Usg): Limit to 31.
12776 (Usj): Limit to 63.
12778 2018-04-27 Jakub Jelinek <jakub@redhat.com>
12780 PR tree-optimization/85529
12781 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
12782 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
12783 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
12784 zero extension or masking of the MSB bit.
12785 (optimize_range_tests): Add FIRST_BB argument, pass it through
12786 to optimize_range_tests_var_bound.
12787 (maybe_optimize_range_tests, reassociate_bb): Adjust
12788 optimize_range_tests callers.
12790 2018-04-26 Richard Biener <rguenther@suse.de>
12791 Jakub Jelinek <jakub@redhat.com>
12793 * cgraph.h (symbol_table): Just declare debug method here.
12794 * symtab.c (symbol_table::debug): Define.
12796 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
12798 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
12800 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
12802 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
12803 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
12804 (*movdi_internal): Substitute Yi and Yj constraint with x
12805 and Ym and Yn constraint with y constraint. Update "isa"
12806 attribute and set "preferred_for_speed" attribute from
12807 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
12808 (*movsi_internal): Ditto.
12809 (*movdf_internal): Ditto.
12810 (*movsf_internal): Ditto.
12811 (*zero_extendsidi2): Ditto.
12812 * config/i386/sse.md (vec_set<mode>_0): Ditto.
12813 (sse2_loadld): Ditto.
12814 (*vec_extract<ssevecmodelower>_0): Ditto.
12815 (*vec_extractv4si_0_zext_sse4): Ditto.
12816 (vec_concatv2di): Ditto.
12817 (*vec_dup<mode>): Ditto.
12818 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
12819 * config/i386/constraints.md (Yi): Remove.
12824 2018-04-26 Nathan Sidwell <nathan@acm.org>
12826 * dumpfile.c (dump_open): New.
12827 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
12828 (dump_finish): Detect stdio/stderr by value not name.
12830 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
12832 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
12834 2018-04-26 Tom de Vries <tom@codesourcery.com>
12837 * config/nvptx/nvptx.c (verify_neutering_jumps)
12838 (verify_neutering_labels): New function
12839 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12841 2018-04-26 Tom de Vries <tom@codesourcery.com>
12844 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12845 (nvptx_single): Use needs_neutering_p to skip over insns that do not
12848 2018-04-26 Richard Biener <rguenther@suse.de>
12849 Tom de Vries <tom@codesourcery.com>
12852 * lto-streamer-out.c (output_function): Fixup loops if required to match
12853 discovery done in the reader.
12855 2018-04-26 Richard Biener <rguenther@suse.de>
12857 PR tree-optimization/85116
12858 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
12859 have a loop exit from the single latch predecessor. Remove
12860 case of header with just condition.
12861 (ch_base::copy_headers): Exclude infinite loops from any
12863 (pass_ch::execute): Record exits.
12865 2018-04-26 Richard Biener <rguenther@suse.de>
12867 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
12868 prologue cost vector and pass it to vect_get_load_cost.
12869 (vect_get_peeling_costs_all_drs): Likewise.
12870 (vect_peeling_hash_get_lowest_cost): Likewise.
12871 (vect_enhance_data_refs_alignment): Likewise.
12873 2018-04-26 Richard Biener <rguenther@suse.de>
12875 PR middle-end/85450
12876 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
12877 checking of integer<->pointer conversions.
12878 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
12879 sign-/zero-extending pointer types.
12880 (expand_omp_for_static_chunk): Likewise.
12882 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
12883 Jean Lee <xiaoyur347@gmail.com>
12885 * config/mips/mips.c (mips_asan_shadow_offset): New function.
12886 (TARGET_ASAN_SHADOW_OFFSET): Define.
12887 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
12888 true for -fsanitize=address.
12890 2018-04-25 Mark Wielaard <mark@klomp.org>
12892 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
12895 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12897 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
12898 than "alu", remove explicit "memory" and "imm_disp" attributes.
12899 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
12901 PR middle-end/85414
12902 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
12903 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
12904 gen_lowpart_no_emit.
12906 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
12909 * config/i386/i386.c (ix86_expand_builtin): Change memory
12910 operand to XI, extend p0 to Pmode.
12911 * config/i386/i386.md: Change unspec volatile and operand
12912 1 mode to XI, change operand 0 mode to P.
12914 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12916 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
12917 GET_MODE_MASK before any checking.
12918 (nds32_can_use_bset_p): Likewise.
12919 (nds32_can_use_btgl_p): Likewise.
12921 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12923 * config/nds32/nds32-doubleword.md: New define_split pattern for
12924 illegal register number.
12926 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12928 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
12930 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12932 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
12934 2018-04-25 Richard Biener <rguenther@suse.de>
12936 * lto-streamer.h (LTO_major_version): Bump to 8.
12938 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12940 * BASE-VER: Set to 9.0.0.
12942 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
12944 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
12945 in __abskf2 and __powikf2.
12947 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12950 * config/aarch64/constraints.md (Usg, Usj): New constraints.
12951 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
12952 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12953 Use the above on operand 2. Reindent.
12954 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12956 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12959 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12961 * config/i386/i386.opt (mcet): Removed.
12962 * doc/install.texi: Remove -mcet documentation.
12963 * doc/invoke.texi: Likewise.
12965 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12968 * doc/install.texi: Remove -mcet from bootstrap-cet.
12970 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12973 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12974 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12978 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12979 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12980 containing a CONST_VECTOR.
12982 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12984 * doc/install.texi: Update newlib dependency for nvptx.
12986 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12989 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12990 instead of INTVAL when shifting x left.
12992 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12994 PR tree-optimization/85478
12995 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12996 vect_grouped_store_supported for single element vectors.
12998 2018-04-24 Richard Biener <rguenther@suse.de>
13001 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13002 load cost increase to the case of non-constant step.
13004 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13007 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13008 destination if any_malformed_asm.
13010 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13012 PR middle-end/85496
13013 * expr.c (store_field): In the bitfield case, if the value comes from
13014 a function call and is returned in registers by means of a PARALLEL,
13015 do not change the mode of the temporary unless BLKmode and VOIDmode.
13017 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13019 PR rtl-optimization/85423
13020 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13021 dependencies to debug insns when the previous insn is non-debug.
13023 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13025 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13026 enums into a single definition.
13027 (fls): Fix predicates and printing.
13030 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13032 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13033 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13034 and short u6 immediate.
13035 (check_if_valid_sleep_operand): Remove.
13036 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13038 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13040 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13041 flag_always_save_lp condition.
13042 * config/nds32/nds32.opt (malways-save-lp): New option.
13044 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13046 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13047 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13048 * config/nds32/nds32.h
13049 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13050 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13052 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13054 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13055 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13057 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13058 Chung-Ju Wu <jasonwucj@gmail.com>
13060 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13062 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13063 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13065 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13067 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13069 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13071 * config/nds32/nds32-protos.h (nds32_data_alignment,
13072 nds32_local_alignment): Declare.
13073 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13074 nds32_local_alignment): New functions.
13075 (TARGET_CONSTANT_ALIGNMENT): Define.
13076 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13078 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13080 * config/nds32/nds32.c
13081 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13082 (TARGET_MODES_TIEABLE_P): Likewise.
13084 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13086 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13087 level Ofast and Og.
13089 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13090 Chung-Ju Wu <jasonwucj@gmail.com>
13092 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13093 for unaligned access.
13094 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13096 * config/nds32/nds32-intrinsic.md: Likewise.
13097 * config/nds32/nds32_intrinsic.h: Likewise.
13098 * config/nds32/nds32.h (nds32_builtins): Likewise.
13099 * config/nds32/nds32.opt (munaligned-access): New option.
13100 * config/nds32/nds32.c (nds32_asm_file_start): Display
13101 flag_unaligned_access status.
13103 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13105 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13106 -mno-relax is present.
13107 * config/riscv/linux.h (LINK_SPEC): Ditto.
13109 2018-04-20 Martin Sebor <msebor@redhat.com>
13112 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13114 (gimple_fold_builtin_stxcpy_chk): Same.
13115 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13117 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13120 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13121 __powikf2 when long double is IEEE 128-bit.
13123 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13125 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13126 step to make sure stack always aligned.
13128 2018-04-20 Carl Love <cel@us.ibm.com>
13131 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13132 size check for arg0.
13134 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13135 Tom de Vries <tom@codesourcery.com>
13138 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13139 Emit insns for calls too.
13140 (nvptx_find_par): Always look for worker-level predecessor insn.
13141 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13143 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13144 (nvptx_process_pars): Propagate frames for calls.
13146 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13149 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13151 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13152 (ix86_handle_option): Don't handle OPT_mibt.
13153 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13155 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13157 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13158 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13159 (ix86_target_macros): Define __CET__ with flag_cf_protection
13160 for -fcf-protection.
13161 * config/i386/i386.c (isa2_opts): Remove -mibt.
13162 * config/i386/i386.h (TARGET_IBT): Removed.
13163 (TARGET_IBT_P): Likewise.
13164 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13165 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13166 * config/i386/i386.opt (mcet): Update help message.
13167 (mshstk): Likewise.
13169 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13170 -mcet as an alias for -mshstk.
13172 2018-04-20 Richard Biener <rguenther@suse.de>
13174 PR middle-end/85475
13175 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13176 complexity by forcing a single use of the multiply operand.
13178 2018-04-20 Martin Jambor <mjambor@suse.cz>
13181 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13182 recursion dependency to only apply to non-clones.
13184 2018-04-20 Martin Jambor <mjambor@suse.cz>
13187 * ipa-cp.c (create_specialized_node): Check that clones of
13188 self-recursive edges exist during IPA-CP.
13190 2018-04-19 Toon Moene <toon@moene.org>
13192 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
13195 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13197 PR tree-optimization/85467
13198 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
13199 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
13200 VECTOR_CST element to type.
13202 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13205 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
13206 * config/i386/i386.md (builtin_setjmp_setup): Removed.
13207 (builtin_longjmp): Likewise.
13208 (save_stack_nonlocal): New pattern.
13209 (restore_stack_nonlocal): Likewise.
13211 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13214 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13215 Replace ASM_OUTPUT_LABEL with fprintf.
13217 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13220 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13221 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
13222 * config/i386/i386-c.c (ix86_target_macros_internal): Also
13223 define __IBT__ and __SHSTK__ for -fcf-protection.
13224 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
13226 (ix86_trampoline_init): Likewise.
13227 (x86_output_mi_thunk): Likewise.
13228 (ix86_notrack_prefixed_insn_p): Likewise.
13229 (ix86_option_override_internal): Don't disallow -fcf-protection.
13230 * config/i386/i386.md (rdssp<mode>): Also enable for
13232 (incssp<mode>): Likewise.
13233 (nop_endbr): Likewise.
13234 * config/i386/i386.opt (mcet): Change help message to built-in
13237 (mshstk): Likewise.
13238 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
13239 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
13240 enable CET built-in functions.
13242 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
13244 * common/config/i386/i386-common.c
13245 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
13246 OPTION_MASK_ISA_MOVDIRI_UNSET,
13247 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
13248 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
13249 * config.gcc (movdirintrin.h): New header.
13250 * config/i386/cpuid.h (bit_MOVDIRI,
13251 bit_MOVDIR64B): New bits.
13252 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
13254 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
13255 (VOID, PVOID, PCVOID)): New function types.
13256 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
13257 __builtin_ia32_directstoreu_u64,
13258 __builtin_ia32_movdir64b): New builtins.
13259 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
13260 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
13262 (ix86_valid_target_attribute_inner_p): Ditto.
13263 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
13264 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
13265 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
13266 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
13267 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
13268 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
13269 (movdiri<mode>, movdir64b_<mode>): New.
13270 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
13271 * config/i386/immintrin.h: Include movdirintrin.h.
13272 * config/i386/movdirintrin.h: New file.
13273 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
13275 2018-04-19 Richard Biener <rguenther@suse.de>
13277 PR middle-end/85455
13278 * cfg.c (clear_bb_flags): When loop state says we have
13279 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
13281 2018-04-19 Richard Biener <rguenther@suse.de>
13283 PR tree-optimization/84737
13284 * tree-vect-data-refs.c (vect_copy_ref_info): New function
13285 copying restrict info.
13286 (vect_setup_realignment): Use it.
13287 * tree-vectorizer.h (vect_copy_ref_info): Declare.
13288 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
13289 the first DR to all generated stores.
13290 (vectorizable_load): Likewise for loads.
13292 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13294 PR tree-optimization/85446
13295 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
13296 the integral and pointer types to have the same precision.
13298 * doc/install.texi: Document --disable-cet being the default and
13301 2018-04-18 Martin Liska <mliska@suse.cz>
13303 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
13306 2018-04-18 Martin Liska <mliska@suse.cz>
13309 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13312 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13313 arguments if they are comparable.
13315 2018-04-18 Martin Liska <mliska@suse.cz>
13318 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13321 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13324 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
13327 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
13328 ENDBR after calling __morestack.
13330 2018-04-18 David Malcolm <dmalcolm@redhat.com>
13333 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
13334 by using gcc_base_ver to generate a gcc_driver_version, and use
13335 it when generating GCC_DRIVER_NAME.
13336 * configure: Regenerate.
13338 2018-04-18 Jakub Jelinek <jakub@redhat.com>
13341 * config.gcc: Obsolete powerpc*-*-*spe*.
13343 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13346 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
13347 (stabstr_D): Change type of unum from unsigned int to
13348 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
13351 2018-04-17 Jim Wilson <jimw@sifive.com>
13354 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
13355 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
13356 Set arg_pointer_offset after using pretend_args_size.
13358 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13360 PR rtl-optimization/85431
13361 * dse.c (record_store): Ignore zero width stores.
13364 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
13365 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
13366 __builtin_stack_restore rather than after it.
13367 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
13368 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
13369 argument instead of virtual_dynamic_stack_rtx.
13371 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
13373 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
13375 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
13376 Add note to error message to explain internal mapping of overloaded
13377 built-in function name to non-overloaded built-in function name.
13378 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
13381 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13384 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
13385 where the inputs overlap with the output.
13387 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13389 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
13390 (=v, v) alternative and explicit "memory" attribute.
13391 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
13392 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13394 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
13395 "sselog1" type instead of "sselog".
13396 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
13397 "sselog". Remove explicit "memory" attribute.
13398 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
13399 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13401 (vec_extract_hi_v32hi): Merge all alternatives into one, use
13402 "sselog1" type instead of "sselog". Remove explicit "memory"
13404 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
13405 use "sselog1" type instead of "sselog". Remove explicit "memory"
13407 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
13408 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13410 (vec_extract_hi_v64qi): Merge all alternatives into one, use
13411 "sselog1" type instead of "sselog". Remove explicit "memory"
13413 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
13414 use "sselog1" type instead of "sselog". Remove explicit "memory"
13418 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
13420 PR middle-end/85414
13421 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
13424 2018-04-17 Martin Jambor <mjambor@suse.cz>
13427 * ipa-cp.c (create_specialized_node): Call
13428 expand_all_artificial_thunks if necessary.
13430 2018-04-17 Martin Liska <mliska@suse.cz>
13433 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
13434 in message, remote space in between '_G' and '('.
13436 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13439 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
13440 avx512f_vmcmp<mode>3<round_saeonly_name>,
13441 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
13442 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
13443 avx512f_rndscale<mode><round_saeonly_name>,
13444 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
13445 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13446 Use %<iptr>2 instead of %2 for -masm=intel.
13447 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
13448 avx512f_vcvttss2usi<round_saeonly_name>,
13449 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
13451 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
13452 avx512f_vcvttsd2usi<round_saeonly_name>,
13453 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
13454 Use %q1 instead of %1 for -masm=intel.
13455 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
13456 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
13457 of %3 for -masm=intel.
13458 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
13460 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
13462 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
13464 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
13465 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
13467 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
13469 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
13470 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
13471 %g1 and one with %0 and %1.
13472 (avx512er_vmrcp28<mode><round_saeonly_name>,
13473 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
13474 %1 for -masm=intel.
13475 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
13476 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
13477 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
13478 of %0 and %{%4%} for -masm=intel.
13479 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
13480 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
13481 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
13482 order of %0 and %{%5%}%{z%} for -masm=intel.
13484 2018-04-17 Jan Hubicka <jh@suse.cz>
13487 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
13489 2018-04-17 Martin Liska <mliska@suse.cz>
13492 * multiple_target.c (create_dispatcher_calls): Set apostrophes
13493 for target_clone error message. Make default implementation
13494 clone to be a local declaration.
13495 (separate_attrs): Add new argument and check for an empty
13497 (expand_target_clones): Handle it.
13498 (ipa_target_clone): Make redirection just for target_clones
13501 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
13502 Tom de Vries <tom@codesourcery.com>
13504 PR middle-end/84955
13505 * omp-expand.c (expand_oacc_for): Add dummy false branch for
13506 tiled basic blocks without omp continue statements.
13508 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
13511 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
13512 vec_extract expression as having side effects to make sure it gets
13515 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
13518 * config/i386/i386.c (get_builtin_code_for_version): Check
13521 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
13524 * config.gcc: Support "skylake".
13525 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13527 * config/i386/i386.c (m_SKYLAKE): Define.
13528 (processor_target_table): Add "skylake".
13529 (ix86_option_override_internal): Add "skylake".
13530 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
13531 PROCESSOR_CANNONLAKE.
13532 (get_builtin_code_for_version): Fix priority for
13533 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
13534 PROCESSOR_SKYLAKE-AVX512.
13535 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
13536 (processor_type): Add PROCESSOR_SKYLAKE.
13538 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
13539 Jason Merrill <jason@redhat.com>
13542 * convert.c (convert_to_integer_1): Use direct recursion for
13543 enumeral types and types with a precision less than the number
13544 of bits in their mode.
13546 2018-04-16 Julia Koval <julia.koval@intel.com>
13549 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
13550 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
13552 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
13555 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
13556 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
13557 and -mno-direct-move.
13559 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
13562 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
13563 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
13564 Ensure negative shifts result in {0}.
13566 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
13568 PR rtl-optimization/79916
13569 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
13570 regs (if any) to define how to gnerate SD moves when LRA is in
13573 2018-04-13 Jakub Jelinek <jakub@redhat.com>
13575 PR rtl-optimization/85393
13576 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
13577 * except.c (expand_dw2_landing_pad_for_region): Make static.
13578 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
13579 a label and unconditional jump to old_bb, rather than
13580 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
13583 PR rtl-optimization/85376
13584 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
13585 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
13586 instead of a specific value.
13588 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13589 Bin Cheng <bin.cheng@arm.com>
13591 PR tree-optimization/82965
13592 PR tree-optimization/83991
13593 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
13594 by_profile_only parameter.
13595 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
13596 information if the loop was predicted to iterate too many times.
13597 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
13599 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13602 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
13605 2018-04-13 Martin Liska <mliska@suse.cz>
13606 Jakub Jelinek <jakub@redhat.com>
13608 PR middle-end/81657
13609 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
13610 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
13611 * builtins.c (expand_builtin_memory_copy_args): Use
13612 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
13613 handle dest_addr == pc_rtx.
13615 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
13618 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
13619 asked to not generate direct moves.
13620 (fix_trunc<mode>si2_stfiwx): Similar.
13621 (fix_trunc<mode>si2_internal): Similar.
13623 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13626 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
13627 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
13628 lookup if dest in some wider mode is known to be const0_rtx and
13629 if so, record permanent equivalence for it to be ZERO_EXTEND of
13630 the narrower mode destination.
13632 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13634 * lto-streamer-out.c (output_function): Revert 259346.
13635 * omp-expand.c (expand_oacc_for): Likewise.
13637 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
13639 PR rtl-optimization/85354
13640 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
13641 * sel-sched.c (sel_global_init): ... here.
13643 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
13646 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
13647 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13648 mode for PE-COFF targets.
13649 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
13650 (i386_pe_asm_lto_end): Likewise.
13651 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
13652 (TARGET_ASM_LTO_END): Likewise.
13653 * config/i386/winnt.c (saved_debug_info_level): New static variable.
13654 (i386_pe_asm_lto_start): New function.
13655 (i386_pe_asm_lto_end): Likewise.
13657 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13658 Richard Biener <rguenther@suse.de>
13660 PR middle-end/84955
13661 * lto-streamer-out.c (output_function): Fix CFG loop state before
13663 * omp-expand.c (expand_oacc_for): Handle calls to internal
13664 functions like regular functions.
13666 2018-04-12 Richard Biener <rguenther@suse.de>
13669 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
13670 for the early LTO debug to properly generate references to it
13671 during DIE emission. Do not re-use that for the skeleton for
13673 (dwarf2out_early_finish): Likewise.
13675 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13678 * config/i386/sse.md
13679 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
13680 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
13681 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
13682 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
13683 and output is a reg, avoid creating invalid lowpart subreg, but
13684 instead split into a 512-bit move. Don't split if not AVX512VL,
13685 input is xmm16+ reg and output is a mem.
13686 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
13687 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
13688 xmm16+ reg and output is a mem.
13690 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13692 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
13693 also for flag_dwarf2_cfi_asm.
13695 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13697 PR rtl-optimization/85342
13698 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
13699 a bool scalar var inside of the loop instead. Don't try to update
13700 recog_data.operand after failed apply_change_group.
13702 2018-04-12 Tom de Vries <tom@codesourcery.com>
13705 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
13706 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13707 array with flexible array member as array without given dimension.
13708 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
13709 argument for undefined param to true.
13711 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
13714 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13715 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
13716 from PowerPC section.
13717 * config/rs6000/sysv4.opt (mcall-): Improve help text.
13718 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
13719 help text that is too long.
13720 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
13721 help text that is too long.
13722 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
13723 help text that is too long.
13725 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
13727 * config/alpha/alpha.md (stack_probe_internal): Rename
13728 from "probe_stack". Update all callers.
13730 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13732 PR rtl-optimization/84566
13733 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
13734 sched_macro_fuse_insns.
13736 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13739 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
13740 (compute_block_dependences): ... from here.
13742 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13744 PR tree-optimization/85331
13745 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
13746 from int to HOST_WIDE_INT.
13748 2018-04-11 Martin Jambor <mjambor@suse.cz>
13751 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
13752 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
13753 not the same as the source val.
13754 (cgraph_edge_brings_value_p): New parameter.
13755 (gather_edges_for_value): Pass destination value to
13756 cgraph_edge_brings_value_p.
13757 (perhaps_add_new_callers): Likewise.
13758 (get_info_about_necessary_edges): Likewise and exclude values brought
13759 only by self-recursive edges.
13760 (create_specialized_node): Redirect only clones of self-calling edges.
13761 (+self_recursive_pass_through_p): New function.
13762 (find_more_scalar_values_for_callers_subset): Use it.
13763 (find_aggregate_values_for_callers_subset): Likewise.
13764 (known_aggs_to_agg_replacement_list): Removed.
13765 (decide_whether_version_node): Re-calculate known constants for all
13766 remaining context clones.
13768 2018-04-11 Richard Biener <rguenther@suse.de>
13771 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
13772 from early DWARF output.
13773 (dwarf2out_early_finish): Output line info unconditionally into
13774 early DWARF and add reference to it.
13776 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13779 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
13780 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
13781 other than V2DFmode using iptr mode attribute.
13782 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
13784 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13786 PR rtl-optimization/84659
13787 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
13789 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13792 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
13794 (output_loc_list): Pass address of a dummy size variable even in the
13795 locview handling loop.
13796 (index_location_lists): Add comment on why skip_loc_list_entry can't
13799 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
13802 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
13805 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
13808 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13809 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
13810 and -mstring-compare-inline-limit.
13812 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13815 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
13816 for stack clash protection in a register whenever we need it to be in
13819 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13821 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
13822 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
13824 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13827 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
13829 (mlong-double-): Ditto.
13830 * config/rs6000/sysv4.opt (msdata=): Ditto.
13831 (mtls-size=): Ditto.
13833 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13835 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13836 erroneous entries for
13837 "vector int vec_ldl (int, long int *)", and
13838 "vector unsigned int vec_ldl (int, unsigned long int *)".
13839 Add comments and entries for
13840 "vector bool char vec_ldl (int, bool char *)",
13841 "vector bool short vec_ldl (int, bool short *)",
13842 "vector bool int vec_ldl (int, bool int *)",
13843 "vector bool long long vec_ldl (int, bool long long *)",
13844 "vector pixel vec_ldl (int, pixel *)",
13845 "vector long long vec_ldl (int, long long *)",
13846 "vector unsigned long long vec_ldl (int, unsigned long long *)".
13847 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
13848 type tree bool_long_long_type_node and correct definition of
13849 bool_V2DI_type_node to make reference to this new type tree.
13850 (rs6000_mangle_type): Replace erroneous reference to
13851 bool_long_type_node with bool_long_long_type_node.
13852 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
13853 comments to emphasize sign distinctions for char and int types and
13854 replace RS6000_BTI_bool_long constant with
13855 RS6000_BTI_bool_long_long constant. Also add comment to restrict
13856 use of RS6000_BTI_pixel.
13857 (bool_long_type_node): Remove this macro definition.
13858 (bool_long_long_type_node): New macro definition
13860 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13862 PR rtl-optimization/85300
13863 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
13864 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
13865 simplify_unary_operation fails.
13867 2018-04-10 Martin Liska <mliska@suse.cz>
13869 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
13870 cgraph_edge and ipa_ref.
13872 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13876 * config/i386/sse.md
13877 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
13878 computation of the VEC_MERGE selector from mask.
13879 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
13880 Fix decoding of the VEC_MERGE selector into mask.
13882 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13884 PR tree-optimization/85286
13885 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13887 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13889 * final.c (final_1): Set insn_last_address as well as
13890 insn_current_address.
13892 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13895 * explow.c (emit_stack_probe): Call validize_mem on memory location
13896 before passing it to gen_probe_stack. Create address operand and
13897 legitimize it for the probe_stack_address case.
13899 2018-04-09 Jan Hubicka <jh@suse.cz>
13902 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
13903 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
13904 * tree.c (free_lang_data_in_type): Fix handling of binfos;
13906 (free_lang_data): Rebuild type inheritance graph.
13908 2018-04-09 Martin Sebor <msebor@redhat.com>
13910 * invoke.texi (-finline-small-functions): Mention other optimization
13912 (-findirect-inlining, -fpartial-inlining): Same.
13913 (-finline-functions-called-once): Same.
13914 (-freorder-blocks-and-partition): Same.
13916 2018-04-09 Jan Hubicka <jh@suse.cz>
13919 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
13920 jumps; choose last target that matches the criteria (i.e.
13921 no partition changes for non-crossing jumps).
13922 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
13923 support for redirecting crossing jumps to non-crossing.
13925 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
13927 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
13928 also for naked functions.
13930 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
13932 * config/arc/arc.md (add_shift): New pattern.
13933 (add_shift2): Likewise.
13934 (sub_shift): Likewise.
13935 (sub_shift_cmp0_noout): Likewise.
13936 (compare_si_ashiftsi): Likewise.
13937 (xbfu_cmp0_noout): New combine pattern.
13938 (xbfu_cmp0"): Likewise.
13939 (movsi_set_cc_insn): Place the predicable variant first.
13940 (commutative_binary_cmp0_noout): Remove clobber.
13941 (commutative_binary_cmp0): New pattern.
13942 (noncommutative_binary_cmp0): Likewise.
13943 (noncommutative_binary_cmp0_noout): Likewise.
13944 (noncommutative_binary_comparison_result_used): Removed.
13945 (rsub_cmp0): New pattern.
13946 (rsub_cmp0_noout): Likewise.
13947 (extzvsi): Changed, keep only meaningful variants.
13948 (SQH, SEZ): New iterators.
13949 (SQH_postfix): New mode attribute.
13950 (SEZ_prefix): New code attribute.
13951 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
13952 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13953 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13954 of numerical value.
13955 (noncommutative_operator): Check the availability of barrel
13958 2018-04-09 Richard Biener <rguenther@suse.de>
13960 PR tree-optimization/85284
13961 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13962 Only use the niter constraining form of simple_iv when the exit
13963 is always executed.
13965 2018-04-09 Tom de Vries <tom@codesourcery.com>
13968 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13969 (define_expand "*memory_barrier"): New define_expand.
13970 (define_insn "memory_barrier"): New insn.
13972 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13974 PR rtl-optimization/80463
13975 PR rtl-optimization/83972
13976 PR rtl-optimization/83480
13978 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13979 correct producer for the insn.
13980 (tidy_control_flow): Fixup seqnos in case of debug insns.
13982 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13984 PR rtl-optimization/83913
13986 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13987 different sched-times when merging exprs.
13989 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13991 PR rtl-optimization/83962
13993 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13994 tidy_fallthru_edge and tidy_control_flow.
13996 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13998 PR rtl-optimization/83530
14000 * sel-sched.c (force_next_insn): New global variable.
14001 (remove_insn_for_debug): When force_next_insn is true, also leave only
14002 next insn in the ready list.
14003 (sel_sched_region): When the region wasn't scheduled, make another pass
14004 over it with force_next_insn set to 1.
14006 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14008 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14010 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14011 for interrupt control.
14012 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14013 functions for interrupt control.
14014 * config/nds32/nds32-intrinsic.md: Likewise.
14015 * config/nds32/nds32_intrinsic.h: Likewise.
14016 * config/nds32/nds32.h (nds32_builtins): Likewise.
14018 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14020 * config/nds32/nds32.c (nds32_init_machine_status,
14021 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14022 strict_aligned_p field.
14023 (nds32_expand_to_rtl_hook): New function.
14024 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14025 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14027 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14028 Chung-Ju Wu <jasonwucj@gmail.com>
14030 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14031 * config/nds32/nds32-n7.md: New file.
14032 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14033 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14035 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14036 * config/nds32/nds32.md (pipeline_model): Add n7.
14037 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14038 * config/nds32/pipelines.md: Include n7 settings.
14040 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14041 Chung-Ju Wu <jasonwucj@gmail.com>
14043 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14044 * config/nds32/nds32-e8.md: New file.
14045 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14046 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14048 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14049 * config/nds32/nds32.md (pipeline_model): Add e8.
14050 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14051 * config/nds32/pipelines.md: Include e8 settings.
14053 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14054 Chung-Ju Wu <jasonwucj@gmail.com>
14056 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14057 * config/nds32/nds32-n8.md: New file.
14058 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14059 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14061 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14062 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14063 * config/nds32/nds32.md (pipeline_model): Add n8.
14064 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14065 * config/nds32/pipelines.md: Include n8 settings.
14067 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14068 Chung-Ju Wu <jasonwucj@gmail.com>
14070 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14071 * config/nds32/nds32-n9-2r1w.md: New file.
14072 * config/nds32/nds32-n9-3r2w.md: New file.
14073 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14074 nds32_register_ports): New or modify for cpu n9.
14075 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14077 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14078 * config/nds32/nds32-utils.c: New file.
14079 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14080 TARGET_MUL_SLOW): Define.
14081 * config/nds32/nds32.md (pipeline_model): New attribute.
14082 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14083 New options that support cpu n9.
14084 * config/nds32/pipelines.md: Include n9 settings.
14085 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14087 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14089 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14090 information if necessary.
14091 (output_cond_branch_compare_zero): Likewise.
14092 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14093 (nds32_target_alignment): Refine for alignment.
14094 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14095 (FUNCTION_BOUNDARY): Modify.
14096 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14098 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14100 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14102 * config/nds32/constants.md (unspec_volatile_element): Add values for
14103 TLB operation and data prefetch.
14104 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14105 functions for TLB operation and data prefetch.
14106 * config/nds32/nds32-intrinsic.md: Likewise.
14107 * config/nds32/nds32_intrinsic.h: Likewise.
14108 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14109 (nds32_print_operand): Likewise.
14110 * config/nds32/nds32.h (nds32_builtins): Likewise.
14112 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14113 Andrew Pinski <pinsika@gcc.gnu.org>
14115 PR middle-end/82976
14116 * match.pd: Use constant_boolean_node of correct type instead of
14117 boolean_true_node or boolean_false_node for simplifying
14118 pointer comparisons to zero.
14120 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14122 PR tree-optimization/80021
14123 * tree.c (verify_type_variant): Make error call in verify_variant_match
14124 translatable and remove final full stop.
14126 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14128 * config/nds32/constants.md (unspec_volatile_element): Add
14129 UNSPEC_VOLATILE_EH_RETURN.
14130 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14131 nds32_output_stack_pop): Support dwarf exception handling process.
14132 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14133 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14134 exception handling process.
14135 (nds32_compute_stack_frame): Likewise.
14136 (nds32_return_addr_rtx): Likewise.
14137 (nds32_initial_elimination_offset): Likewise.
14138 (nds32_expand_prologue): Likewise.
14139 (nds32_expand_epilogue): Likewise.
14140 (nds32_dynamic_chain_address): New function.
14141 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14142 exception handling.
14143 (DYNAMIC_CHAIN_ADDRESS): Define.
14144 (EH_RETURN_DATA_REGNO): Define.
14145 (EH_RETURN_STACKADJ_RTX): Define.
14146 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14147 patterns for dwarf exception handling.
14149 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14151 * config/nds32/nds32.h: Clean up obsolete macros.
14153 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14155 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14156 Add enum values for particular instructions.
14157 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14158 particular intrinsic functions.
14159 * config/nds32/nds32-intrinsic.md: Likewise.
14160 * config/nds32/nds32_intrinsic.h: Likewise.
14161 * config/nds32/nds32.h (nds32_builtins): Likewise.
14162 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14163 (btst, ave): New patterns for particular instructions.
14165 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14167 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14168 Add enum values for atomic load/store and memory sync.
14169 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14171 * config/nds32/nds32-intrinsic.md: Likewise.
14172 * config/nds32/nds32_intrinsic.h: Likewise.
14173 * config/nds32/nds32.h (nds32_builtins): Likewise.
14175 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14177 PR tree-optimization/85257
14178 * fold-const.c (native_encode_vector): If not all elts could fit
14179 and off is -1, return 0 rather than offset.
14180 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14181 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14182 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14183 adjust buffer in native_interpret_expr call.
14185 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14187 * config/nds32/constants.md (unspec_volatile_element): Add cache
14188 control enum values.
14189 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
14190 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
14191 * config/nds32/nds32.c (nds32_cctl_names): New.
14192 (nds32_print_operand): Handle cache control register names.
14193 * config/nds32/nds32.h (nds32_builtins): New enum values.
14194 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
14196 * config/nds32/nds32.md (type): Add mmu.
14197 * config/nds32/pipelines.md (simple_insn): Add mmu.
14199 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14201 * config/nds32/nds32.md (type): Remove call.
14202 * config/nds32/pipelines.md (simple_insn): Likewise.
14204 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14206 * config/nds32/constants.md (unspec_volatile_element): Add
14207 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
14208 UNSPEC_VOLATILE_FMFCFG.
14209 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
14210 description for fmfcfg and fmfcsr.
14211 (bdesc_1arg): Add fmtcsr.
14212 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
14213 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
14214 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
14215 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
14216 unspec_fmfcfg): New patterns.
14217 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
14218 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
14219 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
14220 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
14221 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
14222 __nds32__fmfcfg): Define.
14224 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14226 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
14227 intrinsic register names.
14228 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
14229 intrinsic register enum values and macros.
14231 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14233 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
14234 for load/store addressing form.
14235 (nds32_print_operand_address): Likewise.
14237 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
14240 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
14241 based on LABEL_REF. Remove useless assertion.
14242 (pic_address_needs_scratch): Fix formatting.
14243 (sparc_legitimize_pic_address): Minor tweaks.
14244 (sparc_delegitimize_address): Adjust assertion accordingly.
14245 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
14246 into symbolic_operand.
14247 (movsi_high_pic_label_ref): Likewise.
14248 (movsi_lo_sum_pic_label_ref): Likewise.
14249 (movdi_pic_label_ref): Likewise.
14250 (movdi_high_pic_label_ref): Likewise.
14251 (movdi_lo_sum_pic_label_ref): Likewise.
14253 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
14255 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
14256 custom LIB_SPEC setup.
14258 2018-04-06 Ruslan Bukin <br@bsdpad.com>
14259 Kito Cheng <kito.cheng@gmail.com>
14261 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
14262 * config/riscv/freebsd.h: New.
14264 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14266 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
14267 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
14270 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14271 Kito Cheng <kito.cheng@gmail.com>
14273 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
14274 nds32_output_call, nds32_symbol_binds_local_p): New functions.
14275 * config/nds32/nds32-protos.h (nds32_output_call,
14276 nds32_output_return): Declare.
14277 * config/nds32/nds32.md: Refine all the call and return patterns.
14279 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14282 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
14283 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
14285 PR rtl-optimization/84872
14286 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
14287 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
14288 EDGE_CROSSING edge.
14290 2018-04-06 Tamar Christina <tamar.christina@arm.com>
14292 * expr.c (copy_blkmode_to_reg): Revert 254862.
14293 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
14295 2018-04-06 Richard Biener <rguenther@suse.de>
14297 PR middle-end/85244
14298 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
14299 after seeing a component reference with an adjacent field. Treat
14300 refs to arrays at struct end of external decls similar to
14301 refs to unconstrained commons.
14303 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14306 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
14307 look through SAVE_EXPRs with non-side-effects argument. Adjust
14309 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
14312 2018-04-06 Richard Biener <rguenther@suse.de>
14314 PR middle-end/85180
14315 * alias.c (find_base_term): New wrapper around find_base_term
14316 unwinding CSELIB_VAL_PTR changes.
14317 (find_base_term): Do not restore CSELIB_VAL_PTR during the
14320 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14322 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
14324 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
14325 constant definitions.
14326 ("nop"): lr 0,0 -> nopr r0
14327 ("nop_lr0", "nop_lr1"): New insn definitions.
14329 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14331 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
14332 NDS32_V3PUSH_AVAILABLE_P macro.
14334 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
14335 Chung-Ju Wu <jasonwucj@gmail.com>
14337 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
14338 (nds32*-*-*): Add float and fpu_config into supported_defaults.
14339 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
14340 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
14341 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
14342 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
14343 * config/nds32/constraints.md: New constraints and checking for hard
14344 float configuration.
14345 * config/nds32/iterators.md: New mode iterator and attribute for hard
14346 float configuration.
14347 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
14349 * config/nds32/nds32-fpu.md: New file.
14350 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
14351 deal with hard float code generation.
14352 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
14354 (abi_type, float_reg_number): New enum type.
14355 * config/nds32/nds32-predicates.c: New predicates for hard float.
14356 * config/nds32/nds32-protos.h: Declare functions for hard float.
14357 * config/nds32/nds32.c: Implementation for hard float configuration.
14358 * config/nds32/nds32.h: Definitions for hard float configuration.
14359 * config/nds32/nds32.md: Include hard float machine description and
14360 modify patterns for hard float configuration.
14361 * config/nds32/nds32.opt: New options for hard float configuration.
14362 * config/nds32/predicates.md: New predicates for hard float
14365 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14367 * common/config/nds32/nds32-common.c
14368 (nds32_option_optimization_table): Enable -mreleax-hint by default.
14370 2018-04-05 Jakub Jelinek <jakub@redhat.com>
14372 PR middle-end/85195
14373 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
14374 CONSTRUCTOR_ELT (ctor, ...)->value.
14376 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
14379 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
14381 2018-04-05 Tom de Vries <tom@codesourcery.com>
14384 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
14387 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14388 Kito Cheng <kito.cheng@gmail.com>
14390 * config/nds32/constraints.md (U33): Fine-tune checking condition.
14391 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
14392 * config/nds32/nds32.h (nds32_16bit_address_type): Add
14393 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
14395 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14396 Kito Cheng <kito.cheng@gmail.com>
14398 * config/nds32/constraints.md (Ufe): New memory constraint.
14399 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
14400 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
14401 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
14403 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
14404 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
14406 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14408 * config/nds32/nds32.md: Use optimize_size in the condition for
14409 alu-shift instructions.
14411 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14413 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
14415 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14417 * config/nds32/nds32.md (negsi2): Refine pattern.
14419 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14420 Chung-Ju Wu <jasonwucj@gmail.com>
14422 * config/nds32/iterators.md (shift_rotate): New code iterator.
14423 (shift): New code attribute.
14424 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
14425 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
14426 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
14427 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
14428 bit-wise operations.
14429 (andsi3, *andsi3): Ditto.
14430 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
14431 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
14432 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
14433 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
14434 nds32_ior_operand, nds32_xor_operand): New predicates.
14436 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14438 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
14439 (addsi3, subsi3): ... this.
14441 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14443 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
14445 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14447 * config/nds32/nds32.md: Adjust indention.
14449 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14451 * config/nds32/nds32.md (feature): New attribute.
14453 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14455 * config/nds32/nds32.md (subtype): New attribute.
14457 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
14460 * config/arm/arm-builtins.c (arm_expand_builtin): Change
14461 expansion to perform a bitwise AND of the argument followed by a
14462 boolean negation of the result.
14464 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
14466 PR rtl-optimization/84878
14467 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
14468 the basic block. Assert the use reference is not artificial and that
14469 it has an associated insn.
14471 2018-04-04 Michael Matz <matz@suse.de>
14473 * builtins.c (compute_objsize): Pass correct operand
14474 to array_at_struct_end_p.
14476 2018-04-04 Richard Biener <rguenther@suse.de>
14479 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
14480 from contexts for DINFO_LEVEL_TERSE and below.
14482 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14484 * config/nds32/nds32-doubleword.md (move_<mode>): Require
14485 resiter_operand condition.
14486 * config/nds32/nds32.md (*move<mode>): Ditto.
14488 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14489 Monk Chiang <sh.chiang04@gmail.com>
14491 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
14493 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14495 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
14497 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14498 Kito Cheng <kito.cheng@gmail.com>
14500 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
14501 nds32_cond_code_str, output_cond_branch,
14502 output_cond_branch_compare_zero, nds32_expand_cbranch,
14503 nds32_expand_cstore, nds32_expand_movcc,
14504 nds32_output_cbranchsi4_equality_zero,
14505 nds32_output_cbranchsi4_equality_reg,
14506 nds32_output_cbranchsi4_equality_reg_or_const_int,
14507 nds32_output_cbranchsi4_greater_less_zero: New functions.
14508 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
14509 nds32_expand_cstore, nds32_expand_movcc,
14510 nds32_output_cbranchsi4_equality_zero,
14511 nds32_output_cbranchsi4_equality_reg,
14512 nds32_output_cbranchsi4_equality_reg_or_const_int,
14513 nds32_output_cbranchsi4_greater_less_zero): Declare.
14514 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
14515 nds32_rimm11s_operand): New predicates.
14516 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
14517 * config/nds32/nds32.md: Rewrite all the branch and conditional move
14520 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14522 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
14523 * config/nds32/nds32.md: Ditto.
14524 * config/nds32/pipelines.md: Ditto.
14526 2018-04-04 Richard Biener <rguenther@suse.de>
14528 PR tree-optimization/85168
14529 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
14530 propagating abnormals.
14532 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14534 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
14536 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14537 Kito Cheng <kito.cheng@gmail.com>
14539 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
14540 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
14541 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
14542 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
14543 * config/nds32/nds32.md (sibcall_internal): New.
14544 (sibcall_register): Remove.
14545 (sibcall_immediate): Remove.
14546 (sibcall_value_internal): New.
14547 (sibcall_value_register): Remove.
14548 (sibcall_value_immediate): Remove.
14549 * config/nds32/predicates.md (nds32_general_register_operand): New.
14550 (nds32_call_address_operand): New.
14552 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14554 PR rtl-optimization/85167
14555 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
14556 bb_defs if *split_p, instead preinitialize it to NULL.
14558 PR tree-optimization/85156
14559 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
14560 evaluating the argument multiple times.
14562 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
14564 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
14566 (_mm_cvtpd_ps): Likewise.
14567 (_mm_cvttpd_epi32): Likewise.
14568 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
14569 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
14570 vector, pixel, and bool following altivec.h include.
14572 2018-04-03 Martin Sebor <msebor@redhat.com>
14574 * doc/extend.texi (Common Function Attributes): Clarify.
14575 (const attribute): Likewise.
14576 (pure attribute): Likewise.
14578 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14581 * config/i386/i386.c (ix86_expand_vector_set): Use
14582 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
14584 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
14586 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
14587 instructions when changing rounding bits to preserve precision bits
14588 in the x87 control word.
14590 2018-04-03 Martin Liska <mliska@suse.cz>
14592 PR tree-optimization/82491
14593 * rtl.h (strip_offset_and_add): Replace += suboffset with
14594 poly_uint64 () + suboffset.
14596 2018-03-29 Martin Liska <mliska@suse.cz>
14597 Martin Jambor <mjambor@suse.cz>
14600 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
14601 param_type is not an integral or pointer type.
14603 2018-04-03 Richard Biener <rguenther@suse.de>
14605 * sese.h (recompute_all_dominators): Remove.
14607 2018-04-02 Martin Sebor <msebor@redhat.com>
14609 * doc/invoke.texi (-Wrestrict): Fix typos.
14611 2018-04-02 Jim Wilson <jimw@sifive.com>
14613 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
14614 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
14615 (<optab>di3, <optab>si3_extend): Likewise.
14616 (<optab>si3_mask, <optab>si3_mask_1): New.
14617 (<optab>di3_mask, <optab>di3_mask_1): New.
14618 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
14619 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
14620 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
14622 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
14624 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
14627 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
14629 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
14630 (nds32_canonicalize_comparison): New function.
14632 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14633 Kito Cheng <kito.cheng@gmail.com>
14634 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14636 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
14637 * config/nds32/constants.md (unspec_volatile_element): Add
14638 UNSPEC_VOLATILE_RELAX_GROUP.
14639 * config/nds32/nds32-relax-opt.c: New file.
14640 * config/nds32/nds32-predicates.c
14641 (nds32_symbol_load_store_p): New function.
14642 * config/nds32/nds32-protos.h
14643 (nds32_symbol_load_store_p): Declare function.
14644 (make_pass_nds32_relax_opt): Declare new rtl pass function.
14645 * config/nds32/nds32.c
14646 (nds32_register_pass): New function to register pass.
14647 (nds32_register_passes): New function to register passes.
14648 * config/nds32/nds32.md (relax_group): New pattern.
14649 * config/nds32/nds32.opt (mrelax-hint): New option.
14650 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
14652 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
14654 * config/nds32/t-nds32: Modify files dependency.
14656 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14658 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
14659 (PROFILE_HOOK): Define its implementation.
14661 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14663 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
14664 type and 32-bit size.
14666 2018-04-01 Jakub Jelinek <jakub@redhat.com>
14668 PR middle-end/85090
14669 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
14670 (V_128_256): New mode iterator.
14671 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
14672 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
14673 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
14675 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
14676 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
14678 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
14681 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
14682 NaN inputs correctly.
14684 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
14687 * config/rs6000/vsx.md (??r): New mode attribute.
14688 (*vsx_mov<mode>_64bit): Use it.
14689 (*vsx_mov<mode>_32bit): Likewise.
14691 2018-03-30 Martin Sebor <msebor@redhat.com>
14693 PR tree-optimization/84818
14694 * builtins.c (check_access): Use warning_n.
14696 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14699 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
14701 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
14704 2018-03-30 Julia Koval <julia.koval@intel.com>
14707 * x86-tune.def (movx, partial_reg_dependency): Enable for
14710 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
14712 PR inline-asm/84985
14713 * lra-constraints.c (process_alt_operands): Move setting
14714 this_alternative_matches below.
14716 2018-03-29 Martin Liska <mliska@suse.cz>
14719 * doc/invoke.texi: Document how LTO works with debug info.
14720 Describe auto-load support of binutils. Mention 'x86-64'
14721 as valid option value of -march option.
14723 2018-03-29 Jakub Jelinek <jakub@redhat.com>
14725 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
14728 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
14729 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
14730 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
14733 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14736 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
14737 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
14738 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
14739 for RS6000_BTM_POWERPC64.
14740 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
14741 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
14742 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
14745 (DIVDEU): Likewise.
14747 2018-03-28 Carl Love <cel@us.ibm.com>
14750 2017-09-27 Carl Love <cel@us.ibm.com>
14752 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14753 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14754 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14757 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14759 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14760 instead of __vector bool.
14761 (_mm_max_pu8): Likewise.
14762 (_mm_min_pi16): Likewise.
14764 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14767 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
14768 (DIVWEUO): Likewise.
14769 (DIVDEO): Likewise.
14770 (DIVDEUO): Likewise.
14771 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
14772 DIVWEUO and DIVDEUO.
14773 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
14774 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
14775 (div_extend): Likewise.
14776 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
14778 (__builtin_divweuo): Likewise.
14779 (__builtin_divdeo): Likewise.
14780 (__builtin_divdeuo): Likewise.
14782 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14785 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
14786 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
14788 PR tree-optimization/82004
14789 * gimple-match-head.c (optimize_pow_to_exp): New function.
14790 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
14791 Don't fold to exp if optimize_pow_to_exp is false.
14793 2018-03-28 Martin Liska <mliska@suse.cz>
14796 * calls.c (initialize_argument_information): Fix trailing space.
14797 * common.opt: Fix typo and provide better explanation for
14798 -fsanitize-coverage option.
14799 * config/i386/i386.opt: Fix typo.
14801 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14802 Martin Liska <mliska@suse.cz>
14805 * gimplify.c (asan_poison_variable): Don't do the check for
14806 gimplify_omp_ctxp here.
14807 (gimplify_decl_expr): Do it here.
14808 (gimplify_target_expr): Likewise.
14810 2018-03-28 Martin Liska <mliska@suse.cz>
14813 * config/i386/i386.c (ix86_function_arg_advance): Do not call
14814 chkp_type_bounds_count if MPX is not enabled.
14816 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
14818 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
14820 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
14823 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
14824 function to create the function decl for complex long double
14825 multiply and divide for -mabi=ieeelongdouble.
14826 (init_float128_ieee): Call it.
14828 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
14831 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
14832 -fcf-protection=branch -mibt.
14833 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
14835 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14838 * config/arm/arm.c (arm_valid_symbolic_address): Handle
14839 arm_word_relocations.
14841 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
14844 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
14845 extern array declarations.
14847 2018-03-27 Richard Biener <rguenther@suse.de>
14849 PR middle-end/84067
14850 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
14851 explicit single_use checks.
14853 2018-03-27 Richard Biener <rguenther@suse.de>
14855 PR tree-optimization/85082
14856 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
14859 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14861 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
14862 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
14863 Turn on fasynchronous-unwind-tables and funwind-tables.
14865 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
14868 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
14869 (*bmi_blsr_<mode>_ccz): Ditto.
14871 2018-03-26 Tom de Vries <tom@codesourcery.com>
14873 PR tree-optimization/85063
14874 * omp-general.c (offloading_function_p): New function. Factor out
14876 * omp-offload.c (pass_omp_target_link::gate): ... here.
14877 * omp-general.h (offloading_function_p): Declare.
14878 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
14879 with attribute omp declare target for offloading functions.
14881 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
14883 PR tree-optimization/84005
14884 * tree-data-ref.h (get_base_for_alignment): Declare.
14885 * tree-data-ref.c (get_base_for_alignment_1): New function.
14886 (get_base_for_alignment): Likewise.
14887 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
14888 get_base_for_alignment to find a suitable base object, instead
14889 of always using drb->base_address.
14891 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14893 PR inline-asm/85022
14894 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
14895 known size by default.
14897 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
14899 PR inline-asm/85030
14900 * lra-constraints.c (process_alt_operands): Don't match BLKmode
14901 and non BLKmode operands.
14903 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14906 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
14907 Clean up attributes.
14909 2018-03-23 Richard Biener <rguenther@suse.de>
14912 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
14913 we are going to emit early debug for LTO.
14915 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14917 PR inline-asm/85034
14918 * function.c (match_asm_constraints_1): Don't optimize if input
14919 doesn't satisfy general_operand predicate for output's mode.
14921 PR inline-asm/85022
14922 * alias.c (write_dependence_p): Don't require for x_canonicalized
14923 non-VOIDmode if x has VOIDmode.
14926 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
14927 just don't try to optimize it rather than assert it never happens.
14929 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14931 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
14932 macro expansions for definition of ST_INTERNAL_<mode> and
14933 LD_INTERNAL_<mode> builtins.
14934 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
14936 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
14938 (altivec_expand_st_builtin): Likewise.
14939 (altivec_expand_builtin): Remove calls to deleted functions.
14940 (rs6000_address_for_altivec): Delete this function.
14941 * config/rs6000/vector.md: Remove expands for
14942 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
14944 2018-03-22 Sudakshina Das <sudi.das@arm.com>
14947 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
14948 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
14949 re-computing once computed.
14950 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
14951 (arm_init_machine_status): Initialize
14952 machine->static_chain_stack_bytes.
14954 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14957 * doc/extend.texi: Add four new prototypes for vec_ld.
14958 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14959 definitions for more logical presentation.
14960 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14961 entries for V1TI variants of __builtin_altivec_ld builtin.
14962 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14963 handling of V1TI variant of LVX icode pattern.
14964 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14965 (rs6000_gimple_fold_builtin): Likewise.
14966 (altivec_init_builtins): Add code to define
14967 __builtin_altivec_lvx_v1ti function.
14969 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14971 PR inline-asm/84941
14972 * function.c (match_asm_constraints_1): Don't do the optimization
14973 if input isn't a REG, SUBREG, MEM or constant.
14975 2018-03-22 Tom de Vries <tom@codesourcery.com>
14977 PR tree-optimization/84956
14978 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14979 bb_has_abnormal_pred.
14981 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14984 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14985 DECL_INITIAL (decl) to decl at the end.
14986 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14987 adjust the comment.
14989 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14991 * doc/extend.texi (__builtin_tgmath): Document when complex
14992 integer types are treated as _Complex _Float64.
14994 2018-03-21 Tom de Vries <tom@codesourcery.com>
14996 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14998 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15000 PR tree-optimization/84960
15001 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15002 if it is ENTRY block, move them into single succ of ENTRY in that case.
15004 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15006 PR tree-optimization/84811
15007 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15008 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15009 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15010 (known_subrange_p): Update accordingly. Cast each value involved
15011 in the size comparison, rather than casting the result of the
15014 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15016 PR tree-optimization/84982
15017 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15018 by flipping the least significant bit rather than all bits from
15019 bitpos to bitpos + bitsize - 1.
15021 2018-03-21 Nathan Sidwell <nathan@acm.org>
15023 * doc/extend.texi (Deprecated Features): Remove mention of
15024 long-deleted deprecations.
15026 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15029 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15030 * configure: Regenerate.
15032 2018-03-21 Tom de Vries <tom@codesourcery.com>
15034 PR tree-optimization/83126
15035 * tree-parloops.c (num_phis): New function.
15036 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15038 2018-03-21 Nathan Sidwell <nathan@acm.org>
15040 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15041 mention anon-struct/union members and trailing attributes.
15043 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15045 PR tree-optimization/84969
15046 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15047 builtin memset partitions if they set different rhs values.
15049 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15051 PR rtl-optimization/84989
15052 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15053 VEC_DUPLICATE with scalar result mode.
15055 2018-03-21 Martin Liska <mliska@suse.cz>
15058 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15059 not intended return statement.
15061 2018-03-21 Martin Liska <mliska@suse.cz>
15064 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15065 (chkp_find_bound_slots_1): Limit number of iterations.
15067 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15070 * Minor grammar fixes for x86 options.
15072 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15075 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15076 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15078 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15081 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15082 (altivec_lvx_<mode>_1op): Likewise.
15083 (altivec_stvx_<mode>_2op): Likewise.
15084 (altivec_stvx_<mode>_1op): Likewise.
15085 (altivec_lvx_<VM2:mode>): New define_expand.
15086 (altivec_stvx_<VM2:mode>): Likewise.
15087 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15088 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15089 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15090 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15091 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15092 (rs6000_gen_lvx): Likewise.
15093 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15094 (altivec_expand_stv_builtin): Likewise.
15095 (altivec_expand_builtin): Likewise.
15096 * config/rs6000/vector.md: Likewise.
15098 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15101 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15104 2018-03-20 Richard Biener <rguenther@suse.de>
15107 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15108 sign-conversions as zero, fall back to standard scalar_stmt
15111 2018-03-20 Martin Liska <mliska@suse.cz>
15114 * predict.c (rebuild_frequencies): Handle case when we have
15115 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15117 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15120 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15121 flag_section_anchors.
15122 * varasm.c (use_blocks_for_decl_p): Remove hack for
15123 dw2_force_const_mem.
15126 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15128 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15129 be created, use lowpart_subreg of operands[0] rather than operands[0]
15131 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15132 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15133 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15134 and n constraint instead of aarch64_shift_imm_di and Usd.
15135 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15136 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15138 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15141 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15142 to favor GPR over NEON registers.
15143 (<shift>di3_neon): Likewise.
15145 2018-03-20 Tom de Vries <tom@codesourcery.com>
15148 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15149 (nvptx_process_pars): Emit bar.sync asap and alap.
15151 2018-03-20 Tom de Vries <tom@codesourcery.com>
15154 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15155 seen_label if seen_label is already set.
15157 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15160 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15161 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15162 Use 1U instead of 1. Formatting fixes.
15165 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15166 instead of TREE_TYPE (s1) for the return value.
15168 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15170 PR tree-optimization/84946
15171 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15172 bitsize + bitsize in poly_uint64 rather than poly_int64.
15175 * dwarf2asm.c: Include fold-const.c.
15176 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15177 of decl rather than decl itself.
15179 PR rtl-optimization/84643
15180 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15182 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15185 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15186 calling assemble_variable.
15188 2018-03-19 Sudakshina Das <sudi.das@arm.com>
15191 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
15192 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
15194 2018-03-19 Jim Wilson <jimw@sifive.com>
15197 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
15198 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
15199 (riscv_first_stack_step): Likewise.
15200 (riscv_option_override): Use STACK_BOUNDARY instead of
15201 MIN_STACK_BOUNDARY.
15202 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
15203 MIN_STACK_BOUNDARY.
15204 (BIGGEST_ALIGNMENT): Set to 128.
15205 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
15206 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
15209 2018-03-19 Richard Biener <rguenther@suse.de>
15211 PR tree-optimization/84933
15212 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
15213 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
15215 2018-03-19 Richard Biener <rguenther@suse.de>
15217 PR tree-optimization/84859
15218 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
15219 (cond_if_else_store_replacement): Perform sinking operation on
15220 single-store BBs regardless of MAX_STORES_TO_SINK setting.
15221 Generalize what a BB with a single eligible store is.
15223 2018-03-19 Richard Biener <rguenther@suse.de>
15225 PR tree-optimization/84929
15226 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
15227 chrec_is_positive against non-chrec arg.
15229 2018-03-19 Tamar Christina <tamar.christina@arm.com>
15232 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
15234 2018-03-18 Martin Liska <mliska@suse.cz>
15236 PR rtl-optimization/84635
15237 * regrename.c (build_def_use): Use matches_mode only when
15240 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
15242 PR tree-optimization/84913
15243 * tree-vect-loop.c (vectorizable_reduction): Don't try to
15244 vectorize chains of COND_EXPRs.
15246 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15248 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
15250 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15252 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
15254 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15256 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
15258 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15259 Kito Cheng <kito.cheng@gmail.com>
15261 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
15262 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
15263 (nds32_adjust_reg_alloc_order): New function.
15264 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
15266 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15268 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
15269 nds32_print_operand, nds32_print_operand_address): Use
15270 HOST_WIDE_INT_PRINT_DEC instead.
15272 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15274 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
15276 2018-03-17 Jakub Jelinek <jakub@redhat.com>
15279 * config/i386/i386.c (initial_ix86_tune_features,
15280 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
15281 unsigned long long.
15282 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
15283 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
15284 rather than 1u << ix86_tune. Formatting fix.
15285 (ix86_option_override_internal): Change ix86_arch_mask from
15286 unsigned int to unsigned HOST_WIDE_INT, initialize to
15287 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
15288 (ix86_function_specific_restore): Likewise.
15290 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15293 * postreload.c (reload_combine_recognize_pattern): Perform
15294 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
15295 truncate_int_for_mode the result for the destination's mode.
15298 * hsa-gen.c (mem_type_for_type): Fix comment typo.
15299 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
15301 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15304 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
15307 * lra-assigns.c (lra_split_hard_reg_for): Don't use
15308 regno_allocno_class_array and sorted_pseudos.
15309 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
15310 insns where regno is used.
15312 2018-03-16 Martin Liska <mliska@suse.cz>
15315 * multiple_target.c (create_dispatcher_calls): Redirect
15316 reference in the symbol table.
15318 2018-03-16 Martin Liska <mliska@suse.cz>
15321 * multiple_target.c (create_dispatcher_calls): Redirect also
15324 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15328 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
15329 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
15330 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
15332 2018-03-16 Julia Koval <julia.koval@intel.com>
15334 * doc/invoke.texi (Skylake Server): Add CLWB.
15335 Cannonlake): Remove CLWB.
15337 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15339 PR tree-optimization/84841
15340 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
15342 (FLOAT_ONE_CONST_TYPE): Define.
15343 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
15344 (sort_by_operand_rank): Put entries with higher constant_type last
15345 rather than first to match comments.
15347 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
15349 * config/nios2/nios2.md (movsi_internal): Fix thinko in
15352 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15355 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
15356 check and use address of target always.
15358 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
15361 * config/i386/i386.c (indirect_thunk_needed): Update comments.
15362 (indirect_thunk_bnd_needed): Likewise.
15363 (indirect_thunks_used): Likewise.
15364 (indirect_thunks_bnd_used): Likewise.
15365 (indirect_return_needed): New.
15366 (indirect_return_bnd_needed): Likewise.
15367 (output_indirect_thunk_function): Add a bool argument for
15369 (output_indirect_thunk_function): Don't generate alias for
15370 function return thunk.
15371 (ix86_code_end): Call output_indirect_thunk_function to generate
15372 function return thunks.
15373 (ix86_output_function_return): Set indirect_return_bnd_needed
15374 and indirect_return_needed instead of indirect_thunk_bnd_needed
15375 and indirect_thunk_needed.
15377 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
15379 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
15380 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
15381 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
15383 2018-03-15 David Malcolm <dmalcolm@redhat.com>
15384 Paul Hua <paul.hua.gm@gmail.com>
15387 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
15389 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
15391 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
15392 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
15393 resp. SFmode cases.
15395 2018-03-15 Tamar Christina <tamar.christina@arm.com>
15398 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
15399 instead of GET_MODE_SIZE when comparing Units.
15401 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
15404 * varasm.c (hash_section): Return an unchangeble hash value
15405 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
15406 Return !aarch64_can_use_per_function_literal_pools_p ().
15408 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15411 * optabs.c (emit_conditional_move): Pass address of cmode's copy
15412 rather than address of cmode as last argument to prepare_cmp_insn.
15414 2018-03-15 Julia Koval <julia.koval@intel.com>
15416 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
15417 F_AVX512VNNI, F_AVX512BITALG): New.
15419 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
15422 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
15423 insn for floating-point loads and stores.
15425 2018-03-14 Carl Love <cel@us.ibm.com>
15427 * config/rs6000/rs6000-c.c: Add macro definitions for
15428 ALTIVEC_BUILTIN_VEC_PERMXOR.
15429 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
15430 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
15431 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
15432 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
15434 * config/doc/extend.texi: Add prototypes for vec_permxor.
15436 2018-03-14 David Malcolm <dmalcolm@redhat.com>
15439 * diagnostic-show-locus.c (class layout_point): Convert m_line
15440 from int to linenum_type.
15441 (line_span::comparator): Use linenum "compare" function when
15442 comparing line numbers.
15443 (test_line_span): New function.
15444 (layout_range::contains_point): Convert param "row" from int to
15446 (layout_range::intersects_line_p): Likewise.
15447 (layout::will_show_line_p): Likewise.
15448 (layout::print_source_line): Likewise.
15449 (layout::should_print_annotation_line_p): Likewise.
15450 (layout::print_annotation_line): Likewise.
15451 (layout::print_leading_fixits): Likewise.
15452 (layout::annotation_line_showed_range_p): Likewise.
15453 (struct line_corrections): Likewise for field m_row.
15454 (line_corrections::line_corrections): Likewise for param "row".
15455 (layout::print_trailing_fixits): Likewise.
15456 (layout::get_state_at_point): Likewise.
15457 (layout::get_x_bound_for_row): Likewise.
15458 (layout::print_line): Likewise.
15459 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
15460 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
15461 * input.c (selftest::test_linenum_comparisons): New function.
15462 (selftest::input_c_tests): Call it.
15463 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
15464 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
15465 * selftest.h (ASSERT_GT): New macro.
15466 (ASSERT_GT_AT): New macro.
15467 (ASSERT_LT): New macro.
15468 (ASSERT_LT_AT): New macro.
15470 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
15472 PR rtl-optimization/84780
15473 * combine.c (distribute_links): Don't make a link based on pc_rtx.
15475 2018-03-14 Martin Liska <mliska@suse.cz>
15477 * tree.c (record_node_allocation_statistics): Use
15478 get_stats_node_kind.
15479 (get_stats_node_kind): New function extracted from
15480 record_node_allocation_statistics.
15481 (free_node): Use get_stats_node_kind.
15483 2018-03-14 Richard Biener <rguenther@suse.de>
15485 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
15486 that the value-set of ANTIC_IN doesn't grow.
15489 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15491 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15492 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15494 2018-03-14 Julia Koval <julia.koval@intel.com>
15496 * config.gcc (icelake-client, icelake-server): New.
15498 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
15499 (initial_ix86_arch_features): Ditto.
15500 (PTA_SKYLAKE): Add SGX.
15501 (PTA_ICELAKE): Remove.
15502 (PTA_ICELAKE_CLIENT): New.
15503 (PTA_ICELAKE_SERVER): New.
15504 (ix86_option_override_internal): Split up icelake on icelake client and
15506 (get_builtin_code_for_version): Ditto.
15507 (fold_builtin_cpu): Ditto.
15508 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
15509 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
15510 * config/i386/i386.h (processor_type): Ditto.
15511 * doc/invoke.texi: Ditto.
15513 2018-03-14 Jakub Jelinek <jakub@redhat.com>
15516 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
15517 INTEGER_CST offset, add it together with bitpos / 8 and
15518 sign extend based on POINTER_SIZE.
15522 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
15525 * config/i386/constraints.md (Yc): New register constraint.
15526 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
15527 Use Yc constraint for alternative 2 of operand 0. Remove
15528 preferred_for_speed attribute.
15530 2018-03-14 Richard Biener <rguenther@suse.de>
15532 PR tree-optimization/84830
15533 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
15534 with the old one to avoid oscillations.
15536 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
15539 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
15541 (assign_by_spills): Return a flag of reload assignment failure.
15542 Do not process the reload assignment failures. Do not spill other
15543 reload pseudos if they has the same reg class. Update n if
15545 (lra_assign): Add a return arg. Set up from the result of
15546 assign_by_spills call.
15547 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15548 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15549 usage_insns if it is not NULL.
15550 (spill_hard_reg_in_range): New function.
15551 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15552 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15553 function prototypes.
15554 (lra_assign): Change prototype.
15555 * lra.c (lra): Add code to deal with fails by splitting hard reg
15558 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
15560 * config/riscv/riscv.opt (mrelax): New option.
15561 * config/riscv/riscv.c (riscv_file_start): Emit ".option
15562 "norelax" when riscv_mrelax is disabled.
15563 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
15565 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15568 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
15569 reassociation for int modes.
15571 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15573 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
15574 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
15576 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
15577 * config/aarch64/aarch64-sve.md
15578 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
15579 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
15580 (*extend<mode><Vwide>2): Rename to...
15581 (aarch64_sve_extend<mode><Vwide>2): ...this.
15582 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
15583 renaming the old pattern to...
15584 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
15586 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
15587 define_expand, renaming the old pattern to...
15588 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
15589 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
15590 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
15591 account when deciding which SVE instruction the optab should use.
15592 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
15594 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15596 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
15597 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
15598 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
15599 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
15600 (tlsdesc_small_<mode>): Turn a define_expand and use
15601 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
15602 (tlsdesc_small_advsimd_<mode>): ...this.
15603 (tlsdesc_small_sve_<mode>): New pattern.
15605 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15607 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
15608 (UNSPEC_UMUL_HIGHPART): New constants.
15609 (MUL_HIGHPART): New int iteraor.
15610 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
15611 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
15613 (*<su>mul<mode>3_highpart): New define_insn.
15615 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15618 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15621 2018-03-13 Martin Liska <mliska@suse.cz>
15624 * (sem_item_optimizer::sem_item_optimizer): Initialize new
15626 (sem_item_optimizer::~sem_item_optimizer): Release it.
15627 (sem_item_optimizer::merge_classes): Register variable aliases.
15628 (sem_item_optimizer::fixup_pt_set): New function.
15629 (sem_item_optimizer::fixup_points_to_sets): Likewise.
15630 * ipa-icf.h: Declare new variables and functions.
15632 2018-03-13 Jakub Jelinek <jakub@redhat.com>
15634 PR middle-end/84834
15635 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
15636 integer_pow2p@2 and test integer_pow2p in condition.
15637 (A < 0 ? C : 0): Similarly for @1.
15639 PR middle-end/84831
15640 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
15641 characters starting at p contain '\0' character, don't look beyond
15645 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
15646 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
15649 * reg-stack.c (change_stack): Change update_end var from int to
15650 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
15651 also call set_block_for_insn on the newly added insns and rescan.
15654 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
15655 on the last operand.
15658 * tree.c (stabilize_reference_1): Return save_expr (e) for
15659 STATEMENT_LIST even if it doesn't have side-effects.
15661 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
15663 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
15665 2018-03-12 Renlin Li <renlin.li@arm.com>
15667 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
15668 aarch64_output_scalar_simd_mov_immediate.
15670 2018-03-12 Martin Sebor <msebor@redhat.com>
15672 PR tree-optimization/83456
15673 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
15674 for perfectly overlapping calls to memcpy.
15675 (gimple_fold_builtin_memory_chk): Same.
15676 (gimple_fold_builtin_strcpy): Handle no-warning.
15677 (gimple_fold_builtin_stxcpy_chk): Same.
15678 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
15680 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15682 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
15683 parameter. Use it for SFmode.
15684 (rs6000_function_arg_advance_1): Adjust.
15685 (rs6000_function_arg): Adjust.
15686 (rs6000_gimplify_va_arg): Pass false for that new parameter.
15688 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15690 PR rtl-optimization/84169
15691 PR rtl-optimization/84780
15692 * combine.c (can_combine_p): Check for a 2-insn combination whether
15693 the destination register is used between the two insns, too.
15695 2018-03-12 Richard Biener <rguenther@suse.de>
15697 PR tree-optimization/84803
15698 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
15699 for refs DR analysis didn't process.
15701 2018-03-12 Richard Biener <rguenther@suse.de>
15703 PR tree-optimization/84777
15704 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
15705 force-vectorize loops ignore whether we are optimizing for size.
15707 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
15709 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
15710 (TARGET_MD_ASM_ADJUST): Define.
15712 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
15713 Kito Cheng <kito.cheng@gmail.com>
15714 Chung-Ju Wu <jasonwucj@gmail.com>
15716 * config/nds32/nds32.c (nds32_compute_stack_frame,
15717 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
15718 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
15719 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
15720 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
15721 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
15722 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
15723 * config/nds32/nds32.md (prologue, epilogue): Use macro
15724 NDS32_V3PUSH_AVAILABLE_P to do checking.
15726 2018-03-11 Jakub Jelinek <jakub@redhat.com>
15729 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
15730 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
15731 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
15732 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
15733 addition of most attributes on !orig_type_die or the attribute not
15734 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
15736 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15737 Chung-Ju Wu <jasonwucj@gmail.com>
15739 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
15740 __NDS32_VH__ macro.
15741 * config/nds32/nds32.opt (mvh): New option.
15743 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15744 Chung-Ju Wu <jasonwucj@gmail.com>
15746 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
15748 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
15749 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
15752 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15753 Chung-Ju Wu <jasonwucj@gmail.com>
15755 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
15757 * config/nds32/nds32-multiple.md (strlensi): New pattern.
15758 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
15760 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
15761 Kito Cheng <kito.cheng@gmail.com>
15762 Chung-Ju Wu <jasonwucj@gmail.com>
15764 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
15765 UNSPEC_FFMISM and UNSPEC_FLMISM.
15766 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
15767 for ffb, ffmism and flmism.
15768 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
15769 (unspec_ffmism): Ditto.
15770 (unspec_flmism): Ditto.
15771 (nds32_expand_builtin_impl): Check if string extension is available.
15772 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
15773 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
15775 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
15778 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15781 * lra-assigns.c (assign_by_spills): Return a flag of reload
15782 assignment failure. Do not process the reload assignment
15783 failures. Do not spill other reload pseudos if they has the same
15785 (lra_assign): Add a return arg. Set up from the result of
15786 assign_by_spills call.
15787 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15788 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15789 usage_insns if it is not NULL.
15790 (spill_hard_reg_in_range): New function.
15791 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15792 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15793 function prototypes.
15794 (lra_assign): Change prototype.
15795 * lra.c (lra): Add code to deal with fails by splitting hard reg
15798 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
15801 * config/i386/i386.opt: Replace Enforcment with Enforcement.
15803 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
15806 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
15807 (dw_val_node): Add val_symbolic_view.
15808 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
15809 (symview_upper_bound): New.
15810 (new_line_info_table): Initialize symviews_since_reset.
15811 (dwarf2out_source_line): Count symviews_since_reset and set
15812 symview_upper_bound.
15813 (dw_val_equal_p): Handle symview.
15814 (add_AT_symview): New.
15815 (print_dw_val): Handle symview.
15816 (attr_checksum, attr_checksum_ordered): Likewise.
15817 (same_dw_val_p, size_of_die): Likewise.
15818 (value_format, output_die): Likewise.
15819 (add_high_low_attributes): Use add_AT_symview for entry_view.
15820 (dwarf2out_finish): Reset symview_upper_bound, clear
15823 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
15826 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
15827 Add strict argument and use it.
15828 (rs6000_split_multireg_move): Update for new strict argument.
15829 (mem_operand_gpr): Disallow all non-offsettable addresses.
15830 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
15832 2018-03-09 Jakub Jelinek <jakub@redhat.com>
15835 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
15836 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
15837 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
15840 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
15841 decl, use remap_type if we want to use the type.
15843 2018-03-09 Martin Sebor <msebor@redhat.com>
15845 PR tree-optimization/84526
15846 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15848 (builtin_access::generic_overlap): Be prepared to handle non-array
15851 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
15853 PR rtl-optimization/84682
15854 * lra-constraints.c (process_address_1): Check is_address flag
15855 for address constraints.
15856 (process_alt_operands): Likewise.
15857 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
15858 preprocess_constraints.
15859 * recog.h (preprocess_constraints): Add oploc parameter.
15861 * recog.c (preprocess_constraints): Test address_operand for
15862 CT_ADDRESS constraints.
15864 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15867 * lra-assigns.c (assign_by_spills): Return a flag of reload
15868 assignment failure. Do not process the reload assignment
15869 failures. Do not spill other reload pseudos if they has the same
15871 (lra_assign): Add a return arg. Set up from the result of
15872 assign_by_spills call.
15873 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15874 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15875 usage_insns if it is not NULL.
15876 (spill_hard_reg_in_range): New function.
15877 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15878 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15879 function prototypes.
15880 (lra_assign): Change prototype.
15881 * lra.c (lra): Add code to deal with fails by splitting hard reg
15884 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15887 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
15888 Accept complain bool parameter. Only emit errors if it is true.
15889 (arm_parse_cpu_option_name): Likewise.
15890 (arm_target_thumb_only): Adjust callers of the above.
15891 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
15892 prototype to take a default true bool parameter.
15893 (arm_parse_arch_option_name): Likewise.
15895 2018-03-09 David Malcolm <dmalcolm@redhat.com>
15896 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
15900 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
15901 * configure: Regenerate.
15902 * configure.ac ("linker --version-script option"): New.
15903 ("linker soname option"): New.
15905 2018-03-09 Richard Biener <rguenther@suse.de>
15907 PR tree-optimization/84775
15908 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
15909 immediate uses of predicate stmts and mark them modified.
15912 PR tree-optimization/84178
15913 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15915 (version_loop_for_if_conversion): Delay update_ssa call.
15916 (tree_if_conversion): Delay update_ssa until after predicate
15919 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
15922 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
15923 when the function accesses prior frames.
15925 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15928 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
15929 gen_llsym, otherwise call maybe_gen_llsym.
15931 PR inline-asm/84742
15932 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15933 has ',' character inside of it.
15935 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15938 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15939 as clobbering CC_REGNUM.
15941 2018-03-08 Richard Biener <rguenther@suse.de>
15943 PR middle-end/84552
15944 * tree-scalar-evolution.c: Include tree-into-ssa.h.
15945 (follow_copies_to_constant): Do not follow SSA names registered
15948 2018-03-08 Richard Biener <rguenther@suse.de>
15950 PR tree-optimization/84178
15951 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15953 (version_loop_for_if_conversion): Delay update_ssa call.
15954 (tree_if_conversion): Delay update_ssa until after predicate
15957 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15959 PR tree-optimization/84178
15960 * tree-if-conv.c (release_bb_predicate): Remove the
15961 the assertion that the stmts have NULL use_ops.
15962 Discard the statements, asserting that they haven't
15963 yet been added to a BB.
15965 2018-03-08 Richard Biener <rguenther@suse.de>
15967 PR tree-optimization/84746
15968 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15969 (phi_translate): Pass in destination ANTIC_OUT set.
15970 (phi_translate_1): Likewise. For a simplified result lookup
15971 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15972 (phi_translate_set): Adjust.
15973 (do_pre_regular_insertion): Likewise.
15974 (do_pre_partial_partial_insertion): Likewise.
15976 2018-03-08 Martin Liska <mliska@suse.cz>
15978 PR gcov-profile/84735
15979 * doc/gcov.texi: Document usage of profile files.
15980 * gcov-io.h: Document changes in the format.
15982 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15986 * dwarf2out.c (struct dw_line_info_table): Update comments for
15988 (FORCE_RESET_NEXT_VIEW): New.
15989 (FORCE_RESETTING_VIEW_P): New.
15990 (RESETTING_VIEW_P): Check for -1 too.
15991 (ZERO_VIEW_P): Likewise.
15992 (new_line_info_table): Force-reset next view.
15993 (dwarf2out_begin_function): Likewise.
15994 (dwarf2out_source_line): Simplify zero_view_p initialization.
15995 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15996 view directly. Omit view when omitting .loc at line 0.
15998 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16000 PR tree-optimization/84740
16001 * tree-switch-conversion.c (process_switch): Call build_constructors
16002 only if info.phi_count is non-zero.
16004 PR tree-optimization/84739
16005 * tree-tailcall.c (find_tail_calls): Check call arguments against
16006 DECL_ARGUMENTS (current_function_decl) rather than
16007 DECL_ARGUMENTS (func) when checking for tail recursion.
16009 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16011 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16012 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16013 Volker Reichelt's entry and add entries for people that perform
16014 GCC fuzzy testing and report numerous bugs.
16016 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16019 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16020 readonly data in sdata, if that is disabled.
16021 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16022 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16023 -mreadonly-in-sdata option.
16025 2018-03-07 Martin Sebor <msebor@redhat.com>
16027 PR tree-optimization/84468
16028 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16029 basic block when looking for nul assignment.
16031 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16034 * except.h (output_function_exception_table): Adjust prototype.
16035 * except.c (output_function_exception_table): Remove FNNAME parameter
16036 and add SECTION parameter. Ouput one part of the table at a time.
16037 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16038 the first part of the exception table and emit unwind directives.
16039 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16040 (i386_pe_seh_cold_init): Likewise.
16041 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16042 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16043 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16044 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16045 * config/i386/winnt.c: Include except.h.
16046 (struct seh_frame_state): Add reg_offset, after_prologue and
16047 in_cold_section fields.
16048 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16049 (i386_pe_seh_cold_init): New function.
16050 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16051 to seh->in_cold_section.
16052 (seh_emit_push): Record the offset of the push.
16053 (seh_emit_save): Record the offet of the save.
16054 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16055 Test seh->after_prologue to disregard the epilogue.
16056 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16057 (i386_pe_end_cold_function): New function.
16059 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16062 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16063 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16066 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16068 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16069 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16071 PR middle-end/84723
16072 * multiple_target.c: Include tree-inline.h and intl.h.
16073 (expand_target_clones): Diagnose and fail if node->definition and
16074 !tree_versionable_function_p (node->decl).
16076 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16078 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16080 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16081 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16082 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16084 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16087 * combine.c (try_combine): Use reg_or_subregno instead of handling
16088 just paradoxical SUBREGs and REGs.
16090 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16092 * config/arc/arc.c (arc_finalize_pic): Remove function.
16093 (arc_must_save_register): We use single base PIC register, remove
16094 checks to save/restore the PIC register.
16095 (arc_expand_prologue): Likewise.
16096 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16098 (arc_verify_short): Likewise.
16099 (arc_attr_type): Likewise.
16100 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16101 (walk_stores): Likewise.
16102 (arc_address_cost): Make it static.
16103 (arc_verify_short): Likewise.
16104 (branch_dest): Likewise.
16105 (arc_attr_type): Likewise.
16106 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16107 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16108 (arc_final_prescan_insn): Remove inserting the nops due to
16109 hardware hazards. It is done in reorg step.
16110 (insn_length_variant_t): Remove.
16111 (insn_length_parameters_t): Likewise.
16112 (arc_insn_length_parameters): Likewise.
16113 (arc_get_insn_variants): Likewise.
16114 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16116 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16118 PR inline-asm/84683
16119 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16122 PR tree-optimization/84687
16123 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16125 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16127 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16129 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16130 Rename to ppc_speculation_barrier.
16131 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16132 __builtin_ppc_speculation_barrier.
16134 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16137 * combine.c (combine_simplify_rtx): Don't try to simplify if
16138 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16141 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16143 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16144 to 32 bytes when compiling for POWER9.
16146 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16149 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16150 regparm >= 3 with no arg reg available also for calls with
16151 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16154 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16156 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16158 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16161 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16163 2018-03-05 Richard Biener <rguenther@suse.de>
16165 PR tree-optimization/84486
16166 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16167 When inserting a __builtin_assume_aligned call set the LHS
16168 SSA name alignment info accordingly.
16170 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16172 PR tree-optimization/84114
16173 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16174 Avoid reassociation of FLOAT_MODE addition.
16176 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16178 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16179 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16180 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16181 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16182 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16183 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16186 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16187 __builtin_ia32_wbinvd): New builtins.
16188 (SPECIAL_ARGS2): New.
16189 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
16190 (SPECIAL_ARGS2): New.
16191 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
16192 (ix86_valid_target_attribute_inner_p): Ditto.
16193 (ix86_init_mmx_sse_builtins): Add special_args2.
16194 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
16195 TARGET_WBNOINVD_P): New.
16196 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
16197 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
16198 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
16199 * config/i386/immintrin.h (_wbinvd): New intrinsic.
16200 * config/i386/pconfigintrin.h: New file.
16201 * config/i386/wbnoinvdintrin.h: Ditto.
16202 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
16204 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
16206 2018-03-05 Richard Biener <rguenther@suse.de>
16208 PR tree-optimization/84670
16209 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16211 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16212 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16213 (compute_antic_aux): Only assert the number of values in ANTIC_IN
16214 doesn't grow if all successors (recursively) were visited at least
16217 2018-03-05 Richard Biener <rguenther@suse.de>
16219 PR tree-optimization/84650
16220 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
16221 if executed in the loop pipeline.
16223 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
16225 * doc/configfiles.texi (Configuration Files): Move info about
16226 conditionalizing $target-protos.h to...
16227 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
16228 differs from $target-protos.h.
16230 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
16231 Chung-Ju Wu <jasonwucj@gmail.com>
16233 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
16234 * config/nds32/nds32-multiple.md (setmemsi): Define.
16235 * config/nds32/nds32-memory-manipulation.c
16236 (nds32_gen_dup_4_byte_to_word_value): New.
16237 (emit_setmem_word_loop): New.
16238 (emit_setmem_byte_loop): New.
16239 (nds32_expand_setmem_loop): New.
16240 (nds32_expand_setmem_loop_v3m): New.
16241 (nds32_expand_setmem_unroll): New.
16242 (nds32_expand_setmem): New.
16244 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16245 Chung-Ju Wu <jasonwucj@gmail.com>
16247 * config/nds32/nds32-memory-manipulation.c
16248 (nds32_emit_load_store): New.
16249 (nds32_emit_post_inc_load_store): New.
16250 (nds32_emit_mem_move): New.
16251 (nds32_emit_mem_move_block): New.
16252 (nds32_expand_movmemsi_loop_unknown_size): New.
16253 (nds32_expand_movmemsi_loop_known_size): New.
16254 (nds32_expand_movmemsi_loop): New.
16255 (nds32_expand_movmemsi_unroll): New.
16256 (nds32_expand_movmemqi): Rename ...
16257 (nds32_expand_movmemsi): ... to this.
16258 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
16259 (movmemsi): ... to this.
16260 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
16261 (nds32_expand_movmemsi): ... to this.
16263 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16264 Monk Chiang <sh.chiang04@gmail.com>
16265 Chung-Ju Wu <jasonwucj@gmail.com>
16267 * config/nds32/nds32-protos.h
16268 (nds32_expand_load_multiple): New arguments.
16269 (nds32_expand_store_multiple): Ditto.
16270 (nds32_valid_multiple_load_store): Rename ...
16271 (nds32_valid_multiple_load_store_p): ... to this.
16272 * config/nds32/nds32-memory-manipulation.c
16273 (nds32_expand_load_multiple): Refine implementation.
16274 (nds32_expand_store_multiple): Ditto.
16275 * config/nds32/nds32-multiple.md
16276 (load_multiple): Update nds32_expand_load_multiple interface.
16277 (store_multiple): Update nds32_expand_store_multiple interface.
16278 * config/nds32/nds32-predicates.c
16279 (nds32_valid_multiple_load_store): Rename ...
16280 (nds32_valid_multiple_load_store_p): ... to this and refine
16282 * config/nds32/predicates.md
16283 (nds32_load_multiple_and_update_address_operation): New predicate.
16284 (nds32_store_multiple_and_update_address_operation): New predicate.
16286 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16287 Chung-Ju Wu <jasonwucj@gmail.com>
16289 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
16290 (combo): New attribute.
16291 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
16293 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
16295 * config/nds32/nds32.opt: Change -mcmodel= default value.
16297 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
16298 Monk Chiang <sh.chiang04@gmail.com>
16299 Chung-Ju Wu <jasonwucj@gmail.com>
16301 * config/nds32/constants.md (unspec_element): New enum.
16302 * config/nds32/constraints.md (Umw): New constraint.
16303 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
16304 * config/nds32/nds32-intrinsic.md: Likewise.
16305 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
16306 (nds32_valid_smw_lwm_base_p): New.
16307 (nds32_output_smw_single_word): New.
16308 (nds32_output_lmw_single_word): New.
16309 (nds32_expand_unaligned_load): New.
16310 (nds32_expand_unaligned_store): New.
16311 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
16312 (nds32_output_smw_single_word): Declare.
16313 (nds32_output_lmw_single_word): Declare.
16314 (nds32_expand_unaligned_load): Declare.
16315 (nds32_expand_unaligned_store): Declare.
16316 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
16317 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
16318 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
16319 NDS32_BUILTIN_UASTORE_DW.
16320 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
16323 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
16324 Kito Cheng <kito.cheng@gmail.com>
16325 Chung-Ju Wu <jasonwucj@gmail.com>
16327 * config/nds32/nds32-intrinsic.c
16328 (nds32_expand_builtin_null_ftype_reg): Delete.
16329 (nds32_expand_builtin_reg_ftype_imm): Ditto.
16330 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
16331 (nds32_read_argument): New.
16332 (nds32_legitimize_target): Ditto.
16333 (nds32_legitimize_argument): Ditto.
16334 (nds32_check_constant_argument): Ditto.
16335 (nds32_expand_unop_builtin): Ditto.
16336 (nds32_expand_unopimm_builtin): Ditto.
16337 (nds32_expand_binop_builtin): Ditto.
16338 (nds32_builtin_decl_impl): Ditto.
16339 (builtin_description): Ditto.
16340 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
16341 (nds32_init_builtins_impl): Ditto.
16342 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
16343 (nds32_builtin_decl): New.
16344 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
16345 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
16347 2018-03-02 Jeff Law <law@redhat.com>
16349 * reorg.c (stop_search_p): Handle DEBUG_INSN.
16350 (redundant_insn, fill_simple_delay_slots): Likewise.
16351 (fill_slots_from_thread): Likewise.
16352 * resource.c (mark_referenced_resources): Likewise.
16353 (mark_set_resources, find_dead_or_set_registers): Likewise.
16355 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16357 * substring-locations.h (format_warning_va): Formatting fix for
16358 ATTRIBUTE_GCC_DIAG.
16359 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
16361 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
16362 * substring-locations.c: Include intl.h.
16363 (format_warning_va): Turned into small wrapper around
16364 format_warning_n_va, renamed to ...
16365 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
16366 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
16368 (format_warning_at_substring_n): New function.
16369 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
16370 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
16371 format_warning_at_substring with just a shorter name instead of
16372 const function pointer.
16373 (fmtwarn_n): New function.
16374 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
16375 appropriate, get rid of all the fmtstr temporaries, move conditionals
16376 with G_() wrapped string literals directly into fmtwarn arguments,
16377 cast dir.len to (int), formatting fixes.
16379 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
16381 * doc/invoke.texi: Remove "Cilk Plus" references.
16383 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16384 Richard Biener <rguenther@suse.de>
16387 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
16388 for error or warning attributes if CALL_FROM_THUNK_P is set.
16391 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16394 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
16395 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
16398 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
16399 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
16401 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
16402 instead of -1U in last predictors element's probability member.
16404 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16407 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16408 arguments if they are comparable.
16410 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16412 PR tree-optimization/84634
16413 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
16414 masks and masked_loop_p with a single loop_masks, making sure it's
16415 null for bb vectorization.
16417 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16419 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
16420 (vect_analyze_data_ref_access): Use loop->safe_len rather than
16421 loop->force_vectorize to check whether there is no alias.
16423 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16426 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
16428 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
16430 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
16431 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
16432 instead of a loop around prev_real_insn.
16433 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
16436 PR inline-asm/84625
16437 * config/i386/i386.c (ix86_print_operand): Use conditional
16438 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
16441 2018-03-02 Richard Biener <rguenther@suse.de>
16443 PR tree-optimization/84427
16444 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
16445 (bitmap_set_subtract_values): Rewrite to handle multiple
16448 (prune_clobbered_mems): Likewise.
16449 (phi_translate): Take edge instead of pred/phiblock.
16450 (phi_translate_1): Likewise.
16451 (phi_translate_set): Likewise. Insert all translated
16452 exprs for a value into the set, keeping possibly multiple
16453 expressions per value.
16454 (compute_antic_aux): Adjust for phi_translate changes.
16455 When intersecting union the expressions and prune those
16456 not in the final value set, keeping possibly multiple
16457 expressions per value. Do not use value-insertion
16458 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
16459 all expressions. Add verification that the value-sets
16460 only shrink during iteration.
16461 (compute_partial_antic_aux): Adjust for the phi_translate changes.
16462 (do_pre_regular_insertion): Likewise.
16463 (do_pre_partial_partial_insertion): Likewise.
16465 2018-03-02 Richard Biener <rguenther@suse.de>
16468 * config/darwin.c (saved_debug_info_level): New static global.
16469 (darwin_asm_lto_start): Disable debug info generation for LTO out.
16470 (darwin_asm_lto_end): Restore debug info generation settings.
16472 2018-03-01 Martin Liska <mliska@suse.cz>
16475 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
16476 volatile arguments.
16478 2018-03-01 Richard Biener <rguenther@suse.de>
16481 * dwarf2out.c (gen_variable_die): Properly handle late VLA
16482 type annotation with LTO when debug was disabled at compile-time.
16484 2018-03-01 Matthew Fortune <mfortune@gmail.com>
16486 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
16488 (mips_final_postscan_insn): Likewise.
16490 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
16492 PR rtl-optimization/84528
16493 * alias.c (init_alias_target): Add commentary.
16494 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
16495 a unique base value if the frame pointer is not eliminated
16496 to the stack pointer.
16498 2018-03-01 Tom de Vries <tom@codesourcery.com>
16500 PR rtl-optimization/83327
16501 * lra-int.h (hard_regs_spilled_into): Declare.
16502 * lra.c (hard_regs_spilled_into): Define.
16503 (init_reg_info): Init hard_regs_spilled_into.
16504 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
16505 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
16506 (process_bb_lives): Handle hard_regs_spilled_into.
16507 (lra_create_live_ranges_1): Before doing liveness propagation, clear
16508 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
16510 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
16512 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
16513 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
16514 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
16515 * config/rs6000/aix72.h: New file.
16517 2018-02-28 Jakub Jelinek <jakub@redhat.com>
16519 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
16520 instead of warning_at with conditional singular and plural messages
16524 * stor-layout.c (update_alignment_for_field): For
16525 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
16526 && !DECL_PACKED (field), do the alignment update, just use
16527 only desired_align instead of MAX (type_align, desired_align)
16529 (place_field): Don't do known_align < desired_align handling
16530 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
16531 is non-NULL, instead do it after rli->prev_field handling and
16532 only if not within a bitfield word. For DECL_PACKED (field)
16533 use type_align of BITS_PER_UNIT.
16535 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
16537 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
16538 superfluous parentheses and trailing spaces.
16540 2018-02-28 Richard Biener <rguenther@suse.de>
16542 PR tree-optimization/84584
16543 * graphite-scop-detection.c (scop_detection::add_scop): Discard
16544 SCoPs with fake exit edge.
16546 2018-02-28 Martin Liska <mliska@suse.cz>
16549 * timevar.c (timer::print): Fix format to properly print 100%
16552 2018-02-28 Richard Biener <rguenther@suse.de>
16554 PR middle-end/84607
16555 * genmatch.c (capture_info::walk_match): Do not mark
16556 captured expressions without operands as expr_p given
16557 they act more like predicates and should be subject to
16558 "lost tail" side-effect preserving.
16560 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
16562 PR rtl-optimization/81611
16563 * auto-inc-dec.c (attempt_change): Move dead note from
16564 mem_insn if it's the next use of regno
16565 (find_address): Take address use of reg holding
16566 non-incremented value. Add parm to limit search to the named
16568 (merge_in_block): Attempt to use a mem insn that is the next
16569 use of the original regno.
16571 2018-02-27 Martin Sebor <msebor@redhat.com>
16574 * doc/invoke.texi (-Wmissing-attributes): New option.
16575 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
16577 2018-02-27 Martin Sebor <msebor@redhat.com>
16579 PR translation/84207
16580 * diagnostic-core.h (warning_n, error_n, inform_n): Change
16581 n argument to unsigned HOST_WIDE_INT.
16582 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
16583 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
16584 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
16585 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
16587 2018-02-27 Richard Biener <rguenther@suse.de>
16589 PR tree-optimization/84512
16590 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16591 Do not use the estimate returned from record_stmt_cost for
16592 the scalar iteration cost but sum properly using add_stmt_cost.
16594 2018-02-27 Richard Biener <rguenther@suse.de>
16596 PR tree-optimization/84466
16597 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16598 Adjust last change to less strictly validate use operands.
16600 2018-02-27 Martin Liska <mliska@suse.cz>
16602 PR gcov-profile/84548
16603 * gcov.c (process_file): Allow partial overlap and consider it
16604 also as group functions.
16605 (output_lines): Properly calculate range of lines for a group.
16607 2018-02-27 Martin Liska <mliska@suse.cz>
16609 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
16610 'ggc' suffixes. Change first column width.
16611 (timer::print): Fix formatting of the column.
16613 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
16615 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
16616 preserve inline entry blocks for the sake of debug inline
16617 entry point markers alone.
16618 (remove_unused_locals): Suggest in comments a better place to
16619 force the preservation of inline entry blocks that are
16620 otherwise unused, but do not preserve them.
16622 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16624 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
16626 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16629 * config/i386/constraints.md (Bs): Replace
16630 ix86_indirect_branch_register with
16631 TARGET_INDIRECT_BRANCH_REGISTER.
16633 * config/i386/i386.md (indirect_jump): Likewise.
16634 (tablejump): Likewise.
16635 (*sibcall_memory): Likewise.
16636 (*sibcall_value_memory): Likewise.
16637 Peepholes of indirect call and jump via memory: Likewise.
16638 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
16639 (*sibcall_value_GOT_32): Likewise.
16640 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16641 (GOT_memory_operand): Likewise.
16642 (call_insn_operand): Likewise.
16643 (sibcall_insn_operand): Likewise.
16644 (GOT32_symbol_operand): Likewise.
16645 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
16647 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16649 PR rtl-optimization/83496
16650 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
16651 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
16652 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
16653 redundant insn, if any.
16654 (relax_delay_slots): Likewise.
16655 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
16657 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
16659 PR tree-optimization/83965
16660 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
16661 that grouped statements are part of a reduction chain. Return
16662 true if the statement is not marked as a reduction itself but
16663 is part of a group.
16664 (vect_recog_dot_prod_pattern): Don't check whether the statement
16665 is part of a group here.
16666 (vect_recog_sad_pattern): Likewise.
16667 (vect_recog_widen_sum_pattern): Likewise.
16669 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16672 * final.c (rest_of_clean_state): Also look for calls inside sequences.
16674 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16677 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
16679 (ix86_output_indirect_function_return): New prototype.
16680 (ix86_split_simple_return_pop_internal): Likewise.
16681 * config/i386/i386.c (indirect_return_via_cx): New.
16682 (indirect_return_via_cx_bnd): Likewise.
16683 (indirect_thunk_name): Handle return va CX_REG.
16684 (output_indirect_thunk_function): Create alias for
16685 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
16686 (ix86_output_indirect_jmp): Remove the bool argument.
16687 (ix86_output_indirect_function_return): New function.
16688 (ix86_split_simple_return_pop_internal): Likewise.
16689 * config/i386/i386.md (*indirect_jump): Don't pass false
16690 to ix86_output_indirect_jmp.
16691 (*tablejump_1): Likewise.
16692 (simple_return_pop_internal): Change it to define_insn_and_split.
16693 Call ix86_split_simple_return_pop_internal to split it for
16694 -mfunction-return=.
16695 (simple_return_indirect_internal): Call
16696 ix86_output_indirect_function_return instead of
16697 ix86_output_indirect_jmp.
16699 2018-02-26 Jakub Jelinek <jakub@redhat.com>
16702 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
16703 memset and value initialization afterwards.
16705 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
16707 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
16709 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16712 * common/config/aarch64/aarch64-common.c
16713 (aarch_option_optimization_table[]): Switch
16714 off fomit-frame-pointer
16716 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16717 Chung-Ju Wu <jasonwucj@gmail.com>
16719 * config/nds32/nds32-multiple.md (load_multiple): Disallow
16721 (store_multiple): Ditto.
16723 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16725 * config.gcc: Add --with-cpu support for nds32 target.
16726 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
16727 * config/nds32/nds32.opt: Add -mcpu= option.
16729 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
16731 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
16732 isel=yes): Warn for these deprecated options.
16734 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
16736 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
16737 ISA_2_5_MASKS_EMBEDDED.
16739 2018-02-23 Jakub Jelinek <jakub@redhat.com>
16741 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
16742 p->max as pointers rather than using iterative_hash_expr.
16744 2018-02-23 Carl Love <cel@us.ibm.com>
16746 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
16747 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
16749 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
16750 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
16751 P8V_BUILTIN_VEC_VUNSIGNED2.
16753 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
16756 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
16757 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
16759 * lra-constraints.c (curr_insn_transform): Set up
16760 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
16762 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
16763 * lra-eliminations.c (spill_pseudos): Ditto.
16764 (process_insn_for_elimination): Ditto.
16765 * lra-lives.c (reg_early_clobber_p): Use the new macros.
16766 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
16767 LRA_NON_CLOBBERED_ALT.
16769 2018-02-22 Martin Sebor <msebor@redhat.com>
16771 PR tree-optimization/84480
16772 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
16773 to maybe_diag_stxncpy_trunc. Call it.
16774 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
16775 from gimple_fold_builtin_strcpy. Print inlining stack.
16776 (handle_builtin_stxncpy): Print inlining stack.
16777 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
16779 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
16782 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
16783 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
16784 and -fcheck-pointer-bounds are used together.
16785 (indirect_thunk_prefix): New enum.
16786 (indirect_thunk_need_prefix): New function.
16787 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
16788 "_nt" instead of "_bnd" for NOTRACK prefix.
16789 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
16790 (output_indirect_thunk_function): Likewise.
16792 (ix86_code_end): Update output_indirect_thunk_function calls.
16793 (ix86_output_indirect_branch_via_reg): Replace
16794 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
16795 (ix86_output_indirect_branch_via_push): Likewise.
16796 (ix86_output_function_return): Likewise.
16797 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
16798 incompatible with -fcf-protection=branch and
16799 -fcheck-pointer-bounds.
16801 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16804 * config/aarch64/aarch64.c (aarch64_print_address_internal):
16805 Change gcc_assert call to output_operand_lossage.
16807 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16809 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
16811 2018-02-22 DJ Delorie <dj@redhat.com>
16812 Sebastian Perta <sebastian.perta@renesas.com>
16813 Oleg Endo <olegendo@gcc.gnu.org>
16815 * config/rx/rx.c (rx_rtx_costs): New function.
16816 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
16818 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
16820 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
16822 2018-02-22 Martin Liska <mliska@suse.cz>
16825 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
16826 Add "native" as a possible value.
16828 2018-02-22 Martin Liska <mliska@suse.cz>
16831 * config/i386/i386.c (ix86_option_override_internal):
16832 Add "native" as a possible value for -march and -mtune.
16834 2018-02-22 Jakub Jelinek <jakub@redhat.com>
16837 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
16838 to all type variants.
16840 PR tree-optimization/84503
16841 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
16842 width as info->bitpos + info->bitsize - start.
16843 (merged_store_group::merge_overlapping): Simplify width computation.
16844 (check_no_overlap): New function.
16845 (imm_store_chain_info::try_coalesce_bswap): Compute expected
16846 start + width and last_order of the group, fail if check_no_overlap
16848 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
16849 to group if check_no_overlap fails.
16851 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16853 * config/rs6000/altivec.md: Delete contraint arguments to
16854 define_expand, define_split, and define_peephole2, and in
16855 define_insn_and_split if always unused.
16856 * config/rs6000/darwin.md: Ditto.
16857 * config/rs6000/dfp.md: Ditto.
16858 * config/rs6000/rs6000.md: Ditto.
16859 * config/rs6000/sync.md: Ditto.
16860 * config/rs6000/vector.md: Ditto.
16861 * config/rs6000/vsx.md: Ditto.
16863 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16865 * config/rs6000/altivec.md: Write output control strings as braced
16866 blocks instead of double-quoted strings.
16867 * config/rs6000/darwin.md: Ditto.
16868 * config/rs6000/rs6000.md: Ditto.
16869 * config/rs6000/vector.md: Ditto.
16870 * config/rs6000/vsx.md: Ditto.
16872 2018-02-21 Jason Merrill <jason@redhat.com>
16874 PR c++/84314 - ICE with templates and fastcall attribute.
16875 * attribs.c (build_type_attribute_qual_variant): Remove assert.
16877 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16879 * ipa-cp.c (determine_versionability): Fix comment typos.
16881 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16884 * ipa-cp.c (determine_versionability): Do not version functions caling
16887 2018-02-21 Martin Liska <mliska@suse.cz>
16890 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
16891 Add "native" as a possible value.
16892 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
16893 the macro when native cpu detection is available.
16895 2018-02-21 Martin Liska <mliska@suse.cz>
16898 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
16899 Add "native" as a possible value.
16900 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
16901 when native cpu detection is available.
16903 2018-02-21 Jakub Jelinek <jakub@redhat.com>
16904 Martin Sebor <msebor@redhat.com>
16906 PR tree-optimization/84478
16907 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
16909 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
16910 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
16911 support which is conservatively correct, for 2 only stay conservative
16912 for maxlen. Formatting and comment capitalization fixes. Add STRICT
16913 argument to the 2 argument get_range_strlen, adjust 6 arg
16914 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
16916 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
16917 (gimple_fold_builtin_strlen): Pass true as last argument to
16920 2018-02-20 Martin Sebor <msebor@redhat.com>
16922 PR middle-end/84095
16923 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
16924 (builtin_memref::set_base_and_offset): Same. Handle inner references.
16925 (builtin_memref::builtin_memref): Factor out parts into
16926 set_base_and_offset and call it.
16928 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16930 PR middle-end/84406
16931 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
16932 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
16933 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
16934 search at the associated MODE_INT.
16936 2018-02-20 Jeff Law <law@redhat.com>
16938 PR middle-end/82123
16939 PR tree-optimization/81592
16940 PR middle-end/79257
16941 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
16942 for range data rather than using global data.
16943 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
16944 range data rather than using global data.
16945 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
16946 pass it to children as needed.
16947 (struct directive::fmtresult): Similarly.
16948 (struct directive::set_width): Similarly.
16949 (struct directive::set_precision): Similarly.
16950 (format_integer, format_directive, parse_directive): Similarly.
16951 (format_none): Accept unnamed vr_values parameter.
16952 (format_percent, format_floating, format_character): Similarly.
16953 (format_string, format_plain): Similarly.
16954 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16955 the EVRP range analyzer for range data rather than using global data.
16956 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16957 gimple-ssa-evrp-analyze.h
16958 (class sprintf_dom_walker): Add after_dom_children member function.
16959 Add evrp_range_analyzer member.
16960 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16961 range analyzer as needed.
16962 (sprintf_dom_walker::after_dom_children): New member function.
16963 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16965 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16966 (evrp_range_analyzer::pop_to_marker): Likewise.
16968 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16970 PR tree-optimization/84419
16971 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16972 with the required type if its current type is compatible but
16975 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16977 PR middle-end/82004
16978 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16979 after vectorization.
16981 2018-02-20 Martin Liska <mliska@suse.cz>
16984 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16985 possible values if we don't have a hint.
16987 2018-02-20 Martin Liska <mliska@suse.cz>
16991 * final.c (shorten_branches): Build align_tab array with one
16993 * opts.c (finish_options): Add alignment option limit check.
16994 (MAX_CODE_ALIGN): Likewise.
16995 (MAX_CODE_ALIGN_VALUE): Likewise.
16996 * doc/invoke.texi: Document maximum allowed option value for
16997 all -falign-* options.
16999 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17002 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17003 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17004 * var-tracking.c (emit_note_insn_var_location): Remove all references
17005 to NOTE_INSN_CALL_ARG_LOCATION.
17006 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17007 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17008 Use copy_rtx_if_shared.
17009 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17010 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17011 (dwarf2out_var_location): Remove handling of
17012 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17014 * final.c (final_scan_insn): Remove all references to
17015 NOTE_INSN_CALL_ARG_LOCATION.
17016 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17017 before dumping final insns.
17018 * except.c (emit_note_eh_region_end): Remove all references to
17019 NOTE_INSN_CALL_ARG_LOCATION.
17020 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17021 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17022 * config/arc/arc.c (hwloop_optimize): Likewise.
17023 * config/arm/arm.c (create_fix_barrier): Likewise.
17024 * config/s390/s390.c (s390_chunkify_start): Likewise.
17025 * config/sh/sh.c (find_barrier): Likewise.
17026 * config/i386/i386.c (rest_of_insert_endbranch,
17027 ix86_seh_fixup_eh_fallthru): Likewise.
17028 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17029 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17030 * config/frv/frv.c (frv_function_prologue): Likewise.
17031 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17033 (note_outside_basic_block_p): Remove all references to
17034 NOTE_INSN_CALL_ARG_LOCATION.
17035 * gengtype.c (adjust_field_rtx_def): Likewise.
17036 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17038 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17039 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17042 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17045 PR tree-optimization/84452
17046 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17047 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17050 2018-02-19 Martin Liska <mliska@suse.cz>
17053 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17055 2018-02-19 Martin Liska <mliska@suse.cz>
17056 Richard Sandiford <richard.sandiford@linaro.org>
17058 PR tree-optimization/82491
17059 * gimple-fold.c (get_base_constructor): Make earlier bail out
17062 2018-02-19 Carl Love <cel@us.ibm.com>
17064 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17065 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17067 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17068 P8V_BUILTIN_VEC_NEG.
17070 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17072 * config/rl78/rl78.md (movdf): New define expand.
17074 2018-02-19 Martin Liska <mliska@suse.cz>
17077 * doc/invoke.texi: Fix typo.
17078 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17080 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17082 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17083 handle rs6000_single_float and rs6000_double_float specially for
17086 2018-02-16 Jeff Law <law@redhat.com>
17088 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17089 (pop_regs): Use it.
17091 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17094 * ipa-inline.c (inline_small_functions): Fix a typo.
17096 2018-02-16 Nathan Sidwell <nathan@acm.org>
17098 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17100 2018-02-16 Carl Love <cel@us.ibm.com>
17102 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17103 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17104 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17105 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17106 expansion to P8V_BUILTIN_VEC_FLOAT2.
17108 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17110 PR rtl-optimization/70023
17111 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17112 src_regno into account.
17114 2018-02-16 Carl Love <cel@us.ibm.com>
17116 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17117 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17118 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17119 * config/rs6000/rs6000.c: Remove case statements for
17120 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17121 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17122 and P9V_BUILTIN_VEC_VINSERT4B.
17123 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17124 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17125 * config/rs6000/vsx.md:
17126 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17129 2018-02-16 Carl Love <cel@us.ibm.com>
17131 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17133 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17135 * config/rs6000/rs6000-c.c: Add the definitions for
17136 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17137 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17138 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17139 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17140 definition for insert4b and define insn *insert3b_internal.
17141 * doc/extend.texi: Add documentation for vec_extract4b.
17143 2018-02-16 Nathan Sidwell <nathan@acm.org>
17145 * doc/extend.texi (Backwards Compatibility): Mention friend
17146 injection. Note for-scope is deprecated.
17147 * doc/invoke.texi (-ffriend-injection): Deprecate.
17149 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17151 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17152 that moved to I2, also allow destinations that are a paradoxical
17153 subreg (instead of a normal reg).
17155 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17158 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17161 2018-02-16 Richard Biener <rguenther@suse.de>
17163 PR tree-optimization/84037
17164 PR tree-optimization/84016
17166 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17167 Adjust vec_construct for the fact we need additional higher latency
17168 128bit inserts for AVX256 and AVX512 vector builds.
17169 (ix86_add_stmt_cost): Scale vector construction cost for
17172 2018-02-16 Richard Biener <rguenther@suse.de>
17174 PR tree-optimization/84417
17175 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17176 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17177 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17179 2018-02-16 Martin Liska <mliska@suse.cz>
17182 * internal-fn.def (ASAN_CHECK): Set proper flags.
17183 (ASAN_MARK): Likewise.
17185 2018-02-16 Julia Koval <julia.koval@intel.com>
17187 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17188 from PTA_CANNONLAKE.
17190 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17193 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
17194 Use ++iter rather than iter++ for std::list iterators.
17195 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
17196 defer deleting them until all nodes in the forest are processed. Do
17197 free even leaf nodes. Change to_process into auto_vec.
17200 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
17201 * vec.h (vec_default_construct): Use memset instead of placement new
17202 if BROKEN_VALUE_INITIALIZATION is defined.
17203 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
17204 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
17207 PR rtl-optimization/83723
17208 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
17209 * lra.c (lra_substitute_pseudo): Likewise. If true, use
17210 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
17212 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
17214 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
17216 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
17218 PR rtl-optimization/81443
17219 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
17220 from inner REGs to paradoxical SUBREGs.
17222 2018-02-16 Richard Biener <rguenther@suse.de>
17224 PR tree-optimization/84399
17225 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17226 For operands we can analyze at their definition make sure we can
17227 analyze them at each use as well.
17229 2018-02-16 Richard Biener <rguenther@suse.de>
17231 PR tree-optimization/84190
17232 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
17233 volatile accesses if the decl isn't volatile.
17235 2018-02-15 Jason Merrill <jason@redhat.com>
17237 PR c++/84314 - ICE with templates and fastcall attribute.
17238 * attribs.c (build_type_attribute_qual_variant): Don't clobber
17239 TYPE_CANONICAL on an existing type.
17241 2018-02-15 Jakub Jelinek <jakub@redhat.com>
17243 PR tree-optimization/84383
17244 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
17245 dstoff nor call operand_equal_p if dstbase is NULL.
17247 PR tree-optimization/84334
17248 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
17249 also a CONSTANT_CLASS_P, punt.
17251 2018-02-14 Jim Wilson <jimw@sifive.com>
17253 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
17254 first SMALL_OPERAND check. New local min_second_step. Move assert
17255 to where locals are set. Add TARGET_RVC support.
17256 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
17258 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
17260 * doc/invoke.texi: Correct -Wformat-overflow code sample.
17262 2018-02-14 Martin Sebor <msebor@redhat.com>
17264 PR tree-optimization/83698
17265 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
17266 arrays constrain the offset range to their bounds.
17267 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
17268 (builtin_access::overlap): Avoid setting the size of overlap if it's
17270 (maybe_diag_overlap): Also consider arrays when deciding what values
17271 of offsets to include in diagnostics.
17273 2018-02-14 Martin Sebor <msebor@redhat.com>
17276 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
17277 that correspond to the kind of a declaration.
17279 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
17282 * config/pa/pa.md: Load address of PIC label using the linkage table
17283 if the label is nonlocal.
17285 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
17287 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
17288 warning message if user requests -maltivec=be.
17289 * doc/invoke.texi: Document deprecation of -maltivec=be.
17291 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
17294 * config/rs6000/rs6000-c.c: Update definitions for
17295 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
17296 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
17298 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17301 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
17302 add _get_ssp intrinsics. Remove argument from
17303 __builtin_ia32_rdssp[d|q].
17304 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
17305 * config/i386/i386-builtin.def: Remove argument from
17306 __builtin_ia32_rdssp[d|q].
17307 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
17308 ix86_expand_special_args_builtin for _rdssp[d|q].
17309 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
17310 Clear register before usage.
17311 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
17312 Add documentation for new _get_ssp and _inc_ssp intrinsics.
17314 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
17316 PR tree-optimization/84357
17317 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
17318 operand 1 of an ARRAY_REF too.
17320 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
17323 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
17324 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17326 (set_of_reg): New struct.
17327 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
17328 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
17329 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17331 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
17332 Split into bitclr, bitset, bitinvert patterns if appropriate.
17333 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
17334 use rx_fuse_in_memory_bitop.
17335 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
17336 to named insn, correct maximum insn length.
17338 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
17341 * machmode.def: Define a complex mode for PARTIAL_INT.
17342 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
17344 * doc/rtl.texi: Document CSPImode.
17345 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
17347 (msp430_hard_regno_nregs_with_padding): Likewise.
17349 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
17352 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
17354 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
17356 PR rtl-optimization/84169
17357 * combine.c (try_combine): New variable split_i2i3. Set it to true if
17358 we generated a parallel as new i3 and we split that to new i2 and i3
17359 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
17360 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
17361 those to i2, not i1. Partially rewrite this scan code.
17363 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17366 * stor-layout.c (place_field): For variable length fields, adjust
17367 offset_align afterwards not just based on the field's alignment,
17368 but also on the size.
17370 PR middle-end/84309
17371 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
17372 of exps and logs in the use_exp2 case.
17374 2018-02-13 Jeff Law <law@redhat.com>
17376 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
17377 entry for "vector".
17379 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
17382 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
17386 * common.opt (gas-loc-support, gas-locview-support): New.
17387 (ginline-points, ginternal-reset-location-views): New.
17388 * doc/invoke.texi: Document them. Use @itemx where intended.
17389 (gvariable-location-views): Adjust.
17390 * target.def (reset_location_view): New.
17391 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
17392 (TARGET_RESET_LOCATION_VIEW): New.
17393 * doc/tm.texi: Rebuilt.
17394 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
17395 (dwarf2out_default_as_locview_support): New.
17396 (output_asm_line_debug_info): Use option variables.
17397 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
17398 (output_loc_list): Likewise.
17399 (add_high_low_attributes): Check option variables.
17400 Don't output entry view attribute in strict mode.
17401 (gen_inlined_subroutine_die): Check option variables.
17402 (dwarf2out_inline_entry): Likewise.
17403 (init_sections_and_labels): Likewise.
17404 (dwarf2out_early_finish): Likewise.
17405 (maybe_reset_location_view): New, from...
17406 (dwarf2out_var_location): ... here. Call it.
17407 * debug.h (dwarf2out_default_as_loc_support): Declare.
17408 (dwarf2out_default_as_locview_support): Declare.
17409 * hooks.c (hook_int_rtx_insn_0): New.
17410 * hooks.h (hook_int_rtx_insn_0): Declare.
17411 * toplev.c (process_options): Take -gas-loc-support and
17412 -gas-locview-support from dwarf2out. Enable
17413 -gvariable-location-views by default only with locview
17414 assembler support. Enable -ginternal-reset-location-views by
17415 default only if the target defines the corresponding hook.
17416 Enable -ginline-points by default if location views are
17417 enabled; force it disabled if statement frontiers are
17419 * tree-inline.c (expand_call_inline): Check option variables.
17420 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
17422 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
17424 PR tree-optimization/84321
17425 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
17426 handling. Also check whether the anti-range contains any values
17427 that satisfy the mask; switch to a VR_RANGE if not.
17429 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
17432 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
17434 2018-02-13 Martin Jambor <mjambor@suse.cz>
17437 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
17438 of call statements, also set location of a load to a temporary.
17440 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
17442 * config/rl78/rl78.c (add_vector_labels): New function.
17443 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
17444 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
17445 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
17446 which checks that no arguments are passed.
17447 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
17448 * doc/extend.texi: Documentation for the new attribute.
17450 2018-02-13 Andreas Schwab <schwab@suse.de>
17452 * config/riscv/linux.h (CPP_SPEC): Define.
17454 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17457 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
17458 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
17459 OPTION_MASK_ISA_AES as first argument to def_builtin_const
17460 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
17461 instead of OPTION_MASK_ISA_PCLMUL as first argument to
17462 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
17463 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
17464 temporarily for AES and PCLMUL builtins.
17466 PR tree-optimization/84339
17467 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
17468 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
17471 PR middle-end/84309
17472 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
17473 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
17474 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
17476 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
17478 * omp-simd-clone.h: New file.
17479 * omp-simd-clone.c: Include omp-simd-clone.h.
17480 (expand_simd_clones): No longer static.
17481 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
17482 cgraph.h and omp-simd-clone.h.
17483 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
17484 (vect_recog_widen_shift_pattern): Formatting fix.
17485 (vect_pattern_recog_1): Don't check optab for calls.
17488 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
17489 operands[2] into a REG before using gen_lowpart on it.
17491 2018-02-12 Jeff Law <law@redhat.com>
17494 * config/sh/sh.c (find_barrier): Consider a sibling call
17497 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
17498 successfully back substituting a reg.
17500 2018-02-12 Richard Biener <rguenther@suse.de>
17502 PR tree-optimization/84037
17503 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
17504 parameter, move visited init to caller.
17505 (vect_slp_analyze_operations): Separate cost from validity
17506 check, initialize visited once for all instances.
17507 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
17509 * tree-vect-stmts.c (vect_model_simple_cost): Make early
17511 (vect_model_promotion_demotion_cost): Likewise.
17512 (vectorizable_bswap): Guard cost modeling with !slp_node
17513 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
17515 (vectorizable_call): Likewise.
17516 (vectorizable_conversion): Likewise.
17517 (vectorizable_assignment): Likewise.
17518 (vectorizable_shift): Likewise.
17519 (vectorizable_operation): Likewise.
17520 (vectorizable_store): Likewise.
17521 (vectorizable_load): Likewise.
17522 (vectorizable_condition): Likewise.
17523 (vectorizable_comparison): Likewise.
17525 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
17528 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
17529 (ASAN_MARK): Fix fnspec to account for return value, change pointer
17530 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
17532 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17534 PR middle-end/83665
17535 * params.def (inline-min-speedup): Increase from 8 to 15.
17536 (max-inline-insns-auto): Decrease from 40 to 30.
17537 * ipa-split.c (consider_split): Add some buffer for function to
17538 be considered inlining candidate.
17539 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
17542 2018-02-12 Richard Biener <rguenther@suse.de>
17544 PR tree-optimization/84037
17545 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
17546 matched stmts if we cannot swap the non-matched ones.
17548 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17550 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
17551 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
17552 _mm_maskz_scalef_round_ss): New intrinsics.
17553 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
17554 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
17555 __builtin_ia32_scalefss_round): Remove.
17556 (__builtin_ia32_scalefsd_mask_round,
17557 __builtin_ia32_scalefss_mask_round): New intrinsics.
17558 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
17559 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
17560 ((match_operand:VF_128 2 "<round_nimm_predicate>"
17561 "<round_constraint>")): Changed to ...
17562 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
17563 "<round_scalar_constraint>")): ... this.
17564 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
17565 %0, %1, %2<round_op3>}"): Changed to ...
17566 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
17567 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
17568 %2<round_scalar_mask_op3>}"): ... this.
17569 * config/i386/subst.md (round_scalar_nimm_predicate): New.
17571 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17573 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
17574 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
17575 (_mm_maskz_sqrt_round_ss): New intrinsics.
17576 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
17577 (__builtin_ia32_sqrtsd_mask_round)
17578 (__builtin_ia32_sqrtss_mask_round): New builtins.
17579 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
17580 (__builtin_ia32_sqrtss_round): Remove.
17581 (__builtin_ia32_sqrtsd_mask_round)
17582 (__builtin_ia32_sqrtss_mask_round): New builtins.
17583 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
17584 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
17585 ((match_operand:VF_128 1 "vector_operand"
17586 "xBm,<round_constraint>")): Changed to ...
17587 ((match_operand:VF_128 1 "vector_operand"
17588 "xBm,<round_scalar_constraint>")): ... this.
17589 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
17590 %0, %2, %<iptr>1<round_op3>}): Changed to ...
17591 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
17592 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
17593 %<iptr>1<round_scalar_mask_op3>}): ... this.
17594 ((set_attr "prefix" "<round_prefix>")): Changed to ...
17595 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
17597 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
17600 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
17601 Cast vec_cmpeq result to correct type.
17602 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
17603 Cast vec_cmpgt result to correct type.
17605 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
17607 * final.c (final_scan_insn_1): Renamed from...
17608 (final_scan_insn): ... this. New wrapper, to recover
17609 seen from the outermost call in recursive ones.
17610 * config/sparc/sparc.c (output_return): Drop seen from call.
17611 (output_sibcall): Likewise.
17612 * config/visium/visium.c (output_branch): Likewise.
17614 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
17616 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
17619 2018-02-10 Alan Modra <amodra@gmail.com>
17622 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
17623 Specify LR as an input.
17625 2018-02-10 Jakub Jelinek <jakub@redhat.com>
17628 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
17629 remove_member_access_dummy_vars): New functions.
17630 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
17631 lower_omp_1, execute_lower_omp): Use them.
17633 PR rtl-optimization/84308
17634 * shrink-wrap.c (spread_components): Release todo vector.
17636 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
17638 PR rtl-optimization/57193
17639 * ira-color.c (struct allocno_color_data): Add member
17640 conflict_allocno_hard_prefs.
17641 (update_conflict_allocno_hard_prefs): New.
17642 (bucket_allocno_compare_func): Add a preference based on
17643 conflict_allocno_hard_prefs.
17644 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
17645 (color_allocnos): Remove a dead code. Initiate
17646 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
17648 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17651 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
17652 constraint from =wa to wa. Avoid a subreg on the output operand,
17653 instead use a pseudo and subreg it in a move.
17654 (p9_xxbrd_<mode>): Changed to ...
17655 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
17656 (p9_xxbrd_v2df): New expander.
17657 (p9_xxbrw_<mode>): Changed to ...
17658 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
17659 (p9_xxbrw_v4sf): New expander.
17661 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17663 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
17665 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
17668 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
17669 multiply in 32-bit mode.
17670 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
17671 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
17674 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17676 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
17677 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
17678 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
17679 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
17681 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
17684 * dwarf2out.c (is_trivial_indirect_ref): New function.
17685 (dwarf2out_late_global_decl): Do not generate a location
17686 attribute for variables that have a non-trivial DECL_VALUE_EXPR
17687 and that are not defined in the current unit.
17689 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17691 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
17692 instead of a libcall for UNORDERED.
17694 2018-02-09 Tamar Christina <tamar.christina@arm.com>
17697 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
17698 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
17700 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17703 * config/s390/s390.c (s390_set_current_function): Invoke
17704 s390_indirect_branch_settings also if fndecl didn't change.
17706 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17708 * config/rs6000/rs6000.md (blockage): Set length to zero.
17710 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17712 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
17714 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17717 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
17718 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
17722 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
17723 PARALLEL incoming that failed vt_get_decl_and_offset check.
17725 PR middle-end/84237
17726 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
17727 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
17729 (get_variable_section): For decls in named .bss* sections pass true as
17730 second argument to bss_initializer_p.
17732 2018-02-09 Marek Polacek <polacek@redhat.com>
17733 Jakub Jelinek <jakub@redhat.com>
17736 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
17737 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
17738 Sync some changes from cxx_fold_indirect_ref.
17740 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17742 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
17744 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
17745 (BLOCK_INLINE_ENTRY_LABEL): New.
17746 (dwarf2out_var_location): Disregard inline entry markers.
17747 (inline_entry_data): New struct.
17748 (inline_entry_data_hasher): New hashtable type.
17749 (inline_entry_data_hasher::hash): New.
17750 (inline_entry_data_hasher::equal): New.
17751 (inline_entry_data_table): New variable.
17752 (add_high_low_attributes): Add DW_AT_entry_pc and
17753 DW_AT_GNU_entry_view attributes if a pending entry is found
17754 in inline_entry_data_table. Add old entry_pc attribute only
17755 if debug nonbinding markers are disabled.
17756 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
17757 markers are enabled.
17758 (block_within_block_p, dwarf2out_inline_entry): New.
17759 (dwarf2out_finish): Check that no entries remained in
17760 inline_entry_data_table.
17761 * final.c (reemit_insn_block_notes): Handle inline entry notes.
17762 (final_scan_insn, notice_source_line): Likewise.
17763 (rest_of_clean_state): Skip inline entry markers.
17764 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
17766 * gimple.c (gimple_build_debug_inline_entry): New.
17767 * gimple.h (enum gimple_debug_subcode): Add
17768 GIMPLE_DEBUG_INLINE_ENTRY.
17769 (gimple_build_debug_inline_entry): Declare.
17770 (gimple_debug_inline_entry_p): New.
17771 (gimple_debug_nonbind_marker_p): Adjust.
17772 * insn-notes.def (INLINE_ENTRY): New.
17773 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
17774 inline entry marker notes.
17775 (print_insn): Likewise.
17776 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
17777 (INSN_DEBUG_MARKER_KIND): Likewise.
17778 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
17779 * tree-inline.c (expand_call_inline): Build and insert
17780 debug_inline_entry stmt.
17781 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
17782 inline entry blocks early, if nonbind markers are enabled.
17783 (dump_scope_block): Dump fragment info.
17784 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
17785 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
17786 (gimple_build_debug_inline_entry): New.
17787 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
17788 Enable/disable inline entry points too.
17789 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
17790 (DEBUG_INSN): Describe inline entry markers.
17792 * common.opt (gvariable-location-views): New.
17793 (gvariable-location-views=incompat5): New.
17794 * config.in: Rebuilt.
17795 * configure: Rebuilt.
17796 * configure.ac: Test assembler for view support.
17797 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
17798 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
17799 * dwarf2out.c (var_loc_view): New typedef.
17800 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
17801 (dwarf2out_locviews_in_attribute): New.
17802 (dwarf2out_locviews_in_loclist): New.
17803 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
17804 (enum dw_line_info_opcode): Add LI_adv_address.
17805 (struct dw_line_info_table): Add view.
17806 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
17807 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
17808 (zero_view_p): New variable.
17809 (ZERO_VIEW_P): New macro.
17810 (output_asm_line_debug_info): New.
17811 (struct var_loc_node): Add view.
17812 (add_AT_view_list, AT_loc_list): New.
17813 (add_var_loc_to_decl): Add view param. Test it against last.
17814 (new_loc_list): Add view params. Record them.
17815 (AT_loc_list_ptr): Handle loc and view lists.
17816 (view_list_to_loc_list_val_node): New.
17817 (print_dw_val): Handle dw_val_class_view_list.
17818 (size_of_die): Likewise.
17819 (value_format): Likewise.
17820 (loc_list_has_views): New.
17821 (gen_llsym): Set vl_symbol too.
17822 (maybe_gen_llsym, skip_loc_list_entry): New.
17823 (dwarf2out_maybe_output_loclist_view_pair): New.
17824 (output_loc_list): Output view list or entries too.
17825 (output_view_list_offset): New.
17826 (output_die): Handle dw_val_class_view_list.
17827 (output_dwarf_version): New.
17828 (output_compilation_unit_header): Use it.
17829 (output_skeleton_debug_sections): Likewise.
17830 (output_rnglists, output_line_info): Likewise.
17831 (output_pubnames, output_aranges): Update version comments.
17832 (output_one_line_info_table): Output view numbers in asm comments.
17833 (dw_loc_list): Determine current endview, pass it to new_loc_list.
17834 Call maybe_gen_llsym.
17835 (loc_list_from_tree_1): Adjust.
17836 (add_AT_location_description): Create view list attribute if
17837 needed, check it's absent otherwise.
17838 (convert_cfa_to_fb_loc_list): Adjust.
17839 (maybe_emit_file): Call output_asm_line_debug_info for test.
17840 (dwarf2out_var_location): Reset views as needed. Precompute
17841 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
17842 attribute. Set view.
17843 (new_line_info_table): Reset next view.
17844 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
17845 (dwarf2out_source_line): Likewise. Output view resets and labels to
17846 the assembler, or select appropriate line info opcodes.
17847 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
17848 (optimize_string_length): Catch it. Adjust.
17849 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
17850 dw_val_class_view_list, and remove it if no longer needed.
17851 (hash_loc_list): Hash view numbers.
17852 (loc_list_hasher::equal): Compare them.
17853 (optimize_location_lists): Check whether a view list symbol is
17854 needed, and whether the locview attribute is present, and
17855 whether they match. Remove the locview attribute if no longer
17857 (index_location_lists): Call skip_loc_list_entry for test.
17858 (dwarf2out_finish): Call output_asm_line_debug_info for test.
17859 Use output_dwarf_version.
17860 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
17861 (struct dw_val_node): Add val_view_list.
17862 * final.c (SEEN_NEXT_VIEW): New.
17863 (set_next_view_needed): New.
17864 (clear_next_view_needed): New.
17865 (maybe_output_next_view): New.
17866 (final_start_function): Rename to...
17867 (final_start_function_1): ... this. Take pointer to FIRST,
17868 add SEEN parameter. Emit param bindings in the initial view.
17869 (final_start_function): Reintroduce SEEN-less interface.
17870 (final): Rename to...
17871 (final_1): ... this. Take SEEN parameter. Output final pending
17872 next view at the end.
17873 (final): Reintroduce seen-less interface.
17874 (final_scan_insn): Output pending next view before switching
17875 sections or ending a block. Mark the next view as needed when
17876 outputting variable locations. Notify debug backend of section
17877 changes, and of location view changes.
17878 (rest_of_handle_final): Adjust.
17879 * toplev.c (process_options): Autodetect value for debug variable
17880 location views option. Warn on incompat5 without -gdwarf-5.
17881 * doc/invoke.texi (gvariable-location-views): New.
17882 (gvariable-location-views=incompat5): New.
17883 (gno-variable-location-views): New.
17885 2018-02-08 David Malcolm <dmalcolm@redhat.com>
17887 PR tree-optimization/84136
17888 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
17889 that the result of find_edge is non-NULL.
17891 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17894 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
17895 storing integer register in SImode. Fix cost of 256 and 512
17896 byte aligned SSE register store.
17898 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17900 * config/i386/i386.c (ix86_multiplication_cost): Fix
17901 multiplication cost for TARGET_AVX512DQ.
17903 2018-02-08 Marek Polacek <polacek@redhat.com>
17905 PR tree-optimization/84238
17906 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
17909 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17911 PR tree-optimization/84265
17912 * tree-vect-stmts.c (vectorizable_store): Don't treat
17913 VMAT_CONTIGUOUS accesses as grouped.
17914 (vectorizable_load): Likewise.
17916 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17918 PR tree-optimization/81635
17919 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
17920 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
17921 (test_round_for_mask): New functions.
17922 (wide_int_cc_tests): Call test_round_for_mask.
17923 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
17924 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
17925 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
17926 range returned by get_range_info.
17928 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17931 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
17932 * symtab.c: Include builtins.h
17933 (symtab_node::output_to_lto_symbol_table_p): Move here
17934 from lto-streamer-out.c:output_symbol_p.
17935 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
17936 (output_symbol_p): Move all logic to symtab.c
17937 (produce_symtab): Update.
17939 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17941 * config/s390/s390-opts.h (enum indirect_branch): Define.
17942 * config/s390/s390-protos.h (s390_return_addr_from_memory)
17943 (s390_indirect_branch_via_thunk)
17944 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
17945 (enum s390_indirect_branch_type): Define.
17946 * config/s390/s390.c (struct s390_frame_layout, struct
17947 machine_function): Remove.
17948 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
17949 (indirect_branch_table_label_no, indirect_branch_table_name):
17951 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
17952 (enum s390_indirect_branch_option): Define.
17953 (s390_return_addr_from_memory): New function.
17954 (s390_handle_string_attribute): New function.
17955 (s390_attribute_table): Add new attribute handler.
17956 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17957 (s390_indirect_branch_via_thunk): New function.
17958 (s390_indirect_branch_via_inline_thunk): New function.
17959 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17960 sibling call optimization for non z10 compiles.
17961 (s390_emit_call): Force indirect branch target to be a single
17962 register. Add r1 clobber for non-z10 compiles.
17963 (s390_emit_epilogue): Emit return jump via return_use expander.
17964 (s390_reorg): Handle JUMP_INSNs as execute targets.
17965 (s390_option_override_internal): Perform validity checks for the
17966 new command line options.
17967 (s390_indirect_branch_attrvalue): New function.
17968 (s390_indirect_branch_settings): New function.
17969 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17970 (s390_output_indirect_thunk_function): New function.
17971 (s390_code_end): Implement target hook.
17972 (s390_case_values_threshold): Implement target hook.
17973 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17975 * config/s390/s390.h (struct s390_frame_layout)
17976 (struct machine_function): Move here from s390.c.
17977 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17978 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17979 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17980 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17981 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17982 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17983 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17984 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17985 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17986 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17987 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17988 (mnemonic attribute): Add values which aren't recognized
17990 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17991 pattern for branch conversion. Fix mnemonic attribute.
17992 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17993 indirect branch via thunk if requested.
17994 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17995 ("*indirect_jump"): Disable for branch conversion using out of
17997 ("indirect_jump_via_thunk<mode>_z10")
17998 ("indirect_jump_via_thunk<mode>")
17999 ("indirect_jump_via_inlinethunk<mode>_z10")
18000 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18001 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18002 ("casesi_jump_via_inlinethunk<mode>_z10")
18003 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18004 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18005 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18006 ("*indirect2_jump"): Disable for branch conversion.
18007 ("casesi_jump"): Turn into expander and expand patterns for branch
18009 ("return_use"): New expander.
18010 ("*return"): Emit return via thunk and rename it to ...
18011 ("*return<mode>"): ... this one.
18012 * config/s390/s390.opt: Add new options and and enum for the
18015 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18017 * lra-constraints.c (match_reload): Unconditionally use
18018 gen_lowpart_SUBREG, rather than selecting between that
18019 and equivalent gen_rtx_SUBREG code.
18021 2018-02-08 Richard Biener <rguenther@suse.de>
18023 PR tree-optimization/84233
18024 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18025 changed flag instead of boguously re-using phi_inserted.
18027 2018-02-08 Martin Jambor <mjambor@suse.cz>
18029 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18030 static local variables.
18032 2018-02-08 Richard Biener <rguenther@suse.de>
18034 PR tree-optimization/84278
18035 * tree-vect-stmts.c (vectorizable_store): When looking for
18036 smaller vector types to perform grouped strided loads/stores
18037 make sure the mode is supported by the target.
18038 (vectorizable_load): Likewise.
18040 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18042 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18043 Increase LDP/STP opportunities by adding adjacent callee-saves.
18045 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18047 PR rtl-optimization/84068
18048 PR rtl-optimization/83459
18049 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18051 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18053 PR tree-optimization/84224
18054 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18055 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18056 non-zero arguments.
18058 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18061 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18062 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18063 count, remove one USE.
18065 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18067 * doc/install.texi (Configuration): Document the
18068 --with-long-double-format={ibm,ieee} PowerPC configuration
18072 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18073 Convert from define_expand to be define_insn_and_split. Rework
18074 float/double/_Float128 conversions to QI/HI/SImode to work with
18075 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18076 conversions to QI/HImode types did a store and then a load to
18077 truncate the value. For conversions to VSX registers, don't split
18078 the insn, instead emit the code directly. Use the code iterator
18079 any_fix to combine signed and unsigned conversions.
18080 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18081 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18082 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18083 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18084 (fix_<mode>di2_hw): Likewise.
18085 (fixuns_<mode>di2_hw): Likewise.
18086 (fix_<mode>si2_hw): Likewise.
18087 (fixuns_<mode>si2_hw): Likewise.
18088 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18089 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18090 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18091 fix<uns>_trunc<SFDF:mode>si2_p8.
18092 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18093 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18094 (fix<uns>_<mode>_mem): Likewise.
18095 (fctiw<u>z_<mode>_mem): Likewise.
18096 (fix<uns>_<mode>_mem): Likewise.
18097 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18098 the register allocator from doing a direct move to the GPRs to do
18099 a store, and instead use the ISA 3.0 store byte/half-word from
18100 vector register instruction. For IEEE 128-bit floating point,
18101 also optimize stores of 32-bit ints.
18102 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18104 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18106 * genextract.c (push_pathstr_operand): New function to support
18108 (walk_rtx): Call push_pathstr_operand.
18109 (print_path): Support [a-zA-Z].
18111 2018-02-07 Richard Biener <rguenther@suse.de>
18113 PR tree-optimization/84037
18114 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18115 (cse_and_gimplify_to_preheader): Declare.
18116 (vect_get_place_in_interleaving_chain): Likewise.
18117 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18119 (_loop_vec_info::~_loop_vec_info): Delete it.
18120 (cse_and_gimplify_to_preheader): New function.
18121 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18122 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18123 (vectorizable_load): Likewise. For grouped stores always base
18124 the IV on the first element.
18125 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18126 condition before gimplifying.
18128 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18130 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18131 *DIV_EXPR and *MOD_EXPR.
18133 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18136 * config/i386/i386.c (ix86_option_override_internal): Mask out
18137 the CF_SET bit when checking -fcf-protection.
18139 2018-02-07 Tom de Vries <tom@codesourcery.com>
18142 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18145 2018-02-07 Richard Biener <rguenther@suse.de>
18147 PR tree-optimization/84204
18148 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18151 PR tree-optimization/84205
18152 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18153 special-case isl_ast_op_zdiv_r.
18155 PR tree-optimization/84223
18156 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18157 Only add conditions from within the region.
18158 (gather_bbs::after_dom_children): Adjust.
18160 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18163 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18164 * config/avr/avr.md: Only post-reload split REG-REG moves if
18165 either register is GENERAL_REG_P.
18167 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18169 PR tree-optimization/84235
18170 * tree-ssa-scopedtables.c
18171 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18172 if the subtraction is performed in floating point type where NaNs are
18173 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18174 build 1. Formatting fix.
18176 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18179 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18180 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18181 and skip it regardless of bb boundaries. Use CALL_P macro,
18182 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18183 unnecessarily, formatting fix.
18185 2018-02-06 Michael Collison <michael.collison@arm.com>
18187 * config/arm/thumb2.md:
18188 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
18189 (*thumb_mov_notscc): Ditto.
18191 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
18194 * config/rs6000/rs6000.md (su code attribute): Use "u" for
18195 unsigned_fix, not "s".
18197 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18199 * configure.ac (gcc_fn_eh_frame_ro): New function.
18200 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
18201 correct .eh_frame permissions.
18202 * configure: Regenerate.
18204 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
18206 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
18207 irrelevant options.
18209 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18211 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18212 Display warning message for -mno-speculate-indirect-jumps.
18214 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
18216 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
18218 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
18220 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
18222 PR tree-optimization/84225
18223 * tree-eh.c (find_trapping_overflow): Only call
18224 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
18226 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18229 * config/i386/i386.c: Reimplement the check of possible options
18230 -mibt/-mshstk conbination. Change error messages.
18231 * doc/invoke.texi: Fix a typo: remove extra '='.
18233 2018-02-06 Marek Polacek <polacek@redhat.com>
18235 PR tree-optimization/84228
18236 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
18238 2018-02-06 Tamar Christina <tamar.christina@arm.com>
18241 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
18242 emitted arch directives.
18243 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
18244 __ARM_FEATURE_COPROC before changing architectures.
18246 2018-02-06 Richard Biener <rguenther@suse.de>
18248 * config/i386/i386.c (print_reg): Fix typo.
18249 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
18251 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
18253 * configure: Regenerate.
18255 2018-02-05 Martin Sebor <msebor@redhat.com>
18257 PR tree-optimization/83369
18258 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
18261 2018-02-05 Martin Liska <mliska@suse.cz>
18263 * doc/invoke.texi: Cherry-pick upstream r323995.
18265 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
18267 * ira.c (ira_init_register_move_cost): Adjust comment.
18269 2018-02-05 Martin Liska <mliska@suse.cz>
18271 PR gcov-profile/84137
18272 * doc/gcov.texi: Fix typo in documentation.
18274 2018-02-05 Martin Liska <mliska@suse.cz>
18276 PR gcov-profile/83879
18277 * doc/gcov.texi: Document necessity of --dynamic-list-data when
18278 using dlopen functionality.
18280 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
18282 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
18283 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
18284 _mm_maskz_range_ss, _mm_mask_range_round_ss,
18285 _mm_maskz_range_round_ss): New intrinsics.
18286 (__builtin_ia32_rangesd128_round)
18287 (__builtin_ia32_rangess128_round): Remove.
18288 (__builtin_ia32_rangesd128_mask_round,
18289 __builtin_ia32_rangess128_mask_round): New builtins.
18290 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
18291 __builtin_ia32_rangess128_round): Remove.
18292 (__builtin_ia32_rangesd128_mask_round,
18293 __builtin_ia32_rangess128_mask_round): New builtins.
18294 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
18295 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
18296 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
18297 "<round_saeonly_constraint>")): Changed to ...
18298 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
18299 "<round_saeonly_scalar_constraint>")): ... this.
18300 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
18301 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
18302 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
18303 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
18304 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
18306 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
18308 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
18310 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
18311 Remove all values except native, 8540 and 8548.
18313 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
18315 * config/i386/i386.c (ix86_output_function_return): Pass
18316 INVALID_REGNUM, instead of -1, as invalid register number to
18317 indirect_thunk_name and output_indirect_thunk.
18319 2018-02-02 Julia Koval <julia.koval@intel.com>
18321 * config.gcc: Add -march=icelake.
18322 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
18323 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
18324 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
18325 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
18326 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
18327 (processor_target_table): Add icelake.
18328 (ix86_option_override_internal): Handle new PTAs.
18329 (get_builtin_code_for_version): Handle icelake.
18330 (M_INTEL_COREI7_ICELAKE): New.
18331 (fold_builtin_cpu): Handle icelake.
18332 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
18333 * doc/invoke.texi: Add -march=icelake.
18335 2018-02-02 Julia Koval <julia.koval@intel.com>
18337 * config/i386/i386.c (ix86_option_override_internal): Change flags type
18338 to wide_int_bitmask.
18339 * wide-int-bitmask.h: New.
18341 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18344 * config/i386/i386.md: Replace Pmode with word_mode in
18345 builtin_setjmp_setup and builtin_longjmp to support x32.
18347 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
18351 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
18353 (rs6000_supported_cpu_names): New static variable.
18354 (linux_cpu_translation_table): Likewise.
18355 (elf_platform) <cpu>: Define new static variable and use it.
18356 Translate kernel AT_PLATFORM name to canonical name if needed.
18357 Error if platform name is unknown.
18359 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
18362 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
18364 2018-02-01 Jeff Law <law@redhat.com>
18367 * config/i386/i386.c (release_scratch_register_on_entry): Add new
18368 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
18369 the scratch if RELEASE_VIA_POP is false.
18370 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
18371 If we have to save a temporary register, decrement SIZE appropriately.
18372 Pass new arguments to release_scratch_register_on_entry.
18373 (ix86_adjust_stack_and_probe): Likewise.
18374 (ix86_emit_probe_stack_range): Pass new arguments to
18375 release_scratch_register_on_entry.
18377 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
18379 PR rtl-optimization/84157
18380 * combine.c (change_zero_ext): Use REG_P predicate in
18381 front of HARD_REGISTER_P predicate.
18383 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
18385 * config/avr/avr.c (avr_option_override): Move disabling of
18386 -fdelete-null-pointer-checks to...
18387 * common/config/avr/avr-common.c (avr_option_optimization_table):
18390 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18392 PR tree-optimization/81635
18393 * tree-data-ref.c (split_constant_offset_1): For types that
18394 wrap on overflow, try to use range info to prove that wrapping
18397 2018-02-01 Renlin Li <renlin.li@arm.com>
18400 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
18401 TAILCALL_ADDR_REGS.
18402 (aarch64_register_move_cost): Likewise.
18403 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
18404 TAILCALL_ADDR_REGS.
18405 (REG_CLASS_NAMES): Likewise.
18406 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
18407 TAILCALL_ADDR_REGS. Remove IP registers.
18408 * config/aarch64/aarch64.md (Ucs): Update register constraint.
18410 2018-02-01 Richard Biener <rguenther@suse.de>
18412 * domwalk.h (dom_walker::dom_walker): Add additional constructor
18413 for specifying RPO order and allow NULL for that.
18414 * domwalk.c (dom_walker::dom_walker): Likewise.
18415 (dom_walker::walk): Handle NULL RPO order.
18416 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
18418 (rewrite_update_dom_walker): Likewise.
18419 (mark_def_dom_walker): Likewise.
18421 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18423 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
18424 (aarch64_maybe_expand_sve_subreg_move): Declare.
18425 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
18426 * config/aarch64/predicates.md (aarch64_any_register_operand): New
18428 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
18429 that are semantically a reverse operation.
18430 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
18431 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
18432 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
18434 (aarch64_can_change_mode_class): For big-endian, forbid changes
18435 between two SVE modes if they have different element sizes.
18437 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18439 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
18440 the TImode handling for big-endian targets.
18442 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18444 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
18445 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
18447 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
18448 Remove BSWAP handing for big-endian targets and use the form of
18449 LD1RQ appropariate for the mode.
18451 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18453 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
18454 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
18455 duplicated element.
18457 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18460 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
18461 check for operands that need to go through aarch64_sve_reload_be.
18463 2018-02-01 Jakub Jelinek <jakub@redhat.com>
18465 PR tree-optimization/81661
18466 PR tree-optimization/84117
18467 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
18468 * tree-eh.c: Include gimplify.h.
18469 (find_trapping_overflow, replace_trapping_overflow,
18470 rewrite_to_non_trapping_overflow): New functions.
18471 * tree-vect-loop.c: Include tree-eh.h.
18472 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
18473 * tree-data-ref.c: Include tree-eh.h.
18474 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
18476 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
18478 PR rtl-optimization/84123
18479 * combine.c (change_zero_ext): Check if hard register satisfies
18480 can_change_dest_mode before calling gen_lowpart_SUBREG.
18482 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
18485 * ira.c (ira_init_register_move_cost): Remove assert.
18487 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18489 PR rtl-optimization/84071
18490 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
18491 * doc/tm.texi: Regenerate.
18493 2018-01-31 Richard Biener <rguenther@suse.de>
18495 PR tree-optimization/84132
18496 * tree-data-ref.c (analyze_miv_subscript): Properly
18497 check whether evolution_function_is_affine_multivariate_p
18498 before calling gcd_of_steps_may_divide_p.
18500 2018-01-31 Julia Koval <julia.koval@intel.com>
18503 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
18504 * config/i386/i386.md (rdpid_rex64) New.
18505 (rdpid): Make 32bit only.
18507 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
18510 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
18511 an IDENTIFIER_NODE for FUNCTION_TYPE's.
18513 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18516 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18518 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18520 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18522 PR rtl-optimization/84071
18523 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
18524 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
18526 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
18528 * config/arc/arc.c (arc_handle_aux_attribute): New function.
18529 (arc_attribute_table): Add 'aux' attribute.
18530 (arc_in_small_data_p): Consider aux like variables.
18531 (arc_is_aux_reg_p): New function.
18532 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
18533 (arc_get_aux_arg): New function.
18534 (prepare_move_operands): Handle aux-register access.
18535 (arc_handle_aux_attribute): New function.
18536 * doc/extend.texi (ARC Variable attributes): Add subsection.
18538 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
18540 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
18541 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
18542 (arc_attribute_table): Add 'uncached' attribute.
18543 (arc_print_operand): Print '.di' flag for uncached memory
18545 (arc_in_small_data_p): Do not consider for small data the uncached
18547 (arc_is_uncached_mem_p): New function.
18548 * config/arc/predicates.md (compact_store_memory_operand): Check
18549 for uncached memory accesses.
18550 (nonvol_nonimm_operand): Likewise.
18551 * doc/extend.texi (ARC Type Attribute): New subsection.
18553 2018-01-31 Jakub Jelinek <jakub@redhat.com>
18556 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
18557 falign-loops=): Add Optimization flag.
18559 2018-01-30 Jeff Law <law@redhat.com>
18562 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
18563 INT_REGISTERS_SAVED. Check it prior to calling
18564 get_scratch_register_on_entry.
18565 (ix86_adjust_stack_and_probe): Similarly.
18566 (ix86_emit_probe_stack_range): Similarly.
18567 (ix86_expand_prologue): Corresponding changes.
18569 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18572 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
18573 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
18575 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
18578 * lra-constraints.c (curr_insn_transform): Process AND in the
18581 2018-01-30 Jakub Jelinek <jakub@redhat.com>
18583 PR rtl-optimization/83986
18584 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
18585 dependence against last_pending_memory_flush in addition to
18586 pending_jump_insns.
18588 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
18590 PR tree-optimization/81611
18591 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
18594 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18597 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
18600 2018-01-30 Richard Biener <rguenther@suse.de>
18601 Jakub Jelinek <jakub@redhat.com>
18603 PR tree-optimization/84111
18604 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
18605 inner loops added during recursion, as they don't have up-to-date
18608 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18611 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
18612 (can_inline_edge_by_limits_p): ... here.
18613 (can_early_inline_edge_p, check_callers,
18614 update_caller_keys, update_callee_keys, recursive_inlining,
18615 add_new_edges_to_heap, speculation_useful_p,
18616 inline_small_functions,
18617 inline_small_functions, flatten_function,
18618 inline_to_all_callers_1): Update.
18620 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18622 * profile-count.c (profile_count::combine_with_ipa_count): Handle
18625 2018-01-30 Richard Biener <rguenther@suse.de>
18627 PR tree-optimization/83008
18628 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
18629 invariant and constant vector uses in stmts when they need
18630 more than one stmt.
18632 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18635 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
18636 * configure: Regenerate.
18638 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18640 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
18642 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
18643 Use gen_rtx_REG rather than gen_lowpart.
18645 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18647 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
18648 rather than 0 when creating partial subregs.
18650 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18652 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
18655 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
18658 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
18659 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
18660 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
18661 flags. This restores the settings used before the 2017-07-24.
18662 Turning off pre increment/decrement/modify allows IVOPTS to
18663 optimize DF/SF loops where the index is an int.
18665 2018-01-29 Richard Biener <rguenther@suse.de>
18666 Kelvin Nilsen <kelvin@gcc.gnu.org>
18669 * tree-vect-stmts.c (vectorizable_call): Don't call
18670 targetm.vectorize_builtin_md_vectorized_function if callee is
18673 2018-01-22 Carl Love <cel@us.ibm.com>
18675 * doc/extend.tex: Fix typo in second arg in
18676 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
18678 2018-01-29 Richard Biener <rguenther@suse.de>
18680 PR tree-optimization/84086
18681 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
18682 (flush_ssaname_freelist): When SSA names were released reset
18683 the SCEV hash table.
18685 2018-01-29 Richard Biener <rguenther@suse.de>
18687 PR tree-optimization/84057
18688 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
18689 removed paths when removing edges.
18691 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
18693 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
18694 -mfunction-return=@var{choice}.
18696 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18698 PR diagnostic/84034
18699 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
18700 Handle CR like TAB.
18701 (layout::print_source_line): Likewise.
18702 (test_get_line_width_without_trailing_whitespace): Add test cases.
18704 2018-01-27 Jakub Jelinek <jakub@redhat.com>
18706 PR middle-end/84040
18707 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
18710 2018-01-26 Jim Wilson <jimw@sifive.com>
18712 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
18714 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
18717 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18719 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
18720 and CMP + SUB-immediate -> SUBS.
18722 2018-01-26 Martin Sebor <msebor@redhat.com>
18724 PR tree-optimization/83896
18725 * tree-ssa-strlen.c (get_string_len): Rename...
18726 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
18727 Avoid assuming length is constant.
18728 (handle_char_store): Use HOST_WIDE_INT for string length.
18730 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
18733 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
18734 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
18736 2018-01-26 Richard Biener <rguenther@suse.de>
18738 PR rtl-optimization/84003
18739 * dse.c (record_store): Only record redundant stores when
18740 the earlier store aliases at least all accesses the later one does.
18742 2018-01-26 Jakub Jelinek <jakub@redhat.com>
18744 PR rtl-optimization/83985
18745 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
18746 REG_CFA_RESTORE insns.
18747 (delete_unmarked_insns): Don't ignore separate shrink wrapping
18748 REG_CFA_RESTORE insns here.
18751 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
18752 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
18754 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18756 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
18757 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
18758 (arc_init): Likewise.
18759 (arc_override_options): Likewise.
18760 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
18762 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
18764 * config/arc/arc.h (TARGET_DBNZ): Define.
18765 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
18766 properly set the tune attribute.
18767 (dbnz): Use TARGET_DBNZ guard.
18768 * config/arc/arc.opt (mtune): Add core3 option.
18770 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18772 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
18773 recognize new pic like addresses.
18774 (arc_delegitimize_address): Clean up.
18776 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18778 * config/arc/arc-arches.def: Option mrf16 valid for all
18780 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
18781 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
18782 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
18783 * config/arc/arc-tables.opt: Regenerate.
18784 * config/arc/arc.c (arc_conditional_register_usage): Handle
18785 reduced register file case.
18786 (arc_file_start): Set must have build attributes.
18787 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
18788 mrf16 option value.
18789 * config/arc/arc.opt (mrf16): Add new option.
18790 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
18791 * config/arc/genmultilib.awk: Handle new mrf16 option.
18792 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
18793 * config/arc/t-multilib: Regenerate.
18794 * doc/invoke.texi (ARC Options): Document mrf16 option.
18796 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18798 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
18799 * config/arc/arc.c (arc_handle_secure_attribute): New function.
18800 (arc_attribute_table): Add 'secure_call' attribute.
18801 (arc_print_operand): Print secure call operand.
18802 (arc_function_ok_for_sibcall): Don't optimize tail calls when
18804 (arc_is_secure_call_p): New function. * config/arc/arc.md
18805 (call_i): Add support for sjli instruction.
18806 (call_value_i): Likewise.
18807 * config/arc/constraints.md (Csc): New constraint.
18809 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18810 John Eric Martin <John.Martin@emmicro-us.com>
18812 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
18813 * config/arc/arc.c (_arc_jli_section): New struct.
18814 (arc_jli_section): New type.
18815 (rc_jli_sections): New static variable.
18816 (arc_handle_jli_attribute): New function.
18817 (arc_attribute_table): Add jli_always and jli_fixed attribute.
18818 (arc_file_end): New function.
18819 (TARGET_ASM_FILE_END): Define.
18820 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
18821 (arc_add_jli_section): New function.
18822 (jli_call_scan): Likewise.
18823 (arc_reorg): Call jli_call_scan.
18824 (arc_output_addsi): Remove 'S' from printing asm operand.
18825 (arc_is_jli_call_p): New function.
18826 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
18828 (movhi_insn): Likewise.
18829 (movsi_insn): Likewise.
18830 (movsi_set_cc_insn): Likewise.
18831 (loadqi_update): Likewise.
18832 (load_zeroextendqisi_update): Likewise.
18833 (load_signextendqisi_update): Likewise.
18834 (loadhi_update): Likewise.
18835 (load_zeroextendhisi_update): Likewise.
18836 (load_signextendhisi_update): Likewise.
18837 (loadsi_update): Likewise.
18838 (loadsf_update): Likewise.
18839 (movsicc_insn): Likewise.
18840 (bset_insn): Likewise.
18841 (bxor_insn): Likewise.
18842 (bclr_insn): Likewise.
18843 (bmsk_insn): Likewise.
18844 (bicsi3_insn): Likewise.
18845 (cmpsi_cc_c_insn): Likewise.
18846 (movsi_ne): Likewise.
18847 (movsi_cond_exec): Likewise.
18848 (clrsbsi2): Likewise.
18849 (norm_f): Likewise.
18859 (call_i): Remove 'S' asm letter, add jli instruction.
18860 (call_value_i): Likewise.
18861 * config/arc/arc.op (mjli-always): New option.
18862 * config/arc/constraints.md (Cji): New constraint.
18863 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
18865 (subsf3_fpx): Likewise.
18866 (mulsf3_fpx): Likewise.
18867 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
18869 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
18870 function attrbutes.
18871 * doc/invoke.texi (ARC): Document mjli-always option.
18873 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
18875 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
18876 avoid addition with 0 and use incw and decw where possible.
18878 2018-01-26 Richard Biener <rguenther@suse.de>
18880 PR tree-optimization/81082
18881 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
18882 association if it requires casting to unsigned.
18883 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
18884 from fold_plusminus_mult_expr to catch important cases late when
18885 range info is available.
18887 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18889 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
18890 * configure.ac (hidden_linkonce): New test.
18891 * configure: Regenerate.
18892 * config.in: Regenerate.
18894 2018-01-26 Julia Koval <julia.koval@intel.com>
18896 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
18897 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
18898 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
18899 _mm_mask_bitshuffle_epi64_mask): Fix type.
18900 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
18901 USI_FTYPE_V4DI_V4DI_USI): Remove.
18902 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
18903 __builtin_ia32_vpshufbitqmb256_mask,
18904 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
18905 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
18906 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
18908 2018-01-26 Alan Modra <amodra@gmail.com>
18911 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
18912 UNSPEC_VBPERMQ. Sort other unspecs.
18914 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
18916 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
18918 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
18920 PR middle-end/83055
18921 * predict.c (drop_profile): Do not push/pop cfun; update also
18923 (handle_missing_profiles): Fix logic looking for zero profiles.
18925 2018-01-25 Jakub Jelinek <jakub@redhat.com>
18927 PR middle-end/83977
18928 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
18929 on functions with #pragma omp declare simd or functions with simd
18931 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
18932 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
18933 Remove trailing \n from warning_at calls.
18935 2018-01-25 Tom de Vries <tom@codesourcery.com>
18938 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18939 for neutered workers.
18941 2018-01-24 Joseph Myers <joseph@codesourcery.com>
18944 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
18945 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
18947 2018-01-24 Jeff Law <law@redhat.com>
18950 * i386.c (get_probe_interval): Move to earlier point.
18951 (ix86_compute_frame_layout): If -fstack-clash-protection and
18952 the frame is larger than the probe interval, then use pushes
18953 to save registers rather than reg->mem moves.
18954 (ix86_expand_prologue): Remove conditional for int_registers_saved
18957 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18960 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18961 min/max for never referenced object.
18963 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18965 PR middle-end/83977
18966 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18968 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18969 attributes from DECL_ATTRIBUTES (decl) without affecting
18970 DECL_ATTRIBUTES (current_function_decl).
18971 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18972 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18974 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18976 PR tree-optimization/83979
18977 * fold-const.c (fold_comparison): Use constant_boolean_node
18978 instead of boolean_{true,false}_node.
18980 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18982 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18985 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18987 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18988 Simplify the clause that sets the length attribute.
18989 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18990 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18991 clause that sets the length attribute.
18992 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18994 2018-01-24 Tom de Vries <tom@codesourcery.com>
18997 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18998 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18999 Add strict parameter.
19000 (prevent_branch_around_nothing): Insert dummy insn between branch to
19001 label and label with no ptx insn inbetween.
19002 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19004 2018-01-24 Tom de Vries <tom@codesourcery.com>
19007 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19008 for neutered threads in warp.
19009 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19011 2018-01-24 Richard Biener <rguenther@suse.de>
19013 PR tree-optimization/83176
19014 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19017 2018-01-24 Richard Biener <rguenther@suse.de>
19019 PR tree-optimization/82819
19020 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19021 code generating pluses that are no-ops in the target precision.
19023 2018-01-24 Richard Biener <rguenther@suse.de>
19025 PR middle-end/84000
19026 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19028 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19030 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19031 to merge probabilities.
19032 * predict.c (probably_never_executed): Also mark as cold functions
19033 with global 0 profile and guessed local profile.
19034 * profile-count.c (profile_probability::combine_with_count): New
19036 * profile-count.h (profile_probability::operator*,
19037 profile_probability::operator*=, profile_probability::operator/,
19038 profile_probability::operator/=): Reduce precision to adjusted
19039 and set value to guessed on contradictory divisions.
19040 (profile_probability::combine_with_freq): Remove.
19041 (profile_probability::combine_wiht_count): Declare.
19042 (profile_count::force_nonzero):: Set to adjusted.
19043 (profile_count::probability_in):: Set quality to adjusted.
19044 * tree-ssa-tail-merge.c (replace_block_by): Use
19045 combine_with_count.
19047 2018-01-23 Andrew Waterman <andrew@sifive.com>
19048 Jim Wilson <jimw@sifive.com>
19050 * config/riscv/riscv.c (riscv_stack_boundary): New.
19051 (riscv_option_override): Set riscv_stack_boundary. Handle
19052 riscv_preferred_stack_boundary_arg.
19053 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19054 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19055 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19056 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19057 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19058 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19060 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19063 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19064 of struct ix86_frame.
19065 (ix86_expand_epilogue): Likewise. Add a local variable for
19066 the reg_save_offset field in struct ix86_frame.
19068 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19070 PR tree-optimization/82604
19071 * tree-loop-distribution.c (enum partition_kind): New enum item
19072 PKIND_PARTIAL_MEMSET.
19073 (partition_builtin_p): Support above new enum item.
19074 (generate_code_for_partition): Ditto.
19075 (compute_access_range): Differentiate cases that equality can be
19076 proven at all loops, the innermost loops or no loops.
19077 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19078 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19079 (finalize_partitions, distribute_loop): Don't fuse partition of
19080 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19081 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19082 parloop is enabled.
19084 2018-01-23 Martin Liska <mliska@suse.cz>
19086 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19087 order to ignore the predictor.
19088 (PRED_POLYMORPHIC_CALL): Likewise.
19089 (PRED_RECURSIVE_CALL): Likewise.
19091 2018-01-23 Martin Liska <mliska@suse.cz>
19093 * tree-profile.c (tree_profiling): Print function header to
19094 aware reader which function we are working on.
19095 * value-prof.c (gimple_find_values_to_profile): Do not print
19096 not interesting value histograms.
19098 2018-01-23 Martin Liska <mliska@suse.cz>
19100 * profile-count.h (enum profile_quality): Add
19101 profile_uninitialized as the first value. Do not number values
19102 as they are zero based.
19103 (profile_count::verify): Update sanity check.
19104 (profile_probability::verify): Likewise.
19106 2018-01-23 Nathan Sidwell <nathan@acm.org>
19108 * doc/invoke.texi (ffor-scope): Deprecate.
19110 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19112 PR tree-optimization/83510
19113 * domwalk.c (set_all_edges_as_executable): New function.
19114 (dom_walker::dom_walker): Convert bool param
19115 "skip_unreachable_blocks" to enum reachability. Move setup of
19116 edge flags to set_all_edges_as_executable and only do it when
19117 reachability is REACHABLE_BLOCKS.
19118 * domwalk.h (enum dom_walker::reachability): New enum.
19119 (dom_walker::dom_walker): Convert bool param
19120 "skip_unreachable_blocks" to enum reachability.
19121 (set_all_edges_as_executable): New decl.
19122 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19123 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19125 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19126 but converting true to REACHABLE_BLOCKS.
19127 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19129 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19130 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19131 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19133 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19134 if check_all_array_refs will be called.
19136 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19138 * tree.c (selftest::test_location_wrappers): Add more test
19141 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19143 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19144 (selftest::test_bit_in_range): Likewise.
19146 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19149 * doc/sourcebuild.texi (vect_float): Say that the selector
19150 only describes the situation when -funsafe-math-optimizations is on.
19151 (vect_float_strict): Document.
19153 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19155 PR tree-optimization/83965
19156 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19157 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19158 instead of checking only for a reduction.
19159 (vect_recog_widen_sum_pattern): Likewise.
19161 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19163 * predict.c (probably_never_executed): Only use precise profile info.
19164 (compute_function_frequency): Skip after inlining hack since we now
19165 have quality checking.
19167 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19169 * profile-count.h (profile_probability::very_unlikely,
19170 profile_probability::unlikely, profile_probability::even): Set
19171 precision to guessed.
19173 2018-01-23 Richard Biener <rguenther@suse.de>
19175 PR tree-optimization/83963
19176 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19177 Properly terminate dominator walk when crossing the exit edge not
19178 when visiting its source block.
19180 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19183 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19184 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19186 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19188 PR tree-optimization/83957
19189 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
19190 semicolon after for body surrounded by braces.
19192 PR tree-optimization/83081
19193 * profile-count.h (profile_probability::split): New method.
19194 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
19195 Use profile_probability::split.
19196 (do_compare_rtx_and_jump): Fix adjustment of probabilities
19197 when splitting a single conditional jump into 2.
19199 2018-01-22 David Malcolm <dmalcolm@redhat.com>
19201 PR tree-optimization/69452
19202 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
19205 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19207 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
19208 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
19209 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
19211 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19213 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
19215 * config/rl78/rl78.md (movdi): New define_expand.
19216 * config/rl78/rl78.c (rl78_split_movdi): New function.
19218 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
19221 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
19223 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
19224 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
19225 128-bit to produce an UNSPEC move to get the double word with the
19226 signbit and then a shift directly to do signbit.
19227 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
19228 implementation with a new version that just does either a direct
19229 move or a regular move. Move memory interface to separate insns.
19230 Move insns so they are next to the expander.
19231 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
19232 with signbit move. Split big and little endian case.
19233 (signbit<mode>2_dm_mem_le): Likewise.
19234 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
19235 (signbit<mode>2_dm2): Likewise.
19237 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19239 * config/rl78/rl78.md (anddi3): New define_expand.
19241 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19243 * config/rl78/rl78.md (umindi3): New define_expand.
19245 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19247 * config/rl78/rl78.md (smindi3): New define_expand.
19249 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19251 * config/rl78/rl78.md (smaxdi3): New define_expand.
19253 2018-01-22 Carl Love <cel@us.ibm.com>
19255 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
19256 LVX_V1TI): Add macro expansion.
19257 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
19258 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
19259 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
19260 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
19261 Change check to determine if the instruction is a byte reversing
19262 entry. Fix typo in comment.
19263 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
19264 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
19265 Add def_builtin calls for new builtins.
19266 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
19267 Add define_insn expansion.
19269 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19271 * config/rl78/rl78.md (umaxdi3): New define_expand.
19273 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19275 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
19276 for non-QImode registers.
19278 2018-01-22 Richard Biener <rguenther@suse.de>
19280 PR tree-optimization/83963
19281 * graphite-scop-detection.c (scop_detection::get_sese): Delay
19282 including the loop exit block.
19283 (scop_detection::merge_sese): Likewise.
19284 (scop_detection::add_scop): Do it here instead.
19286 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19288 * doc/sourcebuild.texi (arm_softfloat): Document.
19290 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
19293 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
19294 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
19295 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
19297 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19298 David Edelsohn <dje.gcc@gmail.com>
19301 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19302 Change "crset eq" to "crset 2".
19303 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19304 (*call_indirect_aix<mode>_nospec): Likewise.
19305 (*call_value_indirect_aix<mode>_nospec): Likewise.
19306 (*call_indirect_elfv2<mode>_nospec): Likewise.
19307 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
19308 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
19309 change assembly output from . to $.
19310 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19311 (indirect_jump<mode>_nospec): Change assembly output from . to $.
19312 (*tablejump<mode>_internal1_nospec): Likewise.
19314 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
19317 * config/sh/sh_optimize_sett_clrt.cc:
19318 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
19320 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
19322 PR tree-optimization/83940
19323 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
19324 offset_dt to vect_constant_def rather than vect_unknown_def_type.
19325 (vect_check_load_store_mask): Add a mask_dt_out parameter and
19326 use it to pass back the definition type.
19327 (vect_check_store_rhs): Likewise rhs_dt_out.
19328 (vect_build_gather_load_calls): Add a mask_dt argument and use
19329 it instead of a call to vect_is_simple_use.
19330 (vectorizable_store): Update calls to vect_check_load_store_mask
19331 and vect_check_store_rhs. Use the dt returned by the latter instead
19332 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
19333 instead of calls to vect_is_simple_use. Pass the scalar rather
19334 than the vector operand to vect_is_simple_use when handling
19335 second and subsequent copies of an rhs value.
19336 (vectorizable_load): Update calls to vect_check_load_store_mask
19337 and vect_build_gather_load_calls. Use the cached mask_dt and
19338 gs_info.offset_dt instead of calls to vect_is_simple_use.
19340 2018-01-20 Jakub Jelinek <jakub@redhat.com>
19342 PR middle-end/83945
19343 * tree-emutls.c: Include gimplify.h.
19344 (lower_emutls_2): New function.
19345 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
19346 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
19347 it before further processing.
19350 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
19351 UINTVAL (trueop1) instead of INTVAL (op1).
19353 2018-01-19 Jakub Jelinek <jakub@redhat.com>
19357 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
19358 INCOMING_FRAME_SP_OFFSET if not defined.
19359 (scan_trace): Add ENTRY argument. If true and
19360 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
19361 emit a note to adjust the CFA offset.
19362 (create_cfi_notes): Adjust scan_trace callers.
19363 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
19364 INCOMING_FRAME_SP_OFFSET in the CIE.
19365 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
19366 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
19368 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
19369 * doc/tm.texi: Regenerated.
19371 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19373 PR rtl-optimization/83147
19374 * lra-constraints.c (remove_inheritance_pseudos): Use
19375 lra_substitute_pseudo_within_insn.
19377 2018-01-19 Tom de Vries <tom@codesourcery.com>
19378 Cesar Philippidis <cesar@codesourcery.com>
19381 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
19383 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
19386 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
19387 spaces for function labels.
19389 2018-01-19 Martin Liska <mliska@suse.cz>
19391 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
19392 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
19393 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
19394 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
19395 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
19396 (PRED_CONST_RETURN): Change from 69 to 65.
19397 (PRED_NULL_RETURN): Change from 91 to 71.
19398 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
19399 (PRED_LOOP_GUARD): Change from 66 to 73.
19401 2018-01-19 Martin Liska <mliska@suse.cz>
19403 * predict.c (predict_insn_def): Add new assert.
19404 (struct branch_predictor): Change type to signed integer.
19405 (test_prediction_value_range): Amend test to cover
19406 PROB_UNINITIALIZED.
19407 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
19408 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
19409 (PRED_LOOP_ITERATIONS_MAX): Likewise.
19410 (PRED_LOOP_IV_COMPARE): Likewise.
19411 * predict.h (PROB_UNINITIALIZED): Define new constant.
19413 2018-01-19 Martin Liska <mliska@suse.cz>
19415 * predict.c (dump_prediction): Add new format for
19416 analyze_brprob.py script which is enabled with -details
19418 * profile-count.h (precise_p): New function.
19420 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19422 PR tree-optimization/83922
19423 * tree-vect-loop.c (vect_verify_full_masking): Return false if
19424 there are no statements that need masking.
19425 (vect_active_double_reduction_p): New function.
19426 (vect_analyze_loop_operations): Use it when handling phis that
19427 are not in the loop header.
19429 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19431 PR tree-optimization/83914
19432 * tree-vect-loop.c (vectorizable_induction): Don't convert
19433 init_expr or apply the peeling adjustment for inductions
19434 that are nested within the vectorized loop.
19436 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19438 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
19441 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19445 * function.h (gimplify_parameters): Add gimple_seq * argument.
19446 * function.c: Include gimple.h and options.h.
19447 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
19448 for the added local temporaries if needed.
19449 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
19450 if there are any parameter cleanups, wrap whole body into a
19451 try/finally with the cleanups.
19453 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
19456 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
19457 Use GET_MODE_CLASS for scalar floating point.
19459 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19463 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
19464 Fix call of call_cgraph_insertion_hooks.
19466 2018-01-18 Martin Sebor <msebor@redhat.com>
19468 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
19470 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19473 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
19476 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
19479 * common.opt: (-ffile-prefix-map): New option.
19480 * opts.c (common_handle_option): Defer it.
19481 * opts-global.c (handle_common_deferred_options): Handle it.
19482 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
19483 * file-prefix-map.h: New file.
19484 (remap_debug_filename, add_debug_prefix_map): ...here.
19485 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
19486 * final.c (debug_prefix_map, add_debug_prefix_map
19487 remap_debug_filename): Move to...
19488 * file-prefix-map.c: New file.
19489 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
19490 generalize, get rid of alloca(), use strrchr() instead of strchr().
19491 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
19492 Implement in terms of add_prefix_map().
19493 (remap_macro_filename, remap_debug_filename): Implement in term of
19495 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
19496 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
19497 * dbxout.c: Include file-prefix-map.h.
19498 * varasm.c: Likewise.
19499 * vmsdbgout.c: Likewise.
19500 * xcoffout.c: Likewise.
19501 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
19502 * doc/cppopts.texi (-fmacro-prefix-map): Document.
19503 * doc/invoke.texi (-ffile-prefix-map): Document.
19504 (-fdebug-prefix-map): Update description.
19506 2018-01-18 Martin Liska <mliska@suse.cz>
19508 * config/i386/i386.c (indirect_thunk_name): Document that also
19510 (output_indirect_thunk): Document why both instructions
19511 (pause and lfence) are generated.
19513 2018-01-18 Richard Biener <rguenther@suse.de>
19515 PR tree-optimization/83887
19516 * graphite-scop-detection.c
19517 (scop_detection::get_nearest_dom_with_single_entry): Remove.
19518 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
19519 (scop_detection::merge_sese): Re-implement with a flood-fill
19520 algorithm that properly finds a SESE region if it exists.
19522 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19525 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
19526 pointer_diff optimizations use view_convert instead of convert.
19528 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19530 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19531 Generate different code for -mno-speculate-indirect-jumps.
19532 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19533 (*call_indirect_aix<mode>): Disable for
19534 -mno-speculate-indirect-jumps.
19535 (*call_indirect_aix<mode>_nospec): New define_insn.
19536 (*call_value_indirect_aix<mode>): Disable for
19537 -mno-speculate-indirect-jumps.
19538 (*call_value_indirect_aix<mode>_nospec): New define_insn.
19539 (*sibcall_nonlocal_sysv<mode>): Generate different code for
19540 -mno-speculate-indirect-jumps.
19541 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19543 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
19545 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
19546 long double type, set the flags for noting the default long double
19547 type, even if we don't pass or return a long double type.
19549 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
19552 * ipa-inline.c (flatten_function): Do not overwrite final inlining
19555 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
19557 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
19558 support for merge[hl].
19559 (fold_mergehl_helper): New helper function.
19560 (tree-vector-builder.h): New #include for tree_vector_builder usage.
19561 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
19562 (altivec_vmrglw_direct): Add xxmrglw insn.
19564 2018-01-17 Andrew Waterman <andrew@sifive.com>
19566 * config/riscv/riscv.c (riscv_conditional_register_usage): If
19567 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
19569 2018-01-17 David Malcolm <dmalcolm@redhat.com>
19572 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
19573 call the lto_location_cache before reading the
19574 DECL_SOURCE_LOCATION of the types.
19576 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
19577 Richard Sandiford <richard.sandiford@linaro.org>
19579 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
19580 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
19581 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
19582 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
19583 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
19585 * config/aarch64/constraints.md (aarch64_movti_operand):
19587 * config/aarch64/predicates.md (Uti): Add new constraint.
19589 2018-01-17 Carl Love <cel@us.ibm.com>
19591 * config/rs6000/vsx.md (define_expand xl_len_r,
19592 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
19593 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
19595 (define_expand, define_insn): Move the shift left from the
19596 define_insn to the define_expand for lxvl and stxvl instructions.
19597 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
19598 and XL_LEN_R definitions to PURE.
19600 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
19602 * config/i386/i386.c (indirect_thunk_name): Declare regno
19603 as unsigned int. Compare regno with INVALID_REGNUM.
19604 (output_indirect_thunk): Ditto.
19605 (output_indirect_thunk_function): Ditto.
19606 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
19607 in the call to output_indirect_thunk_function.
19609 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
19611 PR middle-end/83884
19612 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
19613 rather than the size of inner_type to determine the stack slot size
19614 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
19616 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
19619 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
19622 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
19624 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
19625 endian Linux systems to optionally enable multilibs for selecting
19626 the long double type if the user configured an explicit type.
19627 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
19628 have no long double multilibs if not defined.
19629 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
19630 warn if the user used -mabi={ieee,ibm}longdouble and we built
19631 multilibs for long double.
19632 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
19633 appropriate multilib option.
19634 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
19636 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
19637 for building long double multilibs.
19638 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
19640 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
19642 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
19645 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
19647 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
19650 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
19653 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
19656 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19658 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
19659 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
19661 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
19663 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
19664 different rtl trees depending on TARGET_64BIT.
19665 (rs6000_gen_lvx): Likewise.
19667 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19669 * config/visium/visium.md (nop): Tweak comment.
19670 (hazard_nop): Likewise.
19672 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19674 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
19675 -mspeculate-indirect-jumps.
19676 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
19677 for -mno-speculate-indirect-jumps.
19678 (*call_indirect_elfv2<mode>_nospec): New define_insn.
19679 (*call_value_indirect_elfv2<mode>): Disable for
19680 -mno-speculate-indirect-jumps.
19681 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
19682 (indirect_jump): Emit different RTL for
19683 -mno-speculate-indirect-jumps.
19684 (*indirect_jump<mode>): Disable for
19685 -mno-speculate-indirect-jumps.
19686 (*indirect_jump<mode>_nospec): New define_insn.
19687 (tablejump): Emit different RTL for
19688 -mno-speculate-indirect-jumps.
19689 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
19690 (tablejumpsi_nospec): New define_expand.
19691 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
19692 (tablejumpdi_nospec): New define_expand.
19693 (*tablejump<mode>_internal1): Disable for
19694 -mno-speculate-indirect-jumps.
19695 (*tablejump<mode>_internal1_nospec): New define_insn.
19696 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
19699 2018-01-16 Artyom Skrobov tyomitch@gmail.com
19701 * caller-save.c (insert_save): Drop unnecessary parameter. All
19704 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19705 Richard Biener <rguenth@suse.de>
19708 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
19709 return early, inline manually is_gimple_sizepos. Make sure if we
19710 call gimplify_expr we don't end up with a gimple constant.
19711 * tree.c (variably_modified_type_p): Don't return true for
19712 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
19713 * gimplify.h (is_gimple_sizepos): Remove.
19715 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19717 PR tree-optimization/83857
19718 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
19719 vectorizable_live_operation for pure SLP statements.
19720 (vectorizable_live_operation): Handle PHIs.
19722 2018-01-16 Richard Biener <rguenther@suse.de>
19724 PR tree-optimization/83867
19725 * tree-vect-stmts.c (vect_transform_stmt): Precompute
19726 nested_in_vect_loop_p since the scalar stmt may get invalidated.
19728 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19731 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
19732 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
19733 If off is not INTEGER_CST, issue a may not be aligned warning
19734 rather than isn't aligned. Use isn%'t rather than isn't.
19735 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
19737 <case MULT_EXPR>: Improve the case when bottom and one of the
19738 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
19739 operand, in that case check if the other operand is multiple of
19740 bottom divided by the INTEGER_CST operand.
19742 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19745 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
19746 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
19747 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
19748 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
19749 * config/pa/pa.c (pa_function_arg_advance): Likewise.
19750 (pa_function_arg, pa_arg_partial_bytes): Likewise.
19751 (pa_function_arg_size): New function.
19753 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19755 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
19756 in a separate statement.
19758 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19760 PR tree-optimization/83847
19761 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
19762 group gathers and scatters.
19764 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19766 PR rtl-optimization/86620
19767 * params.def (max-sched-ready-insns): Bump minimum value to 1.
19769 PR rtl-optimization/83213
19770 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
19771 to last if both are JUMP_INSNs.
19773 PR tree-optimization/83843
19774 * gimple-ssa-store-merging.c
19775 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
19776 store_immediate_info for bswap/nop orig_stores.
19778 2018-01-15 Andrew Waterman <andrew@sifive.com>
19780 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
19782 <UDIV>: Increase cost if !TARGET_DIV.
19784 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
19786 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
19787 (define_attr "cr_logical_3op"): New.
19788 (cceq_ior_compare): Adjust.
19789 (cceq_ior_compare_complement): Adjust.
19790 (*cceq_rev_compare): Adjust.
19791 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
19792 (is_cracked_insn): Adjust.
19793 (insn_must_be_first_in_group): Adjust.
19794 * config/rs6000/40x.md: Adjust.
19795 * config/rs6000/440.md: Adjust.
19796 * config/rs6000/476.md: Adjust.
19797 * config/rs6000/601.md: Adjust.
19798 * config/rs6000/603.md: Adjust.
19799 * config/rs6000/6xx.md: Adjust.
19800 * config/rs6000/7450.md: Adjust.
19801 * config/rs6000/7xx.md: Adjust.
19802 * config/rs6000/8540.md: Adjust.
19803 * config/rs6000/cell.md: Adjust.
19804 * config/rs6000/e300c2c3.md: Adjust.
19805 * config/rs6000/e500mc.md: Adjust.
19806 * config/rs6000/e500mc64.md: Adjust.
19807 * config/rs6000/e5500.md: Adjust.
19808 * config/rs6000/e6500.md: Adjust.
19809 * config/rs6000/mpc.md: Adjust.
19810 * config/rs6000/power4.md: Adjust.
19811 * config/rs6000/power5.md: Adjust.
19812 * config/rs6000/power6.md: Adjust.
19813 * config/rs6000/power7.md: Adjust.
19814 * config/rs6000/power8.md: Adjust.
19815 * config/rs6000/power9.md: Adjust.
19816 * config/rs6000/rs64.md: Adjust.
19817 * config/rs6000/titan.md: Adjust.
19819 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19821 * config/i386/predicates.md (indirect_branch_operand): Rewrite
19822 ix86_indirect_branch_register logic.
19824 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19826 * config/i386/constraints.md (Bs): Update
19827 ix86_indirect_branch_register check. Don't check
19828 ix86_indirect_branch_register with GOT_memory_operand.
19830 * config/i386/predicates.md (GOT_memory_operand): Don't check
19831 ix86_indirect_branch_register here.
19832 (GOT32_symbol_operand): Likewise.
19834 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19836 * config/i386/predicates.md (constant_call_address_operand):
19837 Rewrite ix86_indirect_branch_register logic.
19838 (sibcall_insn_operand): Likewise.
19840 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19842 * config/i386/constraints.md (Bs): Replace
19843 ix86_indirect_branch_thunk_register with
19844 ix86_indirect_branch_register.
19846 * config/i386/i386.md (indirect_jump): Likewise.
19847 (tablejump): Likewise.
19848 (*sibcall_memory): Likewise.
19849 (*sibcall_value_memory): Likewise.
19850 Peepholes of indirect call and jump via memory: Likewise.
19851 * config/i386/i386.opt: Likewise.
19852 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19853 (GOT_memory_operand): Likewise.
19854 (call_insn_operand): Likewise.
19855 (sibcall_insn_operand): Likewise.
19856 (GOT32_symbol_operand): Likewise.
19858 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19860 PR middle-end/83837
19861 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
19862 type rather than type addr's type points to.
19863 (expand_omp_atomic_mutex): Likewise.
19864 (expand_omp_atomic): Likewise.
19866 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19869 * config/i386/i386.c (output_indirect_thunk_function): Use
19870 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
19871 for __x86_return_thunk.
19873 2018-01-15 Richard Biener <rguenther@suse.de>
19875 PR middle-end/83850
19876 * expmed.c (extract_bit_field_1): Fix typo.
19878 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19881 * config/arm/iterators.md (VF): New mode iterator.
19882 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
19883 Remove integer-related logic from pattern.
19884 (neon_vabd<mode>_3): Likewise.
19886 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19888 PR middle-end/82694
19889 * common.opt (fstrict-overflow): No longer an alias.
19890 (fwrapv-pointer): New option.
19891 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
19892 also for pointer types based on flag_wrapv_pointer.
19893 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
19894 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
19895 opts->x_flag_wrapv got set.
19896 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
19897 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
19898 POINTER_TYPE_OVERFLOW_UNDEFINED.
19899 * match.pd: Likewise in address comparison pattern.
19900 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
19902 2018-01-15 Richard Biener <rguenther@suse.de>
19905 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
19906 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
19907 Reset type names to their identifier if their TYPE_DECL doesn't
19908 have linkage (and thus is used for ODR and devirt).
19909 (save_debug_info_for_decl): Remove.
19910 (save_debug_info_for_type): Likewise.
19911 (add_tree_to_fld_list): Adjust.
19912 * tree-pretty-print.c (dump_generic_node): Make dumping of
19913 type names more robust.
19915 2018-01-15 Richard Biener <rguenther@suse.de>
19917 * BASE-VER: Bump to 8.0.1.
19919 2018-01-14 Martin Sebor <msebor@redhat.com>
19922 * builtins.c (check_access): Avoid warning when the no-warning bit
19925 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
19927 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
19928 * ira-color (allocno_hard_regs_compare): Likewise.
19930 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
19933 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
19934 Use .pushsection/.popsection.
19936 2018-01-14 Martin Sebor <msebor@redhat.com>
19939 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
19941 2018-01-14 Jakub Jelinek <jakub@redhat.com>
19943 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
19944 entry from extra_headers.
19945 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
19946 extra_headers, make the list bitwise identical to the i?86-*-* one.
19948 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19950 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
19951 -mcmodel=large with -mindirect-branch=thunk,
19952 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19953 -mfunction-return=thunk-extern.
19954 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19955 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19956 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19958 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19960 * config/i386/i386.c (print_reg): Print the name of the full
19961 integer register without '%'.
19962 (ix86_print_operand): Handle 'V'.
19963 * doc/extend.texi: Document 'V' modifier.
19965 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19967 * config/i386/constraints.md (Bs): Disallow memory operand for
19968 -mindirect-branch-register.
19970 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19971 (GOT_memory_operand): Likewise.
19972 (call_insn_operand): Likewise.
19973 (sibcall_insn_operand): Likewise.
19974 (GOT32_symbol_operand): Likewise.
19975 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19976 for -mindirect-branch-register.
19977 (tablejump): Likewise.
19978 (*sibcall_memory): Likewise.
19979 (*sibcall_value_memory): Likewise.
19980 Disallow peepholes of indirect call and jump via memory for
19981 -mindirect-branch-register.
19982 (*call_pop): Replace m with Bw.
19983 (*call_value_pop): Likewise.
19984 (*sibcall_pop_memory): Replace m with Bs.
19985 * config/i386/i386.opt (mindirect-branch-register): New option.
19986 * doc/invoke.texi: Document -mindirect-branch-register option.
19988 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19990 * config/i386/i386-protos.h (ix86_output_function_return): New.
19991 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19992 set function_return_type.
19993 (indirect_thunk_name): Add ret_p to indicate thunk for function
19995 (output_indirect_thunk_function): Pass false to
19996 indirect_thunk_name.
19997 (ix86_output_indirect_branch_via_reg): Likewise.
19998 (ix86_output_indirect_branch_via_push): Likewise.
19999 (output_indirect_thunk_function): Create alias for function
20000 return thunk if regno < 0.
20001 (ix86_output_function_return): New function.
20002 (ix86_handle_fndecl_attribute): Handle function_return.
20003 (ix86_attribute_table): Add function_return.
20004 * config/i386/i386.h (machine_function): Add
20005 function_return_type.
20006 * config/i386/i386.md (simple_return_internal): Use
20007 ix86_output_function_return.
20008 (simple_return_internal_long): Likewise.
20009 * config/i386/i386.opt (mfunction-return=): New option.
20010 (indirect_branch): Mention -mfunction-return=.
20011 * doc/extend.texi: Document function_return function attribute.
20012 * doc/invoke.texi: Document -mfunction-return= option.
20014 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20016 * config/i386/i386-opts.h (indirect_branch): New.
20017 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20018 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20019 with local indirect jump when converting indirect call and jump.
20020 (ix86_set_indirect_branch_type): New.
20021 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20022 (indirectlabelno): New.
20023 (indirect_thunk_needed): Likewise.
20024 (indirect_thunk_bnd_needed): Likewise.
20025 (indirect_thunks_used): Likewise.
20026 (indirect_thunks_bnd_used): Likewise.
20027 (INDIRECT_LABEL): Likewise.
20028 (indirect_thunk_name): Likewise.
20029 (output_indirect_thunk): Likewise.
20030 (output_indirect_thunk_function): Likewise.
20031 (ix86_output_indirect_branch_via_reg): Likewise.
20032 (ix86_output_indirect_branch_via_push): Likewise.
20033 (ix86_output_indirect_branch): Likewise.
20034 (ix86_output_indirect_jmp): Likewise.
20035 (ix86_code_end): Call output_indirect_thunk_function if needed.
20036 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20038 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20039 (ix86_attribute_table): Add indirect_branch.
20040 * config/i386/i386.h (machine_function): Add indirect_branch_type
20041 and has_local_indirect_jump.
20042 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20044 (tablejump): Likewise.
20045 (*indirect_jump): Use ix86_output_indirect_jmp.
20046 (*tablejump_1): Likewise.
20047 (simple_return_indirect_internal): Likewise.
20048 * config/i386/i386.opt (mindirect-branch=): New option.
20049 (indirect_branch): New.
20052 (thunk-inline): Likewise.
20053 (thunk-extern): Likewise.
20054 * doc/extend.texi: Document indirect_branch function attribute.
20055 * doc/invoke.texi: Document -mindirect-branch= option.
20057 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20060 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20062 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20064 * ipa-inline.c (want_inline_small_function_p): Return false if
20065 inlining has already failed with CIF_FINAL_ERROR.
20066 (update_caller_keys): Call want_inline_small_function_p before
20068 (update_callee_keys): Likewise.
20070 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20072 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20074 (rs6000_quadword_masked_address_p): Likewise.
20075 (quad_aligned_load_p): Likewise.
20076 (quad_aligned_store_p): Likewise.
20077 (const_load_sequence_p): Add comment to describe the outer-most loop.
20078 (mimic_memory_attributes_and_flags): New function.
20079 (rs6000_gen_stvx): Likewise.
20080 (replace_swapped_aligned_store): Likewise.
20081 (rs6000_gen_lvx): Likewise.
20082 (replace_swapped_aligned_load): Likewise.
20083 (replace_swapped_load_constant): Capitalize argument name in
20084 comment describing this function.
20085 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20086 and stores that access quad-word aligned addresses and replace
20087 with stvx or lvx instructions when appropriate.
20088 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20089 New function prototype.
20090 (rs6000_quadword_masked_address_p): Likewise.
20091 (rs6000_gen_lvx): Likewise.
20092 (rs6000_gen_stvx): Likewise.
20093 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20094 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20095 when memory address is aligned.
20096 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20097 this split to select lvx instruction when memory address is aligned.
20098 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20099 instruction when memory address is aligned.
20100 (*vsx_le_perm_load_v16qi): Likewise.
20101 (four unnamed splitters): Modify to select the stvx instruction
20102 when memory is aligned.
20104 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20106 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20107 which appears in the queue multiple times.
20109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20110 Alan Hayward <alan.hayward@arm.com>
20111 David Sherwood <david.sherwood@arm.com>
20113 * tree-vectorizer.h (vec_lower_bound): New structure.
20114 (_loop_vec_info): Add check_nonzero and lower_bounds.
20115 (LOOP_VINFO_CHECK_NONZERO): New macro.
20116 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20117 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20118 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20119 fields. Make seg_len the distance travelled, not including the
20121 (dr_direction_indicator): Declare.
20122 (dr_zero_step_indicator): Likewise.
20123 (dr_known_forward_stride_p): Likewise.
20124 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20126 (runtime_alias_check_p): Allow runtime alias checks with
20128 (operator ==): Compare access_size and align.
20129 (prune_runtime_alias_test_list): Rework for new distinction between
20130 the access_size and seg_len.
20131 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20133 (get_segment_min_max): New function.
20134 (create_intersect_range_checks): Use it.
20135 (dr_step_indicator): New function.
20136 (dr_direction_indicator): Likewise.
20137 (dr_zero_step_indicator): Likewise.
20138 (dr_known_forward_stride_p): Likewise.
20139 * tree-loop-distribution.c (data_ref_segment_size): Return
20140 DR_STEP * (niters - 1).
20141 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20143 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20144 (vect_preserves_scalar_order_p): New function, split out from...
20145 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20146 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20147 (vect_vfa_access_size): New function.
20148 (vect_vfa_align): Likewise.
20149 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20150 (dump_lower_bound): New function.
20151 (vect_check_lower_bound): Likewise.
20152 (vect_small_gap_p): Likewise.
20153 (vectorizable_with_step_bound_p): Likewise.
20154 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20155 depencies if the vectorization factor is 1. Convert the checks
20156 for nonzero steps into checks on the bounds of DR_STEP. Try using
20157 a bunds check for variable steps if the minimum required step is
20158 relatively small. Update calls to the dr_with_seg_len
20159 constructor and to vect_compile_time_alias.
20160 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20162 (vect_loop_versioning): Call it.
20163 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20165 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20168 Alan Hayward <alan.hayward@arm.com>
20169 David Sherwood <david.sherwood@arm.com>
20171 * doc/sourcebuild.texi (vect_scatter_store): Document.
20172 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20174 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20176 * genopinit.c (main): Add supports_vec_scatter_store and
20177 supports_vec_scatter_store_cached to target_optabs.
20178 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20179 IFN_MASK_SCATTER_STORE.
20180 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20182 * internal-fn.h (internal_store_fn_p): Declare.
20183 (internal_fn_stored_value_index): Likewise.
20184 * internal-fn.c (scatter_store_direct): New macro.
20185 (expand_scatter_store_optab_fn): New function.
20186 (direct_scatter_store_optab_supported_p): New macro.
20187 (internal_store_fn_p): New function.
20188 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
20189 IFN_MASK_SCATTER_STORE.
20190 (internal_fn_mask_index): Likewise.
20191 (internal_fn_stored_value_index): New function.
20192 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
20193 for scatter stores.
20194 * optabs-query.h (supports_vec_scatter_store_p): Declare.
20195 * optabs-query.c (supports_vec_scatter_store_p): New function.
20196 * tree-vectorizer.h (vect_get_store_rhs): Declare.
20197 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
20198 true for scatter stores.
20199 (vect_gather_scatter_fn_p): Handle scatter stores too.
20200 (vect_check_gather_scatter): Consider using scatter stores if
20201 supports_vec_scatter_store_p.
20202 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
20203 scatter stores too.
20204 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20205 internal_fn_stored_value_index.
20206 (check_load_store_masking): Handle scatter stores too.
20207 (vect_get_store_rhs): Make public.
20208 (vectorizable_call): Use internal_store_fn_p.
20209 (vectorizable_store): Handle scatter store internal functions.
20210 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
20211 when deciding whether the end of the group has been reached.
20212 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
20213 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
20214 (mask_scatter_store<mode>): New insns.
20216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20217 Alan Hayward <alan.hayward@arm.com>
20218 David Sherwood <david.sherwood@arm.com>
20220 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
20221 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
20222 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
20224 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
20225 Use vect_truncate_gather_scatter_offset if we can't treat the
20226 operation as a normal gather load or scatter store.
20227 (get_group_load_store_type): Take the gather_scatter_info
20228 as argument. Try using a gather load or scatter store for
20229 single-element groups.
20230 (get_load_store_type): Update calls to get_group_load_store_type
20231 and vect_use_strided_gather_scatters_p.
20233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20234 Alan Hayward <alan.hayward@arm.com>
20235 David Sherwood <david.sherwood@arm.com>
20237 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
20238 optional tree argument.
20239 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
20241 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
20242 but continue to use the current value as a fallback.
20243 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
20244 to compare the updates.
20245 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
20246 (get_load_store_type): Use it when handling a strided access.
20247 (vect_get_strided_load_store_ops): New function.
20248 (vect_get_data_ptr_increment): Likewise.
20249 (vectorizable_load): Handle strided gather loads. Always pass
20250 a step to vect_create_data_ref_ptr and bump_vector_ptr.
20252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20253 Alan Hayward <alan.hayward@arm.com>
20254 David Sherwood <david.sherwood@arm.com>
20256 * doc/md.texi (gather_load@var{m}): Document.
20257 (mask_gather_load@var{m}): Likewise.
20258 * genopinit.c (main): Add supports_vec_gather_load and
20259 supports_vec_gather_load_cached to target_optabs.
20260 * optabs-tree.c (init_tree_optimization_optabs): Use
20261 ggc_cleared_alloc to allocate target_optabs.
20262 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
20263 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
20265 * internal-fn.h (internal_load_fn_p): Declare.
20266 (internal_gather_scatter_fn_p): Likewise.
20267 (internal_fn_mask_index): Likewise.
20268 (internal_gather_scatter_fn_supported_p): Likewise.
20269 * internal-fn.c (gather_load_direct): New macro.
20270 (expand_gather_load_optab_fn): New function.
20271 (direct_gather_load_optab_supported_p): New macro.
20272 (direct_internal_fn_optab): New function.
20273 (internal_load_fn_p): Likewise.
20274 (internal_gather_scatter_fn_p): Likewise.
20275 (internal_fn_mask_index): Likewise.
20276 (internal_gather_scatter_fn_supported_p): Likewise.
20277 * optabs-query.c (supports_at_least_one_mode_p): New function.
20278 (supports_vec_gather_load_p): Likewise.
20279 * optabs-query.h (supports_vec_gather_load_p): Declare.
20280 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
20281 and memory_type field.
20282 (NUM_PATTERNS): Bump to 15.
20283 * tree-vect-data-refs.c: Include internal-fn.h.
20284 (vect_gather_scatter_fn_p): New function.
20285 (vect_describe_gather_scatter_call): Likewise.
20286 (vect_check_gather_scatter): Try using internal functions for
20287 gather loads. Recognize existing calls to a gather load function.
20288 (vect_analyze_data_refs): Consider using gather loads if
20289 supports_vec_gather_load_p.
20290 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
20291 (vect_get_gather_scatter_offset_type): Likewise.
20292 (vect_convert_mask_for_vectype): Likewise.
20293 (vect_add_conversion_to_patterm): Likewise.
20294 (vect_try_gather_scatter_pattern): Likewise.
20295 (vect_recog_gather_scatter_pattern): New pattern recognizer.
20296 (vect_vect_recog_func_ptrs): Add it.
20297 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20298 internal_fn_mask_index and internal_gather_scatter_fn_p.
20299 (check_load_store_masking): Take the gather_scatter_info as an
20300 argument and handle gather loads.
20301 (vect_get_gather_scatter_ops): New function.
20302 (vectorizable_call): Check internal_load_fn_p.
20303 (vectorizable_load): Likewise. Handle gather load internal
20305 (vectorizable_store): Update call to check_load_store_masking.
20306 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
20307 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
20308 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
20309 (aarch64_gather_scale_operand_d): New predicates.
20310 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
20311 (mask_gather_load<mode>): New insns.
20313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20314 Alan Hayward <alan.hayward@arm.com>
20315 David Sherwood <david.sherwood@arm.com>
20317 * optabs.def (fold_left_plus_optab): New optab.
20318 * doc/md.texi (fold_left_plus_@var{m}): Document.
20319 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
20320 * internal-fn.c (fold_left_direct): Define.
20321 (expand_fold_left_optab_fn): Likewise.
20322 (direct_fold_left_optab_supported_p): Likewise.
20323 * fold-const-call.c (fold_const_fold_left): New function.
20324 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
20325 * tree-parloops.c (valid_reduction_p): New function.
20326 (gather_scalar_reductions): Use it.
20327 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
20328 (vect_finish_replace_stmt): Declare.
20329 * tree-vect-loop.c (fold_left_reduction_fn): New function.
20330 (needs_fold_left_reduction_p): New function, split out from...
20331 (vect_is_simple_reduction): ...here. Accept reductions that
20332 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
20333 (vect_force_simple_reduction): Also store the reduction type in
20334 the assignment's STMT_VINFO_REDUC_TYPE.
20335 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
20336 (merge_with_identity): New function.
20337 (vect_expand_fold_left): Likewise.
20338 (vectorize_fold_left_reduction): Likewise.
20339 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
20340 scalar phi in place for it. Check for target support and reject
20341 cases that would reassociate the operation. Defer the transform
20342 phase to vectorize_fold_left_reduction.
20343 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
20344 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
20345 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
20347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20349 * tree-if-conv.c (predicate_mem_writes): Remove redundant
20350 call to ifc_temp_var.
20352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20353 Alan Hayward <alan.hayward@arm.com>
20354 David Sherwood <david.sherwood@arm.com>
20356 * target.def (legitimize_address_displacement): Take the original
20357 offset as a poly_int.
20358 * targhooks.h (default_legitimize_address_displacement): Update
20360 * targhooks.c (default_legitimize_address_displacement): Likewise.
20361 * doc/tm.texi: Regenerate.
20362 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
20363 as an argument, moving assert of ad->disp == ad->disp_term to...
20364 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
20365 Try calling targetm.legitimize_address_displacement before expanding
20366 the address rather than afterwards, and adjust for the new interface.
20367 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
20368 Match the new hook interface. Handle SVE addresses.
20369 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
20370 new hook interface.
20372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20374 * Makefile.in (OBJS): Add early-remat.o.
20375 * target.def (select_early_remat_modes): New hook.
20376 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
20377 * doc/tm.texi: Regenerate.
20378 * targhooks.h (default_select_early_remat_modes): Declare.
20379 * targhooks.c (default_select_early_remat_modes): New function.
20380 * timevar.def (TV_EARLY_REMAT): New timevar.
20381 * passes.def (pass_early_remat): New pass.
20382 * tree-pass.h (make_pass_early_remat): Declare.
20383 * early-remat.c: New file.
20384 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
20386 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
20388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20389 Alan Hayward <alan.hayward@arm.com>
20390 David Sherwood <david.sherwood@arm.com>
20392 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
20393 vfm1 with a bound_epilog parameter.
20394 (vect_do_peeling): Update calls accordingly, and move the prologue
20395 call earlier in the function. Treat the base bound_epilog as 0 for
20396 fully-masked loops and retain vf - 1 for other loops. Add 1 to
20397 this base when peeling for gaps.
20398 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
20399 with fully-masked loops.
20400 (vect_estimate_min_profitable_iters): Handle the single peeled
20401 iteration in that case.
20403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20404 Alan Hayward <alan.hayward@arm.com>
20405 David Sherwood <david.sherwood@arm.com>
20407 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
20408 single-element interleaving even if the size is not a power of 2.
20409 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
20410 accesses for single-element interleaving if the group size is
20413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20414 Alan Hayward <alan.hayward@arm.com>
20415 David Sherwood <david.sherwood@arm.com>
20417 * doc/md.texi (fold_extract_last_@var{m}): Document.
20418 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
20419 * optabs.def (fold_extract_last_optab): New optab.
20420 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
20421 * internal-fn.c (fold_extract_direct): New macro.
20422 (expand_fold_extract_optab_fn): Likewise.
20423 (direct_fold_extract_optab_supported_p): Likewise.
20424 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
20425 * tree-vect-loop.c (vect_model_reduction_cost): Handle
20426 EXTRACT_LAST_REDUCTION.
20427 (get_initial_def_for_reduction): Do not create an initial vector
20428 for EXTRACT_LAST_REDUCTION reductions.
20429 (vectorizable_reduction): Leave the scalar phi in place for
20430 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
20431 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
20432 epilogue code for EXTRACT_LAST_REDUCTION and defer the
20433 transform phase to vectorizable_condition.
20434 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
20436 (vect_finish_stmt_generation): ...here.
20437 (vect_finish_replace_stmt): New function.
20438 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
20439 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
20441 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
20443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20444 Alan Hayward <alan.hayward@arm.com>
20445 David Sherwood <david.sherwood@arm.com>
20447 * doc/md.texi (extract_last_@var{m}): Document.
20448 * optabs.def (extract_last_optab): New optab.
20449 * internal-fn.def (EXTRACT_LAST): New internal function.
20450 * internal-fn.c (cond_unary_direct): New macro.
20451 (expand_cond_unary_optab_fn): Likewise.
20452 (direct_cond_unary_optab_supported_p): Likewise.
20453 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
20454 loops using EXTRACT_LAST.
20455 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
20456 (extract_last_<mode>): ...this optab.
20457 (vec_extract<mode><Vel>): Update accordingly.
20459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20460 Alan Hayward <alan.hayward@arm.com>
20461 David Sherwood <david.sherwood@arm.com>
20463 * target.def (empty_mask_is_expensive): New hook.
20464 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
20465 * doc/tm.texi: Regenerate.
20466 * targhooks.h (default_empty_mask_is_expensive): Declare.
20467 * targhooks.c (default_empty_mask_is_expensive): New function.
20468 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
20469 if the target says that empty masks are expensive.
20470 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
20472 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
20474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20475 Alan Hayward <alan.hayward@arm.com>
20476 David Sherwood <david.sherwood@arm.com>
20478 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
20479 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
20480 (vect_use_loop_mask_for_alignment_p): New function.
20481 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
20482 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
20483 niters_skip argument. Make sure that the first niters_skip elements
20484 of the first iteration are inactive.
20485 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
20486 Update call to vect_set_loop_masks_directly.
20487 (get_misalign_in_elems): New function, split out from...
20488 (vect_gen_prolog_loop_niters): ...here.
20489 (vect_update_init_of_dr): Take a code argument that specifies whether
20490 the adjustment should be added or subtracted.
20491 (vect_update_init_of_drs): Likewise.
20492 (vect_prepare_for_masked_peels): New function.
20493 (vect_do_peeling): Skip prologue peeling if we're using a mask
20494 instead. Update call to vect_update_inits_of_drs.
20495 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20497 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
20498 alignment. Do not include the number of peeled iterations in
20499 the minimum threshold in that case.
20500 (vectorizable_induction): Adjust the start value down by
20501 LOOP_VINFO_MASK_SKIP_NITERS iterations.
20502 (vect_transform_loop): Call vect_prepare_for_masked_peels.
20503 Take the number of skipped iterations into account when calculating
20505 * tree-vect-stmts.c (vect_gen_while_not): New function.
20507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20508 Alan Hayward <alan.hayward@arm.com>
20509 David Sherwood <david.sherwood@arm.com>
20511 * doc/sourcebuild.texi (vect_fully_masked): Document.
20512 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
20513 default value to 0.
20514 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
20516 (vect_analyze_loop_2): ...here. Don't check the vectorization
20517 factor against the number of loop iterations if the loop is
20520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20521 Alan Hayward <alan.hayward@arm.com>
20522 David Sherwood <david.sherwood@arm.com>
20524 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
20525 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
20526 (dump_groups): Update accordingly.
20527 (iv_use::mem_type): New member variable.
20528 (address_p): New function.
20529 (record_use): Add a mem_type argument and initialize the new
20531 (record_group_use): Add a mem_type argument. Use address_p.
20532 Remove obsolete null checks of base_object. Update call to record_use.
20533 (find_interesting_uses_op): Update call to record_group_use.
20534 (find_interesting_uses_cond): Likewise.
20535 (find_interesting_uses_address): Likewise.
20536 (get_mem_type_for_internal_fn): New function.
20537 (find_address_like_use): Likewise.
20538 (find_interesting_uses_stmt): Try find_address_like_use before
20539 calling find_interesting_uses_op.
20540 (addr_offset_valid_p): Use the iv mem_type field as the type
20541 of the addressed memory.
20542 (add_autoinc_candidates): Likewise.
20543 (get_address_cost): Likewise.
20544 (split_small_address_groups_p): Use address_p.
20545 (split_address_groups): Likewise.
20546 (add_iv_candidate_for_use): Likewise.
20547 (autoinc_possible_for_pair): Likewise.
20548 (rewrite_groups): Likewise.
20549 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
20550 (determine_group_iv_cost): Update after split of USE_ADDRESS.
20551 (get_alias_ptr_type_for_ptr_address): New function.
20552 (rewrite_use_address): Rewrite address uses in calls that were
20553 identified by find_address_like_use.
20555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20556 Alan Hayward <alan.hayward@arm.com>
20557 David Sherwood <david.sherwood@arm.com>
20559 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
20561 * gimple-expr.h (is_gimple_addressable: Likewise.
20562 * gimple-expr.c (is_gimple_address): Likewise.
20563 * internal-fn.c (expand_call_mem_ref): New function.
20564 (expand_mask_load_optab_fn): Use it.
20565 (expand_mask_store_optab_fn): Likewise.
20567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20568 Alan Hayward <alan.hayward@arm.com>
20569 David Sherwood <david.sherwood@arm.com>
20571 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
20572 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
20573 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
20574 (cond_umax@var{mode}): Document.
20575 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
20576 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
20577 (cond_umin_optab, cond_umax_optab): New optabs.
20578 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
20579 (COND_IOR, COND_XOR): New internal functions.
20580 * internal-fn.h (get_conditional_internal_fn): Declare.
20581 * internal-fn.c (cond_binary_direct): New macro.
20582 (expand_cond_binary_optab_fn): Likewise.
20583 (direct_cond_binary_optab_supported_p): Likewise.
20584 (get_conditional_internal_fn): New function.
20585 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
20586 Cope with reduction statements that are vectorized as calls rather
20588 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
20589 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
20590 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
20591 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
20592 (UNSPEC_COND_EOR): New unspecs.
20593 (optab): Add mappings for them.
20594 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
20595 (sve_int_op, sve_fp_op): New int attributes.
20597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20598 Alan Hayward <alan.hayward@arm.com>
20599 David Sherwood <david.sherwood@arm.com>
20601 * optabs.def (while_ult_optab): New optab.
20602 * doc/md.texi (while_ult@var{m}@var{n}): Document.
20603 * internal-fn.def (WHILE_ULT): New internal function.
20604 * internal-fn.h (direct_internal_fn_supported_p): New override
20605 that takes two types as argument.
20606 * internal-fn.c (while_direct): New macro.
20607 (expand_while_optab_fn): New function.
20608 (convert_optab_supported_p): Likewise.
20609 (direct_while_optab_supported_p): New macro.
20610 * wide-int.h (wi::udiv_ceil): New function.
20611 * tree-vectorizer.h (rgroup_masks): New structure.
20612 (vec_loop_masks): New typedef.
20613 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
20614 and fully_masked_p.
20615 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
20616 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
20617 (vect_max_vf): New function.
20618 (slpeel_make_loop_iterate_ntimes): Delete.
20619 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
20620 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
20621 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
20622 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
20623 internal-fn.h, stor-layout.h and optabs-query.h.
20624 (vect_set_loop_mask): New function.
20625 (add_preheader_seq): Likewise.
20626 (add_header_seq): Likewise.
20627 (interleave_supported_p): Likewise.
20628 (vect_maybe_permute_loop_masks): Likewise.
20629 (vect_set_loop_masks_directly): Likewise.
20630 (vect_set_loop_condition_masked): Likewise.
20631 (vect_set_loop_condition_unmasked): New function, split out from
20632 slpeel_make_loop_iterate_ntimes.
20633 (slpeel_make_loop_iterate_ntimes): Rename to..
20634 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
20635 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
20636 (vect_do_peeling): Update call accordingly.
20637 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
20639 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20640 mask_compare_type, can_fully_mask_p and fully_masked_p.
20641 (release_vec_loop_masks): New function.
20642 (_loop_vec_info): Use it to free the loop masks.
20643 (can_produce_all_loop_masks_p): New function.
20644 (vect_get_max_nscalars_per_iter): Likewise.
20645 (vect_verify_full_masking): Likewise.
20646 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
20647 retries, and free the mask rgroups before retrying. Check loop-wide
20648 reasons for disallowing fully-masked loops. Make the final decision
20649 about whether use a fully-masked loop or not.
20650 (vect_estimate_min_profitable_iters): Do not assume that peeling
20651 for the number of iterations will be needed for fully-masked loops.
20652 (vectorizable_reduction): Disable fully-masked loops.
20653 (vectorizable_live_operation): Likewise.
20654 (vect_halve_mask_nunits): New function.
20655 (vect_double_mask_nunits): Likewise.
20656 (vect_record_loop_mask): Likewise.
20657 (vect_get_loop_mask): Likewise.
20658 (vect_transform_loop): Handle the case in which the final loop
20659 iteration might handle a partial vector. Call vect_set_loop_condition
20660 instead of slpeel_make_loop_iterate_ntimes.
20661 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
20662 (check_load_store_masking): New function.
20663 (prepare_load_store_mask): Likewise.
20664 (vectorizable_store): Handle fully-masked loops.
20665 (vectorizable_load): Likewise.
20666 (supportable_widening_operation): Use vect_halve_mask_nunits for
20668 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
20669 (vect_gen_while): New function.
20670 * config/aarch64/aarch64.md (umax<mode>3): New expander.
20671 (aarch64_uqdec<mode>): New insn.
20673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20674 Alan Hayward <alan.hayward@arm.com>
20675 David Sherwood <david.sherwood@arm.com>
20677 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
20678 (reduc_xor_scal_optab): New optabs.
20679 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
20680 (reduc_xor_scal_@var{m}): Document.
20681 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
20682 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
20683 internal functions.
20684 * fold-const-call.c (fold_const_call): Handle them.
20685 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
20686 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
20687 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
20688 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
20689 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
20690 (UNSPEC_XORV): New unspecs.
20691 (optab): Add entries for them.
20692 (BITWISEV): New int iterator.
20693 (bit_reduc_op): New int attributes.
20695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20696 Alan Hayward <alan.hayward@arm.com>
20697 David Sherwood <david.sherwood@arm.com>
20699 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
20700 * internal-fn.def (VEC_SHL_INSERT): New internal function.
20701 * optabs.def (vec_shl_insert_optab): New optab.
20702 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
20703 (duplicate_and_interleave): Likewise.
20704 * tree-vect-loop.c: Include internal-fn.h.
20705 (neutral_op_for_slp_reduction): New function, split out from
20706 get_initial_defs_for_reduction.
20707 (get_initial_def_for_reduction): Handle option 2 for variable-length
20708 vectors by loading the neutral value into a vector and then shifting
20709 the initial value into element 0.
20710 (get_initial_defs_for_reduction): Replace the code argument with
20711 the neutral value calculated by neutral_op_for_slp_reduction.
20712 Use gimple_build_vector for constant-length vectors.
20713 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
20714 but the first group_size elements have a neutral value.
20715 Use duplicate_and_interleave otherwise.
20716 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
20717 Update call to get_initial_defs_for_reduction. Handle SLP
20718 reductions for variable-length vectors by creating one vector
20719 result for each scalar result, with the elements associated
20720 with other scalar results stubbed out with the neutral value.
20721 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
20722 Require IFN_VEC_SHL_INSERT for double reductions on
20723 variable-length vectors, or SLP reductions that have
20724 a neutral value. Require can_duplicate_and_interleave_p
20725 support for variable-length unchained SLP reductions if there
20726 is no neutral value, such as for MIN/MAX reductions. Also require
20727 the number of vector elements to be a multiple of the number of
20728 SLP statements when doing variable-length unchained SLP reductions.
20729 Update call to vect_create_epilog_for_reduction.
20730 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
20731 and remove initial values.
20732 (duplicate_and_interleave): Make public.
20733 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
20734 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
20736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20737 Alan Hayward <alan.hayward@arm.com>
20738 David Sherwood <david.sherwood@arm.com>
20740 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
20741 (can_duplicate_and_interleave_p): New function.
20742 (vect_get_and_check_slp_defs): Take the vector of statements
20743 rather than just the current one. Remove excess parentheses.
20744 Restriction rejectinon of vect_constant_def and vect_external_def
20745 for variable-length vectors to boolean types, or types for which
20746 can_duplicate_and_interleave_p is false.
20747 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
20748 (duplicate_and_interleave): New function.
20749 (vect_get_constant_vectors): Use gimple_build_vector for
20750 constant-length vectors and suitable variable-length constant
20751 vectors. Use duplicate_and_interleave for other variable-length
20752 vectors. Don't defer the update when inserting new statements.
20754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20755 Alan Hayward <alan.hayward@arm.com>
20756 David Sherwood <david.sherwood@arm.com>
20758 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
20759 min_profitable_iters doesn't go negative.
20761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20762 Alan Hayward <alan.hayward@arm.com>
20763 David Sherwood <david.sherwood@arm.com>
20765 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
20766 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
20767 * optabs.def (vec_mask_load_lanes_optab): New optab.
20768 (vec_mask_store_lanes_optab): Likewise.
20769 * internal-fn.def (MASK_LOAD_LANES): New internal function.
20770 (MASK_STORE_LANES): Likewise.
20771 * internal-fn.c (mask_load_lanes_direct): New macro.
20772 (mask_store_lanes_direct): Likewise.
20773 (expand_mask_load_optab_fn): Handle masked operations.
20774 (expand_mask_load_lanes_optab_fn): New macro.
20775 (expand_mask_store_optab_fn): Handle masked operations.
20776 (expand_mask_store_lanes_optab_fn): New macro.
20777 (direct_mask_load_lanes_optab_supported_p): Likewise.
20778 (direct_mask_store_lanes_optab_supported_p): Likewise.
20779 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
20781 (vect_load_lanes_supported): Likewise.
20782 * tree-vect-data-refs.c (strip_conversion): New function.
20783 (can_group_stmts_p): Likewise.
20784 (vect_analyze_data_ref_accesses): Use it instead of checking
20785 for a pair of assignments.
20786 (vect_store_lanes_supported): Take a masked_p parameter.
20787 (vect_load_lanes_supported): Likewise.
20788 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
20789 vect_store_lanes_supported and vect_load_lanes_supported.
20790 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
20791 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
20792 parameter. Don't allow gaps for masked accesses.
20793 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
20794 and vect_load_lanes_supported.
20795 (get_load_store_type): Take a masked_p parameter and update
20796 call to get_group_load_store_type.
20797 (vectorizable_store): Update call to get_load_store_type.
20798 Handle IFN_MASK_STORE_LANES.
20799 (vectorizable_load): Update call to get_load_store_type.
20800 Handle IFN_MASK_LOAD_LANES.
20802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20803 Alan Hayward <alan.hayward@arm.com>
20804 David Sherwood <david.sherwood@arm.com>
20806 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
20808 * config/aarch64/aarch64-protos.h
20809 (aarch64_sve_struct_memory_operand_p): Declare.
20810 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
20811 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
20812 (VPRED, vpred): Handle SVE structure modes.
20813 * config/aarch64/constraints.md (Utx): New constraint.
20814 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
20815 (aarch64_sve_struct_nonimmediate_operand): New predicates.
20816 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
20817 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
20818 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
20819 structure modes. Split into pieces after RA.
20820 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
20821 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
20823 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
20824 SVE structure modes.
20825 (aarch64_classify_address): Likewise.
20826 (sizetochar): Move earlier in file.
20827 (aarch64_print_operand): Handle SVE register lists.
20828 (aarch64_array_mode): New function.
20829 (aarch64_sve_struct_memory_operand_p): Likewise.
20830 (TARGET_ARRAY_MODE): Redefine.
20832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20833 Alan Hayward <alan.hayward@arm.com>
20834 David Sherwood <david.sherwood@arm.com>
20836 * target.def (array_mode): New target hook.
20837 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
20838 * doc/tm.texi: Regenerate.
20839 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
20840 * hooks.c (hook_optmode_mode_uhwi_none): New function.
20841 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
20842 targetm.array_mode.
20843 * stor-layout.c (mode_for_array): Likewise. Support polynomial
20846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20847 Alan Hayward <alan.hayward@arm.com>
20848 David Sherwood <david.sherwood@arm.com>
20850 * fold-const.c (fold_binary_loc): Check the argument types
20851 rather than the result type when testing for a vector operation.
20853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20855 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
20856 * doc/tm.texi: Regenerate.
20858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20859 Alan Hayward <alan.hayward@arm.com>
20860 David Sherwood <david.sherwood@arm.com>
20862 * doc/invoke.texi (-msve-vector-bits=): Document new option.
20863 (sve): Document new AArch64 extension.
20864 * doc/md.texi (w): Extend the description of the AArch64
20865 constraint to include SVE vectors.
20866 (Upl, Upa): Document new AArch64 predicate constraints.
20867 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
20869 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
20870 (msve-vector-bits=): New option.
20871 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
20872 SVE when these are disabled.
20873 (sve): New extension.
20874 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
20875 modes. Adjust their number of units based on aarch64_sve_vg.
20876 (MAX_BITSIZE_MODE_ANY_MODE): Define.
20877 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
20878 aarch64_addr_query_type.
20879 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
20880 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
20881 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
20882 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
20883 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
20884 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
20885 (aarch64_simd_imm_zero_p): Delete.
20886 (aarch64_check_zero_based_sve_index_immediate): Declare.
20887 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20888 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20889 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20890 (aarch64_sve_float_mul_immediate_p): Likewise.
20891 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20892 rather than an rtx.
20893 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
20894 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
20895 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
20896 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
20897 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
20898 (aarch64_regmode_natural_size): Likewise.
20899 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
20900 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
20902 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
20903 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
20904 for VG and the SVE predicate registers.
20905 (V_ALIASES): Add a "z"-prefixed alias.
20906 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
20907 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
20908 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
20909 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
20910 (REG_CLASS_NAMES): Add entries for them.
20911 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
20912 and the predicate registers.
20913 (aarch64_sve_vg): Declare.
20914 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
20915 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
20916 (REGMODE_NATURAL_SIZE): Define.
20917 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
20919 * config/aarch64/aarch64.c: Include cfgrtl.h.
20920 (simd_immediate_info): Add a constructor for series vectors,
20921 and an associated step field.
20922 (aarch64_sve_vg): New variable.
20923 (aarch64_dbx_register_number): Handle VG and the predicate registers.
20924 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
20925 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
20926 (VEC_ANY_DATA, VEC_STRUCT): New constants.
20927 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
20928 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
20929 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
20930 (aarch64_get_mask_mode): New functions.
20931 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
20932 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
20933 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
20934 predicate modes and predicate registers. Explicitly restrict
20935 GPRs to modes of 16 bytes or smaller. Only allow FP registers
20936 to store a vector mode if it is recognized by
20937 aarch64_classify_vector_mode.
20938 (aarch64_regmode_natural_size): New function.
20939 (aarch64_hard_regno_caller_save_mode): Return the original mode
20941 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
20942 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
20943 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
20944 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
20946 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
20947 does not overlap dest if the function is frame-related. Handle
20949 (aarch64_split_add_offset): New function.
20950 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
20951 them aarch64_add_offset.
20952 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20953 and update call to aarch64_sub_sp.
20954 (aarch64_add_cfa_expression): New function.
20955 (aarch64_expand_prologue): Pass extra temporary registers to the
20956 functions above. Handle the case in which we need to emit new
20957 DW_CFA_expressions for registers that were originally saved
20958 relative to the stack pointer, but now have to be expressed
20959 relative to the frame pointer.
20960 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20962 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20963 IP0 and IP1 values for SVE frames.
20964 (aarch64_expand_vec_series): New function.
20965 (aarch64_expand_sve_widened_duplicate): Likewise.
20966 (aarch64_expand_sve_const_vector): Likewise.
20967 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20968 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20969 into the register, rather than emitting a SET directly.
20970 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20971 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20972 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20973 (offset_9bit_signed_scaled_p): New functions.
20974 (aarch64_replicate_bitmask_imm): New function.
20975 (aarch64_bitmask_imm): Use it.
20976 (aarch64_cannot_force_const_mem): Reject expressions involving
20977 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20978 (aarch64_classify_index): Handle SVE indices, by requiring
20979 a plain register index with a scale that matches the element size.
20980 (aarch64_classify_address): Handle SVE addresses. Assert that
20981 the mode of the address is VOIDmode or an integer mode.
20982 Update call to aarch64_classify_symbol.
20983 (aarch64_classify_symbolic_expression): Update call to
20984 aarch64_classify_symbol.
20985 (aarch64_const_vec_all_in_range_p): New function.
20986 (aarch64_print_vector_float_operand): Likewise.
20987 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20988 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20989 and the FP immediates 1.0 and 0.5.
20990 (aarch64_print_address_internal): Handle SVE addresses.
20991 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20992 (aarch64_regno_regclass): Handle predicate registers.
20993 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20995 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20996 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20997 (aarch64_convert_sve_vector_bits): New function.
20998 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20999 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21000 rather than an rtx.
21001 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21002 Handle SVE vector and predicate modes. Accept VL-based constants
21003 that need only one temporary register, and VL offsets that require
21004 no temporary registers.
21005 (aarch64_conditional_register_usage): Mark the predicate registers
21006 as fixed if SVE isn't available.
21007 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21008 Return true for SVE vector and predicate modes.
21009 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21010 rather than an unsigned int. Handle SVE modes.
21011 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21013 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21015 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21016 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21017 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21018 (aarch64_sve_float_mul_immediate_p): New functions.
21019 (aarch64_sve_valid_immediate): New function.
21020 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21021 Explicitly reject structure modes. Check for INDEX constants.
21022 Handle PTRUE and PFALSE constants.
21023 (aarch64_check_zero_based_sve_index_immediate): New function.
21024 (aarch64_simd_imm_zero_p): Delete.
21025 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21026 vector modes. Accept constants in the range of CNT[BHWD].
21027 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21028 ask for an Advanced SIMD mode.
21029 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21030 (aarch64_simd_vector_alignment): Handle SVE predicates.
21031 (aarch64_vectorize_preferred_vector_alignment): New function.
21032 (aarch64_simd_vector_alignment_reachable): Use it instead of
21034 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21035 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21037 (MAX_VECT_LEN): Delete.
21038 (expand_vec_perm_d): Add a vec_flags field.
21039 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21040 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21041 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21043 (aarch64_evpc_rev): Rename to...
21044 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21045 (aarch64_evpc_rev_global): New function.
21046 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21047 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21049 (aarch64_evpc_sve_tbl): New function.
21050 (aarch64_expand_vec_perm_const_1): Update after rename of
21051 aarch64_evpc_rev. Handle SVE permutes too, trying
21052 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21053 than aarch64_evpc_tbl.
21054 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21055 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21056 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21057 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21058 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21059 (aarch64_expand_sve_vcond): New functions.
21060 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21061 of aarch64_vector_mode_p.
21062 (aarch64_dwarf_poly_indeterminate_value): New function.
21063 (aarch64_compute_pressure_classes): Likewise.
21064 (aarch64_can_change_mode_class): Likewise.
21065 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21066 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21067 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21068 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21069 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21070 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21071 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21072 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21074 (Dn, Dl, Dr): Accept const as well as const_vector.
21075 (Dz): Likewise. Compare against CONST0_RTX.
21076 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21077 of "vector" where appropriate.
21078 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21079 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21080 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21081 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21082 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21083 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21084 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21085 (v_int_equiv): Extend to SVE modes.
21086 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21088 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21089 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21090 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21091 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21092 (SVE_COND_FP_CMP): New int iterators.
21093 (perm_hilo): Handle the new unpack unspecs.
21094 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21096 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21097 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21098 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21099 (aarch64_equality_operator, aarch64_constant_vector_operand)
21100 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21101 (aarch64_sve_nonimmediate_operand): Likewise.
21102 (aarch64_sve_general_operand): Likewise.
21103 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21104 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21105 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21106 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21107 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21108 (aarch64_sve_float_arith_immediate): Likewise.
21109 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21110 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21111 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21112 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21113 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21114 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21115 (aarch64_sve_float_arith_operand): Likewise.
21116 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21117 (aarch64_sve_float_mul_operand): Likewise.
21118 (aarch64_sve_vec_perm_operand): Likewise.
21119 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21120 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21121 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21122 as well as const_vector.
21123 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21124 in file. Use CONST0_RTX and CONSTM1_RTX.
21125 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21126 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21127 Use aarch64_simd_imm_zero.
21128 * config/aarch64/aarch64-sve.md: New file.
21129 * config/aarch64/aarch64.md: Include it.
21130 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21131 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21132 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21133 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21134 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21135 (sve): New attribute.
21136 (enabled): Disable instructions with the sve attribute unless
21138 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21139 aarch64_expand_mov_immediate.
21140 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21141 CNT[BHSD] immediates.
21142 (movti): Split CONST_POLY_INT moves into two halves.
21143 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21144 Split additions that need a temporary here if the destination
21145 is the stack pointer.
21146 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21147 (*add<mode>3_poly_1): New instruction.
21148 (set_clobber_cc): New expander.
21150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21152 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21153 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21154 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21155 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21156 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21157 Change innermode from fixed_mode_size to machine_mode.
21158 (simplify_subreg): Update call accordingly. Handle a constant-sized
21159 subreg of a variable-length CONST_VECTOR.
21161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21162 Alan Hayward <alan.hayward@arm.com>
21163 David Sherwood <david.sherwood@arm.com>
21165 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21166 (add_offset_to_base): New function, split out from...
21167 (create_mem_ref): ...here. When handling a scale other than 1,
21168 check first whether the address is valid without the offset.
21169 Add it into the base if so, leaving the index and scale as-is.
21171 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21174 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21175 fold_for_warn before checking if arg2 is INTEGER_CST.
21177 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21179 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21180 (store_multiple_operation): Delete.
21181 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21182 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21183 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21184 guarded by TARGET_STRING.
21185 (rs6000_output_load_multiple): Delete.
21186 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21187 OPTION_MASK_STRING / TARGET_STRING handling.
21188 (print_operand) <'N', 'O'>: Add comment that these are unused now.
21189 (const rs6000_opt_masks) <"string">: Change mask to 0.
21190 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
21191 (MASK_STRING): Delete.
21192 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
21194 (load_multiple): Delete.
21201 (store_multiple): Delete.
21208 (movmemsi_8reg): Delete.
21209 (corresponding unnamed define_insn): Delete.
21210 (movmemsi_6reg): Delete.
21211 (corresponding unnamed define_insn): Delete.
21212 (movmemsi_4reg): Delete.
21213 (corresponding unnamed define_insn): Delete.
21214 (movmemsi_2reg): Delete.
21215 (corresponding unnamed define_insn): Delete.
21216 (movmemsi_1reg): Delete.
21217 (corresponding unnamed define_insn): Delete.
21218 * config/rs6000/rs6000.opt (mno-string): New.
21219 (mstring): Replace by deprecation warning stub.
21220 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
21222 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21224 * regrename.c (regrename_do_replace): If replacing the same
21225 reg multiple times, try to reuse last created gen_raw_REG.
21228 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
21229 main to workaround a bug in GDB.
21231 2018-01-12 Tom de Vries <tom@codesourcery.com>
21234 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
21236 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
21238 PR rtl-optimization/80481
21239 * ira-color.c (get_cap_member): New function.
21240 (allocnos_conflict_by_live_ranges_p): Use it.
21241 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
21242 (setup_slot_coalesced_allocno_live_ranges): Ditto.
21244 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
21247 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
21248 (*saddl_se_1): Ditto.
21249 (*ssubsi_1): Ditto.
21250 (*ssubl_se_1): Ditto.
21252 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21254 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
21255 rather than wi::to_widest for DR_INITs.
21256 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
21257 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
21258 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
21260 (vect_analyze_group_access_1): Note that here.
21262 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21264 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
21265 polynomial type sizes.
21267 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21269 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
21270 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
21271 (gimple_add_tmp_var): Likewise.
21273 2018-01-12 Martin Liska <mliska@suse.cz>
21275 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
21276 (gimple_alloc_sizes): Likewise.
21277 (dump_gimple_statistics): Use PRIu64 in printf format.
21278 * gimple.h: Change uint64_t to int.
21280 2018-01-12 Martin Liska <mliska@suse.cz>
21282 * tree-core.h: Use uint64_t instead of int.
21283 * tree.c (tree_node_counts): Likewise.
21284 (tree_node_sizes): Likewise.
21285 (dump_tree_statistics): Use PRIu64 in printf format.
21287 2018-01-12 Martin Liska <mliska@suse.cz>
21289 * Makefile.in: As qsort_chk is implemented in vec.c, add
21290 vec.o to linkage of gencfn-macros.
21291 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
21292 passing the info to record_node_allocation_statistics.
21293 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
21295 * ggc-common.c (struct ggc_usage): Add operator== and use
21296 it in operator< and compare function.
21297 * mem-stats.h (struct mem_usage): Likewise.
21298 * vec.c (struct vec_usage): Remove operator< and compare
21299 function. Can be simply inherited.
21301 2018-01-12 Martin Jambor <mjambor@suse.cz>
21304 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
21305 * tree-ssa-math-opts.c: Include domwalk.h.
21306 (convert_mult_to_fma_1): New function.
21307 (fma_transformation_info): New type.
21308 (fma_deferring_state): Likewise.
21309 (cancel_fma_deferring): New function.
21310 (result_of_phi): Likewise.
21311 (last_fma_candidate_feeds_initial_phi): Likewise.
21312 (convert_mult_to_fma): Added deferring logic, split actual
21313 transformation to convert_mult_to_fma_1.
21314 (math_opts_dom_walker): New type.
21315 (math_opts_dom_walker::after_dom_children): New method, body moved
21316 here from pass_optimize_widening_mul::execute, added deferring logic
21318 (pass_optimize_widening_mul::execute): Moved most of code to
21319 math_opts_dom_walker::after_dom_children.
21320 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
21321 * config/i386/i386.c (ix86_option_override_internal): Added
21322 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
21324 2018-01-12 Richard Biener <rguenther@suse.de>
21327 * dwarf2out.c (gen_variable_die): Do not reset old_die for
21328 inline instance vars.
21330 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
21333 * config/rx/rx.c (rx_is_restricted_memory_address):
21334 Handle SUBREG case.
21336 2018-01-12 Richard Biener <rguenther@suse.de>
21338 PR tree-optimization/80846
21339 * target.def (split_reduction): New target hook.
21340 * targhooks.c (default_split_reduction): New function.
21341 * targhooks.h (default_split_reduction): Declare.
21342 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
21343 target requests first reduce vectors by combining low and high
21345 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
21346 (get_vectype_for_scalar_type_and_size): Export.
21347 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
21348 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
21349 * doc/tm.texi: Regenerate.
21350 * config/i386/i386.c (ix86_split_reduction): Implement
21351 TARGET_VECTORIZE_SPLIT_REDUCTION.
21353 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21356 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
21357 in PIC mode except for TARGET_VXWORKS_RTP.
21358 * config/sparc/sparc.c: Include cfgrtl.h.
21359 (TARGET_INIT_PIC_REG): Define.
21360 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
21361 (sparc_pic_register_p): New predicate.
21362 (sparc_legitimate_address_p): Use it.
21363 (sparc_legitimize_pic_address): Likewise.
21364 (sparc_delegitimize_address): Likewise.
21365 (sparc_mode_dependent_address_p): Likewise.
21366 (gen_load_pcrel_sym): Remove 4th parameter.
21367 (load_got_register): Adjust call to above. Remove obsolete stuff.
21368 (sparc_expand_prologue): Do not call load_got_register here.
21369 (sparc_flat_expand_prologue): Likewise.
21370 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
21371 (sparc_use_pseudo_pic_reg): New function.
21372 (sparc_init_pic_reg): Likewise.
21373 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21374 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
21376 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
21378 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
21379 Add item for branch_cost.
21381 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21383 PR rtl-optimization/83565
21384 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
21385 not extend the result to a larger mode for rotate operations.
21386 (num_sign_bit_copies1): Likewise.
21388 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21391 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
21393 Use values-Xc.o for -pedantic.
21394 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
21396 2018-01-12 Martin Liska <mliska@suse.cz>
21399 * ipa-devirt.c (final_warning_record::grow_type_warnings):
21401 (possible_polymorphic_call_targets): Use it.
21402 (ipa_devirt): Likewise.
21404 2018-01-12 Martin Liska <mliska@suse.cz>
21406 * profile-count.h (enum profile_quality): Use 0 as invalid
21407 enum value of profile_quality.
21409 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
21411 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
21412 -mext-string options.
21414 2018-01-12 Richard Biener <rguenther@suse.de>
21416 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
21417 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
21418 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21420 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21422 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
21424 * configure.ac (--with-long-double-format): Add support for the
21425 configuration option to change the default long double format on
21427 * config.gcc (powerpc*-linux*-*): Likewise.
21428 * configure: Regenerate.
21429 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
21430 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
21431 used without modification.
21433 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21435 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
21436 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
21437 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
21438 MISC_BUILTIN_SPEC_BARRIER.
21439 (rs6000_init_builtins): Likewise.
21440 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
21442 (speculation_barrier): New define_insn.
21443 * doc/extend.texi: Document __builtin_speculation_barrier.
21445 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21448 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
21449 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
21450 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
21452 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
21453 integral modes instead of "ss" and "sd".
21454 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
21455 vectors with 32-bit and 64-bit elements.
21456 (vecdupssescalarmodesuffix): New mode attribute.
21457 (vec_dup<mode>): Use it.
21459 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
21462 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
21463 frame if argument is passed on stack.
21465 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21468 * ree.c (combine_reaching_defs): Optimize also
21469 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
21470 reg2=any_extend(exp); reg1=reg2;, formatting fix.
21472 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21474 PR middle-end/83189
21475 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
21477 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21479 PR middle-end/83718
21480 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
21481 after they are computed.
21483 2018-01-11 Bin Cheng <bin.cheng@arm.com>
21485 PR tree-optimization/83695
21486 * gimple-loop-linterchange.cc
21487 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
21488 reset cached scev information after interchange.
21489 (pass_linterchange::execute): Remove call to scev_reset_htab.
21491 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21493 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
21494 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
21495 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
21496 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
21497 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
21498 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
21499 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
21500 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
21501 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
21502 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
21503 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
21504 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
21505 (V_lane_reg): Likewise.
21506 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
21508 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
21509 (vfmal_lane_low<mode>_intrinsic,
21510 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
21511 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
21512 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
21513 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
21514 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
21515 vfmsl_lane_high<mode>_intrinsic): New define_insns.
21517 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21519 * config/arm/arm-cpus.in (fp16fml): New feature.
21520 (ALL_SIMD): Add fp16fml.
21521 (armv8.2-a): Add fp16fml as an option.
21522 (armv8.3-a): Likewise.
21523 (armv8.4-a): Add fp16fml as part of fp16.
21524 * config/arm/arm.h (TARGET_FP16FML): Define.
21525 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
21527 * config/arm/arm-modes.def (V2HF): Define.
21528 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
21529 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
21530 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
21531 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
21532 vfmsl_low, vfmsl_high): New set of builtins.
21533 * config/arm/iterators.md (PLUSMINUS): New code iterator.
21534 (vfml_op): New code attribute.
21535 (VFMLHALVES): New int iterator.
21536 (VFML, VFMLSEL): New mode attributes.
21537 (V_reg): Define mapping for V2HF.
21538 (V_hi, V_lo): New mode attributes.
21539 (VF_constraint): Likewise.
21540 (vfml_half, vfml_half_selector): New int attributes.
21541 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
21543 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
21544 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
21546 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
21547 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
21548 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
21549 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
21551 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
21552 Document new effective target and option set.
21554 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21556 * config/arm/arm-cpus.in (armv8_4): New feature.
21557 (ARMv8_4a): New fgroup.
21558 (armv8.4-a): New arch.
21559 * config/arm/arm-tables.opt: Regenerate.
21560 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
21561 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
21562 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
21563 Add matching rules for -march=armv8.4-a and extensions.
21564 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
21566 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
21569 * config/rx/rx.md (BW): New mode attribute.
21570 (sync_lock_test_and_setsi): Add mode suffix to insn output.
21572 2018-01-11 Richard Biener <rguenther@suse.de>
21574 PR tree-optimization/83435
21575 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
21576 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
21577 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
21579 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21580 Alan Hayward <alan.hayward@arm.com>
21581 David Sherwood <david.sherwood@arm.com>
21583 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
21585 (aarch64_classify_address): Initialize it. Track polynomial offsets.
21586 (aarch64_print_address_internal): Use it to check for a zero offset.
21588 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21589 Alan Hayward <alan.hayward@arm.com>
21590 David Sherwood <david.sherwood@arm.com>
21592 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
21593 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
21594 Return a poly_int64 rather than a HOST_WIDE_INT.
21595 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
21596 rather than a HOST_WIDE_INT.
21597 * config/aarch64/aarch64.h (aarch64_frame): Protect with
21598 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
21599 hard_fp_offset, frame_size, initial_adjust, callee_offset and
21600 final_offset from HOST_WIDE_INT to poly_int64.
21601 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
21602 to_constant when getting the number of units in an Advanced SIMD
21604 (aarch64_builtin_vectorized_function): Check for a constant number
21606 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
21608 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
21609 attribute instead of GET_MODE_NUNITS.
21610 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
21611 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
21612 GET_MODE_SIZE for fixed-size registers.
21613 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
21614 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
21615 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
21616 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
21617 (aarch64_print_operand, aarch64_print_address_internal)
21618 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
21619 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
21620 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
21621 Handle polynomial GET_MODE_SIZE.
21622 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
21623 wider than SImode without modification.
21624 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
21625 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
21626 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
21627 passing and returning SVE modes.
21628 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
21629 rather than GEN_INT.
21630 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
21631 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
21632 (aarch64_allocate_and_probe_stack_space): Likewise.
21633 (aarch64_layout_frame): Cope with polynomial offsets.
21634 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
21635 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
21636 polynomial offsets.
21637 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
21638 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
21639 poly_int64 rather than a HOST_WIDE_INT.
21640 (aarch64_get_separate_components, aarch64_process_components)
21641 (aarch64_expand_prologue, aarch64_expand_epilogue)
21642 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
21643 (aarch64_anchor_offset): New function, split out from...
21644 (aarch64_legitimize_address): ...here.
21645 (aarch64_builtin_vectorization_cost): Handle polynomial
21646 TYPE_VECTOR_SUBPARTS.
21647 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
21649 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
21650 number of elements from the PARALLEL rather than the mode.
21651 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
21652 rather than GET_MODE_BITSIZE.
21653 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
21654 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
21655 (aarch64_expand_vec_perm_const_1): Handle polynomial
21656 d->perm.length () and d->perm elements.
21657 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
21658 Apply to_constant to d->perm elements.
21659 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
21660 polynomial CONST_VECTOR_NUNITS.
21661 (aarch64_move_pointer): Take amount as a poly_int64 rather
21663 (aarch64_progress_pointer): Avoid temporary variable.
21664 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
21665 the mode attribute instead of GET_MODE.
21667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21668 Alan Hayward <alan.hayward@arm.com>
21669 David Sherwood <david.sherwood@arm.com>
21671 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
21672 x exists before using it.
21673 (aarch64_add_constant_internal): Rename to...
21674 (aarch64_add_offset_1): ...this. Replace regnum with separate
21675 src and dest rtxes. Handle the case in which they're different,
21676 including when the offset is zero. Replace scratchreg with an rtx.
21677 Use 2 additions if there is no spare register into which we can
21678 move a 16-bit constant.
21679 (aarch64_add_constant): Delete.
21680 (aarch64_add_offset): Replace reg with separate src and dest
21681 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
21682 Use aarch64_add_offset_1.
21683 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
21684 an rtx rather than an int. Take the delta as a poly_int64
21685 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
21686 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
21687 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
21688 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
21689 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
21690 and aarch64_add_sp.
21691 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
21692 aarch64_add_constant.
21694 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21696 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
21697 Use scalar_float_mode.
21699 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21701 * config/aarch64/aarch64-simd.md
21702 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
21703 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
21704 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
21705 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
21706 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
21707 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
21708 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
21709 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
21710 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
21711 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
21713 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21716 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
21717 targ_options->x_arm_arch_string is non NULL.
21719 2018-01-11 Tamar Christina <tamar.christina@arm.com>
21721 * config/aarch64/aarch64.h
21722 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
21724 2018-01-11 Sudakshina Das <sudi.das@arm.com>
21727 * expmed.c (emit_store_flag_force): Swap if const op0
21728 and change VOIDmode to mode of op0.
21730 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21732 PR rtl-optimization/83761
21733 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
21734 than bytes to mode_for_size.
21736 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21738 PR middle-end/83189
21739 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
21740 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
21743 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21745 PR middle-end/83575
21746 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
21747 when in layout mode.
21748 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
21749 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
21752 2018-01-10 Michael Collison <michael.collison@arm.com>
21754 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
21755 * config/aarch64/aarch64-option-extension.def: Add
21756 AARCH64_OPT_EXTENSION of 'fp16fml'.
21757 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21758 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
21759 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
21760 * config/aarch64/constraints.md (Ui7): New constraint.
21761 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
21762 (VFMLA_SEL_W): Ditto.
21765 (VFMLA16_LOW): New int iterator.
21766 (VFMLA16_HIGH): Ditto.
21767 (UNSPEC_FMLAL): New unspec.
21768 (UNSPEC_FMLSL): Ditto.
21769 (UNSPEC_FMLAL2): Ditto.
21770 (UNSPEC_FMLSL2): Ditto.
21771 (f16mac): New code attribute.
21772 * config/aarch64/aarch64-simd-builtins.def
21773 (aarch64_fmlal_lowv2sf): Ditto.
21774 (aarch64_fmlsl_lowv2sf): Ditto.
21775 (aarch64_fmlalq_lowv4sf): Ditto.
21776 (aarch64_fmlslq_lowv4sf): Ditto.
21777 (aarch64_fmlal_highv2sf): Ditto.
21778 (aarch64_fmlsl_highv2sf): Ditto.
21779 (aarch64_fmlalq_highv4sf): Ditto.
21780 (aarch64_fmlslq_highv4sf): Ditto.
21781 (aarch64_fmlal_lane_lowv2sf): Ditto.
21782 (aarch64_fmlsl_lane_lowv2sf): Ditto.
21783 (aarch64_fmlal_laneq_lowv2sf): Ditto.
21784 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
21785 (aarch64_fmlalq_lane_lowv4sf): Ditto.
21786 (aarch64_fmlsl_lane_lowv4sf): Ditto.
21787 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
21788 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
21789 (aarch64_fmlal_lane_highv2sf): Ditto.
21790 (aarch64_fmlsl_lane_highv2sf): Ditto.
21791 (aarch64_fmlal_laneq_highv2sf): Ditto.
21792 (aarch64_fmlsl_laneq_highv2sf): Ditto.
21793 (aarch64_fmlalq_lane_highv4sf): Ditto.
21794 (aarch64_fmlsl_lane_highv4sf): Ditto.
21795 (aarch64_fmlalq_laneq_highv4sf): Ditto.
21796 (aarch64_fmlsl_laneq_highv4sf): Ditto.
21797 * config/aarch64/aarch64-simd.md:
21798 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
21799 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21800 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
21801 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21802 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
21803 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
21804 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
21805 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
21806 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
21807 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
21808 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
21809 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
21810 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
21811 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
21812 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
21813 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
21814 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
21815 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
21816 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
21817 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
21818 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
21819 (vfmlsl_low_u32): Ditto.
21820 (vfmlalq_low_u32): Ditto.
21821 (vfmlslq_low_u32): Ditto.
21822 (vfmlal_high_u32): Ditto.
21823 (vfmlsl_high_u32): Ditto.
21824 (vfmlalq_high_u32): Ditto.
21825 (vfmlslq_high_u32): Ditto.
21826 (vfmlal_lane_low_u32): Ditto.
21827 (vfmlsl_lane_low_u32): Ditto.
21828 (vfmlal_laneq_low_u32): Ditto.
21829 (vfmlsl_laneq_low_u32): Ditto.
21830 (vfmlalq_lane_low_u32): Ditto.
21831 (vfmlslq_lane_low_u32): Ditto.
21832 (vfmlalq_laneq_low_u32): Ditto.
21833 (vfmlslq_laneq_low_u32): Ditto.
21834 (vfmlal_lane_high_u32): Ditto.
21835 (vfmlsl_lane_high_u32): Ditto.
21836 (vfmlal_laneq_high_u32): Ditto.
21837 (vfmlsl_laneq_high_u32): Ditto.
21838 (vfmlalq_lane_high_u32): Ditto.
21839 (vfmlslq_lane_high_u32): Ditto.
21840 (vfmlalq_laneq_high_u32): Ditto.
21841 (vfmlslq_laneq_high_u32): Ditto.
21842 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
21843 (AARCH64_FL_FOR_ARCH8_4): New.
21844 (AARCH64_ISA_F16FML): New ISA flag.
21845 (TARGET_F16FML): New feature flag for fp16fml.
21846 (doc/invoke.texi): Document new fp16fml option.
21848 2018-01-10 Michael Collison <michael.collison@arm.com>
21850 * config/aarch64/aarch64-builtins.c:
21851 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
21852 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21853 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
21854 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
21855 (AARCH64_ISA_SHA3): New ISA flag.
21856 (TARGET_SHA3): New feature flag for sha3.
21857 * config/aarch64/iterators.md (sha512_op): New int attribute.
21858 (CRYPTO_SHA512): New int iterator.
21859 (UNSPEC_SHA512H): New unspec.
21860 (UNSPEC_SHA512H2): Ditto.
21861 (UNSPEC_SHA512SU0): Ditto.
21862 (UNSPEC_SHA512SU1): Ditto.
21863 * config/aarch64/aarch64-simd-builtins.def
21864 (aarch64_crypto_sha512hqv2di): New builtin.
21865 (aarch64_crypto_sha512h2qv2di): Ditto.
21866 (aarch64_crypto_sha512su0qv2di): Ditto.
21867 (aarch64_crypto_sha512su1qv2di): Ditto.
21868 (aarch64_eor3qv8hi): Ditto.
21869 (aarch64_rax1qv2di): Ditto.
21870 (aarch64_xarqv2di): Ditto.
21871 (aarch64_bcaxqv8hi): Ditto.
21872 * config/aarch64/aarch64-simd.md:
21873 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
21874 (aarch64_crypto_sha512su0qv2di): Ditto.
21875 (aarch64_crypto_sha512su1qv2di): Ditto.
21876 (aarch64_eor3qv8hi): Ditto.
21877 (aarch64_rax1qv2di): Ditto.
21878 (aarch64_xarqv2di): Ditto.
21879 (aarch64_bcaxqv8hi): Ditto.
21880 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
21881 (vsha512h2q_u64): Ditto.
21882 (vsha512su0q_u64): Ditto.
21883 (vsha512su1q_u64): Ditto.
21884 (veor3q_u16): Ditto.
21885 (vrax1q_u64): Ditto.
21886 (vxarq_u64): Ditto.
21887 (vbcaxq_u16): Ditto.
21888 * config/arm/types.md (crypto_sha512): New type attribute.
21889 (crypto_sha3): Ditto.
21890 (doc/invoke.texi): Document new sha3 option.
21892 2018-01-10 Michael Collison <michael.collison@arm.com>
21894 * config/aarch64/aarch64-builtins.c:
21895 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
21896 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21897 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
21898 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
21899 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
21900 (AARCH64_ISA_SM4): New ISA flag.
21901 (TARGET_SM4): New feature flag for sm4.
21902 * config/aarch64/aarch64-simd-builtins.def
21903 (aarch64_sm3ss1qv4si): Ditto.
21904 (aarch64_sm3tt1aq4si): Ditto.
21905 (aarch64_sm3tt1bq4si): Ditto.
21906 (aarch64_sm3tt2aq4si): Ditto.
21907 (aarch64_sm3tt2bq4si): Ditto.
21908 (aarch64_sm3partw1qv4si): Ditto.
21909 (aarch64_sm3partw2qv4si): Ditto.
21910 (aarch64_sm4eqv4si): Ditto.
21911 (aarch64_sm4ekeyqv4si): Ditto.
21912 * config/aarch64/aarch64-simd.md:
21913 (aarch64_sm3ss1qv4si): Ditto.
21914 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
21915 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
21916 (aarch64_sm4eqv4si): Ditto.
21917 (aarch64_sm4ekeyqv4si): Ditto.
21918 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
21919 (sm3part_op): Ditto.
21920 (CRYPTO_SM3TT): Ditto.
21921 (CRYPTO_SM3PART): Ditto.
21922 (UNSPEC_SM3SS1): New unspec.
21923 (UNSPEC_SM3TT1A): Ditto.
21924 (UNSPEC_SM3TT1B): Ditto.
21925 (UNSPEC_SM3TT2A): Ditto.
21926 (UNSPEC_SM3TT2B): Ditto.
21927 (UNSPEC_SM3PARTW1): Ditto.
21928 (UNSPEC_SM3PARTW2): Ditto.
21929 (UNSPEC_SM4E): Ditto.
21930 (UNSPEC_SM4EKEY): Ditto.
21931 * config/aarch64/constraints.md (Ui2): New constraint.
21932 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
21933 * config/arm/types.md (crypto_sm3): New type attribute.
21934 (crypto_sm4): Ditto.
21935 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
21936 (vsm3tt1aq_u32): Ditto.
21937 (vsm3tt1bq_u32): Ditto.
21938 (vsm3tt2aq_u32): Ditto.
21939 (vsm3tt2bq_u32): Ditto.
21940 (vsm3partw1q_u32): Ditto.
21941 (vsm3partw2q_u32): Ditto.
21942 (vsm4eq_u32): Ditto.
21943 (vsm4ekeyq_u32): Ditto.
21944 (doc/invoke.texi): Document new sm4 option.
21946 2018-01-10 Michael Collison <michael.collison@arm.com>
21948 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
21949 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
21950 (AARCH64_FL_FOR_ARCH8_4): New.
21951 (AARCH64_FL_V8_4): New flag.
21952 (doc/invoke.texi): Document new armv8.4-a option.
21954 2018-01-10 Michael Collison <michael.collison@arm.com>
21956 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21957 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21958 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21959 * config/aarch64/aarch64-option-extension.def: Add
21960 AARCH64_OPT_EXTENSION of 'sha2'.
21961 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21962 (crypto): Disable sha2 and aes if crypto disabled.
21963 (crypto): Enable aes and sha2 if enabled.
21964 (simd): Disable sha2 and aes if simd disabled.
21965 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21967 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21968 (TARGET_SHA2): New feature flag for sha2.
21969 (TARGET_AES): New feature flag for aes.
21970 * config/aarch64/aarch64-simd.md:
21971 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21972 conditional on TARGET_AES.
21973 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21974 (aarch64_crypto_sha1hsi): Make pattern conditional
21976 (aarch64_crypto_sha1hv4si): Ditto.
21977 (aarch64_be_crypto_sha1hv4si): Ditto.
21978 (aarch64_crypto_sha1su1v4si): Ditto.
21979 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21980 (aarch64_crypto_sha1su0v4si): Ditto.
21981 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21982 (aarch64_crypto_sha256su0v4si): Ditto.
21983 (aarch64_crypto_sha256su1v4si): Ditto.
21984 (doc/invoke.texi): Document new aes and sha2 options.
21986 2018-01-10 Martin Sebor <msebor@redhat.com>
21988 PR tree-optimization/83781
21989 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21992 2018-01-11 Martin Sebor <msebor@gmail.com>
21993 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21995 PR tree-optimization/83501
21996 PR tree-optimization/81703
21998 * tree-ssa-strlen.c (get_string_cst): Rename...
21999 (get_string_len): ...to this. Handle global constants.
22000 (handle_char_store): Adjust.
22002 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22003 Jim Wilson <jimw@sifive.com>
22005 * config/riscv/riscv-protos.h (riscv_output_return): New.
22006 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22007 (riscv_attribute_table, riscv_output_return),
22008 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22009 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22010 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22012 (riscv_expand_prologue): Add early return for naked function.
22013 (riscv_expand_epilogue): Likewise.
22014 (riscv_function_ok_for_sibcall): Return false for naked function.
22015 (riscv_set_current_function): New.
22016 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22017 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22018 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22019 * doc/extend.texi (RISC-V Function Attributes): New.
22021 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22023 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22024 check for 128-bit long double before checking TCmode.
22025 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22026 128-bit long doubles before checking TFmode or TCmode.
22027 (FLOAT128_IBM_P): Likewise.
22029 2018-01-10 Martin Sebor <msebor@redhat.com>
22031 PR tree-optimization/83671
22032 * builtins.c (c_strlen): Unconditionally return zero for the empty
22034 Use -Warray-bounds for warnings.
22035 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22036 for non-constant array indices with COMPONENT_REF, arrays of
22037 arrays, and pointers to arrays.
22038 (gimple_fold_builtin_strlen): Determine and set length range for
22039 non-constant character arrays.
22041 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22043 PR middle-end/81897
22044 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22047 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22049 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22051 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22054 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22055 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22056 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22057 indexed_or_indirect_operand predicate.
22058 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22059 (*vsx_le_perm_load_v8hi): Likewise.
22060 (*vsx_le_perm_load_v16qi): Likewise.
22061 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22062 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22063 (*vsx_le_perm_store_v8hi): Likewise.
22064 (*vsx_le_perm_store_v16qi): Likewise.
22065 (eight unnamed splitters): Likewise.
22067 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22069 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22070 * config/rs6000/emmintrin.h: Likewise.
22071 * config/rs6000/mmintrin.h: Likewise.
22072 * config/rs6000/xmmintrin.h: Likewise.
22074 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22077 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22079 * tree.c (tree_nop_conversion): Return true for location wrapper
22081 (maybe_wrap_with_location): New function.
22082 (selftest::check_strip_nops): New function.
22083 (selftest::test_location_wrappers): New function.
22084 (selftest::tree_c_tests): Call it.
22085 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22086 (maybe_wrap_with_location): New decl.
22087 (EXPR_LOCATION_WRAPPER_P): New macro.
22088 (location_wrapper_p): New inline function.
22089 (tree_strip_any_location_wrapper): New inline function.
22091 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22094 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22095 stack_realign_offset for the largest alignment of stack slot
22097 (ix86_find_max_used_stack_alignment): New function.
22098 (ix86_finalize_stack_frame_flags): Use it. Set
22099 max_used_stack_alignment if we don't realign stack.
22100 * config/i386/i386.h (machine_function): Add
22101 max_used_stack_alignment.
22103 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22105 * config/arm/arm.opt (-mbranch-cost): New option.
22106 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22109 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22112 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22113 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22115 2018-01-10 Richard Biener <rguenther@suse.de>
22118 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22119 early out so it also covers the case where we have a non-NULL
22122 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22124 PR tree-optimization/83753
22125 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22126 for non-strided grouped accesses if the number of elements is 1.
22128 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22131 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22132 * i386.h (TARGET_USE_GATHER): Define.
22133 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22135 2018-01-10 Martin Liska <mliska@suse.cz>
22138 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22139 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22141 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22142 CLEANUP_NO_PARTITIONING is not set.
22144 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22146 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22147 for vectors, as a partial revert of r254296.
22148 * rtl.h (const_vec_p): Delete.
22149 (const_vec_duplicate_p): Don't test for vector CONSTs.
22150 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22151 * expmed.c (make_tree): Likewise.
22154 * common.md (E, F): Use CONSTANT_P instead of checking for
22156 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22157 checking for CONST_VECTOR.
22159 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22161 PR middle-end/83575
22162 * predict.c (force_edge_cold): Handle in more sane way edges
22163 with no prediction.
22165 2018-01-09 Carl Love <cel@us.ibm.com>
22167 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22169 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22170 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22171 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22172 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22173 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22174 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22175 * config/rs6000/rs6000-protos.h: Add extern defition for
22176 rs6000_generate_float2_double_code.
22177 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22179 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22180 (float2_v2df): Add define_expand.
22182 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22185 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22186 op_mode in the force_to_mode call.
22188 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22190 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
22191 instead of checking each element individually.
22192 (aarch64_evpc_uzp): Likewise.
22193 (aarch64_evpc_zip): Likewise.
22194 (aarch64_evpc_ext): Likewise.
22195 (aarch64_evpc_rev): Likewise.
22196 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
22197 instead of checking each element individually. Return true without
22199 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
22200 whether all selected elements come from the same input, instead of
22201 checking each element individually. Remove calls to gen_rtx_REG,
22202 start_sequence and end_sequence and instead assert that no rtl is
22205 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22207 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
22208 order of HIGH and CONST checks.
22210 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22212 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
22213 if the destination isn't an SSA_NAME.
22215 2018-01-09 Richard Biener <rguenther@suse.de>
22217 PR tree-optimization/83668
22218 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
22220 (canonicalize_loop_form): ... here, renamed from ...
22221 (canonicalize_loop_closed_ssa_form): ... this and amended to
22222 swap successor edges for loop exit blocks to make us use
22223 the RPO order we need for initial schedule generation.
22225 2018-01-09 Joseph Myers <joseph@codesourcery.com>
22227 PR tree-optimization/64811
22228 * match.pd: When optimizing comparisons with Inf, avoid
22229 introducing or losing exceptions from comparisons with NaN.
22231 2018-01-09 Martin Liska <mliska@suse.cz>
22234 * asan.c (shadow_mem_size): Add gcc_assert.
22236 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
22238 Don't save registers in main().
22241 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
22242 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
22243 * config/avr/avr.c (avr_set_current_function): Don't error if
22244 naked, OS_task or OS_main are specified at the same time.
22245 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
22247 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
22249 * common/config/avr/avr-common.c (avr_option_optimization_table):
22250 Switch on -mmain-is-OS_task for optimizing compilations.
22252 2018-01-09 Richard Biener <rguenther@suse.de>
22254 PR tree-optimization/83572
22255 * graphite.c: Include cfganal.h.
22256 (graphite_transform_loops): Connect infinite loops to exit
22257 and remove fake edges at the end.
22259 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22261 * ipa-inline.c (edge_badness): Revert accidental checkin.
22263 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22266 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
22267 symbols; not inline clones.
22269 2018-01-09 Jakub Jelinek <jakub@redhat.com>
22272 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
22273 hard registers. Formatting fixes.
22275 PR preprocessor/83722
22276 * gcc.c (try_generate_repro): Pass
22277 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
22278 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
22281 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
22282 Kito Cheng <kito.cheng@gmail.com>
22284 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
22285 (riscv_leaf_function_p): Delete.
22286 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
22288 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22290 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
22292 (do_ifelse): New function.
22293 (do_isel): New function.
22294 (do_sub3): New function.
22295 (do_add3): New function.
22296 (do_load_mask_compare): New function.
22297 (do_overlap_load_compare): New function.
22298 (expand_compare_loop): New function.
22299 (expand_block_compare): Call expand_compare_loop() when appropriate.
22300 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
22301 option description.
22302 (-mblock-compare-inline-loop-limit): New option.
22304 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22307 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
22308 Reverse order of second and third operands in first alternative.
22309 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
22310 of first and second elements in UNSPEC_VPERMR vector.
22311 (altivec_expand_vec_perm_le): Likewise.
22313 2018-01-08 Jeff Law <law@redhat.com>
22315 PR rtl-optimizatin/81308
22316 * tree-switch-conversion.c (cfg_altered): New file scoped static.
22317 (process_switch): If group_case_labels makes a change, then set
22319 (pass_convert_switch::execute): If a switch is converted, then
22320 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
22322 PR rtl-optimization/81308
22323 * recog.c (split_all_insns): Conditionally cleanup the CFG after
22326 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
22328 PR target/83663 - Revert r255946
22329 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
22330 generation for cases where splatting a value is not useful.
22331 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
22332 across a vec_duplicate and a paradoxical subreg forming a vector
22333 mode to a vec_concat.
22335 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22337 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
22338 -march=armv8.3-a variants.
22339 * config/arm/t-multilib: Likewise.
22340 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
22342 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22344 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
22346 (cceq_ior_compare_complement): Give it a name so I can use it, and
22347 change boolean_or_operator predicate to boolean_operator so it can
22348 be used to generate a crand.
22349 (eqne): New code iterator.
22350 (bd/bd_neg): New code_attrs.
22351 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
22352 a single define_insn.
22353 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
22354 decrement (bdnzt/bdnzf/bdzt/bdzf).
22355 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
22356 with the new names of the branch decrement patterns, and added the
22357 names of the branch decrement conditional patterns.
22359 2018-01-08 Richard Biener <rguenther@suse.de>
22361 PR tree-optimization/83563
22362 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
22365 2018-01-08 Richard Biener <rguenther@suse.de>
22367 PR middle-end/83713
22368 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
22370 2018-01-08 Richard Biener <rguenther@suse.de>
22372 PR tree-optimization/83685
22373 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
22374 references to abnormals.
22376 2018-01-08 Richard Biener <rguenther@suse.de>
22379 * dwarf2out.c (output_indirect_strings): Handle empty
22380 skeleton_debug_str_hash.
22381 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
22383 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22385 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
22386 (emit_store_direct): Likewise.
22387 (arc_trampoline_adjust_address): Likewise.
22388 (arc_asm_trampoline_template): New function.
22389 (arc_initialize_trampoline): Use asm_trampoline_template.
22390 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
22391 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
22392 * config/arc/arc.md (flush_icache): Delete pattern.
22394 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22396 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
22397 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
22400 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22403 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
22404 by not USED_FOR_TARGET.
22405 (make_pass_resolve_sw_modes): Likewise.
22407 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22409 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
22412 2018-01-08 Richard Biener <rguenther@suse.de>
22414 PR middle-end/83580
22415 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
22417 2018-01-08 Richard Biener <rguenther@suse.de>
22419 PR middle-end/83517
22420 * match.pd ((t * 2) / 2) -> t): Add missing :c.
22422 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
22424 PR middle-end/81897
22425 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
22426 basic blocks with a small number of successors.
22427 (convert_control_dep_chain_into_preds): Improve handling of
22429 (dump_predicates): Split apart into...
22430 (dump_pred_chain): ...here...
22431 (dump_pred_info): ...and here.
22432 (can_one_predicate_be_invalidated_p): Add debugging printfs.
22433 (can_chain_union_be_invalidated_p): Improve check for invalidation
22435 (uninit_uses_cannot_happen): Avoid unnecessary if
22436 convert_control_dep_chain_into_preds yielded nothing.
22438 2018-01-06 Martin Sebor <msebor@redhat.com>
22440 PR tree-optimization/83640
22441 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
22442 subtracting negative offset from size.
22443 (builtin_access::overlap): Adjust offset bounds of the access to fall
22444 within the size of the object if possible.
22446 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
22448 PR rtl-optimization/83699
22449 * expmed.c (extract_bit_field_1): Restrict the vector usage of
22450 extract_bit_field_as_subreg to cases in which the extracted
22451 value is also a vector.
22453 * lra-constraints.c (process_alt_operands): Test for the equivalence
22454 substitutions when detecting a possible reload cycle.
22456 2018-01-06 Jakub Jelinek <jakub@redhat.com>
22459 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
22460 by default if flag_selective_schedling{,2}. Formatting fixes.
22462 PR rtl-optimization/83682
22463 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
22464 if it has non-VECTOR_MODE element mode.
22465 (vec_duplicate_p): Likewise.
22467 PR middle-end/83694
22468 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
22469 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
22471 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22474 * config/i386/i386-builtin.def
22475 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
22476 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
22477 Require also OPTION_MASK_ISA_AVX512F in addition to
22478 OPTION_MASK_ISA_GFNI.
22479 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
22480 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
22481 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
22482 to OPTION_MASK_ISA_GFNI.
22483 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
22484 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
22485 OPTION_MASK_ISA_AVX512BW.
22486 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
22487 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
22488 addition to OPTION_MASK_ISA_GFNI.
22489 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
22490 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
22491 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
22492 to OPTION_MASK_ISA_GFNI.
22493 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
22494 a requirement for all ISAs rather than any of them with a few
22496 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
22498 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
22499 bitmasks to be enabled with 3 exceptions, instead of requiring any
22500 enabled ISA with lots of exceptions.
22501 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
22502 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
22503 Change avx512bw in isa attribute to avx512f.
22504 * config/i386/sgxintrin.h: Add license boilerplate.
22505 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
22506 to __AVX512F__ and __AVX512VL to __AVX512VL__.
22507 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
22508 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
22510 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
22511 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
22512 temporarily sse2 rather than sse if not enabled already.
22515 * config/i386/sse.md (VI248_VLBW): Rename to ...
22516 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
22517 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
22518 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
22519 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
22520 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
22521 mode iterator instead of VI248_VLBW.
22523 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
22525 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
22526 (record_modified): Skip clobbers; add debug output.
22527 (param_change_prob): Use sreal frequencies.
22529 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
22531 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
22532 punt for user-aligned variables.
22534 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
22536 * tree-chrec.c (chrec_contains_symbols): Return true for
22539 2018-01-05 Sudakshina Das <sudi.das@arm.com>
22542 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
22543 of (x|y) == x for BICS pattern.
22545 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22547 PR tree-optimization/83605
22548 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
22549 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
22552 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
22554 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
22555 * config/epiphany/rtems.h: New file.
22557 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22558 Uros Bizjak <ubizjak@gmail.com>
22561 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
22562 QIreg_operand instead of register_operand predicate.
22563 * config/i386/i386.c (ix86_rop_should_change_byte_p,
22564 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
22565 comments instead of -fmitigate[-_]rop.
22567 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22570 * cgraphunit.c (symbol_table::compile): Switch to text_section
22571 before calling assembly_start debug hook.
22572 * run-rtl-passes.c (run_rtl_passes): Likewise.
22575 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22577 * tree-vrp.c (extract_range_from_binary_expr_1): Check
22578 range_int_cst_p rather than !symbolic_range_p before calling
22579 extract_range_from_multiplicative_op_1.
22581 2018-01-04 Jeff Law <law@redhat.com>
22583 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
22584 redundant test in assertion.
22586 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22588 * doc/rtl.texi: Document machine_mode wrapper classes.
22590 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22592 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
22593 using tree_to_uhwi.
22595 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22597 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
22598 the VEC_PERM_EXPR fold to fail.
22600 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22603 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
22604 to switched_sections.
22606 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22609 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
22610 test for d.testing.
22612 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
22615 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
22616 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
22618 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22621 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
22622 is BLKmode and bitpos not zero or mode change is needed.
22624 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22627 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
22630 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
22633 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
22634 instead of MULT rtx. Update all corresponding splitters.
22635 (*saddl_se): Ditto.
22636 (*ssub<modesuffix>): Ditto.
22637 (*ssubl_se): Ditto.
22638 (*cmp_sadd_di): Update split patterns.
22639 (*cmp_sadd_si): Ditto.
22640 (*cmp_sadd_sidi): Ditto.
22641 (*cmp_ssub_di): Ditto.
22642 (*cmp_ssub_si): Ditto.
22643 (*cmp_ssub_sidi): Ditto.
22644 * config/alpha/predicates.md (const23_operand): New predicate.
22645 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
22646 Look for ASHIFT, not MULT inner operand.
22647 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
22649 2018-01-04 Martin Liska <mliska@suse.cz>
22651 PR gcov-profile/83669
22652 * gcov.c (output_intermediate_file): Add version to intermediate
22654 * doc/gcov.texi: Document new field 'version' in intermediate
22655 file format. Fix location of '-k' option of gcov command.
22657 2018-01-04 Martin Liska <mliska@suse.cz>
22660 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
22662 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22664 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
22666 2018-01-03 Martin Sebor <msebor@redhat.com>
22668 PR tree-optimization/83655
22669 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
22670 checking calls with invalid arguments.
22672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22674 * tree-vect-stmts.c (vect_get_store_rhs): New function.
22675 (vectorizable_mask_load_store): Delete.
22676 (vectorizable_call): Return false for masked loads and stores.
22677 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
22678 instead of gimple_assign_rhs1.
22679 (vectorizable_load): Handle IFN_MASK_LOAD.
22680 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
22682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22684 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
22686 (vectorizable_mask_load_store): ...here.
22687 (vectorizable_load): ...and here.
22689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22691 * tree-vect-stmts.c (vect_build_all_ones_mask)
22692 (vect_build_zero_merge_argument): New functions, split out from...
22693 (vectorizable_load): ...here.
22695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22697 * tree-vect-stmts.c (vect_check_store_rhs): New function,
22699 (vectorizable_mask_load_store): ...here.
22700 (vectorizable_store): ...and here.
22702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22704 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
22706 (vectorizable_mask_load_store): ...here.
22708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22710 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
22711 (vect_model_store_cost): Take a vec_load_store_type instead of a
22713 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
22714 (vect_model_store_cost): Take a vec_load_store_type instead of a
22716 (vectorizable_mask_load_store): Update accordingly.
22717 (vectorizable_store): Likewise.
22718 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
22720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22722 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
22723 IFN_MASK_LOAD calls here rather than...
22724 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
22726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22727 Alan Hayward <alan.hayward@arm.com>
22728 David Sherwood <david.sherwood@arm.com>
22730 * expmed.c (extract_bit_field_1): For vector extracts,
22731 fall back to extract_bit_field_as_subreg if vec_extract
22734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22735 Alan Hayward <alan.hayward@arm.com>
22736 David Sherwood <david.sherwood@arm.com>
22738 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
22739 they are variable or constant sized.
22740 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
22741 slots for constant-sized data.
22743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22744 Alan Hayward <alan.hayward@arm.com>
22745 David Sherwood <david.sherwood@arm.com>
22747 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
22748 handling COND_EXPRs with boolean comparisons, try to find a better
22749 basis for the mask type than the boolean itself.
22751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22753 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
22754 is calculated and how it can be overridden.
22755 * genmodes.c (max_bitsize_mode_any_mode): New variable.
22756 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
22758 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
22761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22762 Alan Hayward <alan.hayward@arm.com>
22763 David Sherwood <david.sherwood@arm.com>
22765 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
22766 Remove the mode argument.
22767 (aarch64_simd_valid_immediate): Remove the mode and inverse
22769 * config/aarch64/iterators.md (bitsize): New iterator.
22770 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
22771 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
22772 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
22773 aarch64_simd_valid_immediate.
22774 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
22775 (aarch64_reg_or_bic_imm): Likewise.
22776 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
22777 with an insn_type enum and msl with a modifier_type enum.
22778 Replace element_width with a scalar_mode. Change the shift
22779 to unsigned int. Add constructors for scalar_float_mode and
22780 scalar_int_mode elements.
22781 (aarch64_vect_float_const_representable_p): Delete.
22782 (aarch64_can_const_movi_rtx_p)
22783 (aarch64_simd_scalar_immediate_valid_for_move)
22784 (aarch64_simd_make_constant): Update call to
22785 aarch64_simd_valid_immediate.
22786 (aarch64_advsimd_valid_immediate_hs): New function.
22787 (aarch64_advsimd_valid_immediate): Likewise.
22788 (aarch64_simd_valid_immediate): Remove mode and inverse
22789 arguments. Rewrite to use the above. Use const_vec_duplicate_p
22790 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
22791 and aarch64_float_const_representable_p on the result.
22792 (aarch64_output_simd_mov_immediate): Remove mode argument.
22793 Update call to aarch64_simd_valid_immediate and use of
22794 simd_immediate_info.
22795 (aarch64_output_scalar_simd_mov_immediate): Update call
22798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22799 Alan Hayward <alan.hayward@arm.com>
22800 David Sherwood <david.sherwood@arm.com>
22802 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
22803 (mode_nunits): Likewise CONST_MODE_NUNITS.
22804 * machmode.def (ADJUST_NUNITS): Document.
22805 * genmodes.c (mode_data::need_nunits_adj): New field.
22806 (blank_mode): Update accordingly.
22807 (adj_nunits): New variable.
22808 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
22810 (emit_mode_size_inline): Set need_bytesize_adj for all modes
22811 listed in adj_nunits.
22812 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
22813 listed in adj_nunits. Don't emit case statements for such modes.
22814 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
22815 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
22816 nothing if adj_nunits is nonnull.
22817 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
22818 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
22819 (emit_mode_fbit): Update use of print_maybe_const_decl.
22820 (emit_move_size): Likewise. Treat the array as non-const
22822 (emit_mode_adjustments): Handle adj_nunits.
22824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22826 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
22827 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
22828 (VECTOR_MODES): Use it.
22829 (make_vector_modes): Take the prefix as an argument.
22831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22832 Alan Hayward <alan.hayward@arm.com>
22833 David Sherwood <david.sherwood@arm.com>
22835 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
22836 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
22837 for MODE_VECTOR_BOOL.
22838 * machmode.def (VECTOR_BOOL_MODE): Document.
22839 * genmodes.c (VECTOR_BOOL_MODE): New macro.
22840 (make_vector_bool_mode): New function.
22841 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
22843 * lto-streamer-in.c (lto_input_mode_table): Likewise.
22844 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
22846 * stor-layout.c (int_mode_for_mode): Likewise.
22847 * tree.c (build_vector_type_for_mode): Likewise.
22848 * varasm.c (output_constant_pool_2): Likewise.
22849 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
22850 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
22851 for MODE_VECTOR_BOOL.
22852 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
22853 of mode class checks.
22854 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
22855 instead of a list of mode class checks.
22856 (expand_vector_scalar_condition): Likewise.
22857 (type_for_widest_vector_mode): Handle BImode as an inner mode.
22859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22860 Alan Hayward <alan.hayward@arm.com>
22861 David Sherwood <david.sherwood@arm.com>
22863 * machmode.h (mode_size): Change from unsigned short to
22865 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
22866 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22867 or if measurement_type is not polynomial.
22868 (fixed_size_mode::includes_p): Check for constant-sized modes.
22869 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
22870 return a poly_uint16 rather than an unsigned short.
22871 (emit_mode_size): Change the type of mode_size from unsigned short
22872 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
22873 (emit_mode_adjustments): Cope with polynomial vector sizes.
22874 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22876 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22878 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
22879 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
22880 * caller-save.c (setup_save_areas): Likewise.
22881 (replace_reg_with_saved_mem): Likewise.
22882 * calls.c (emit_library_call_value_1): Likewise.
22883 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
22884 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
22885 (gen_lowpart_for_combine): Likewise.
22886 * convert.c (convert_to_integer_1): Likewise.
22887 * cse.c (equiv_constant, cse_insn): Likewise.
22888 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
22889 (cselib_subst_to_values): Likewise.
22890 * dce.c (word_dce_process_block): Likewise.
22891 * df-problems.c (df_word_lr_mark_ref): Likewise.
22892 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
22893 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
22894 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
22895 (rtl_for_decl_location): Likewise.
22896 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
22897 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
22898 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
22899 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
22900 (expand_expr_real_1): Likewise.
22901 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
22902 (pad_below): Likewise.
22903 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22904 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
22905 * ira.c (get_subreg_tracking_sizes): Likewise.
22906 * ira-build.c (ira_create_allocno_objects): Likewise.
22907 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
22908 (ira_sort_regnos_for_alter_reg): Likewise.
22909 * ira-costs.c (record_operand_costs): Likewise.
22910 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
22911 (resolve_simple_move): Likewise.
22912 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
22913 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
22914 (lra_constraints): Likewise.
22915 (CONST_POOL_OK_P): Reject variable-sized modes.
22916 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
22917 (add_pseudo_to_slot, lra_spill): Likewise.
22918 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22919 * optabs-query.c (get_best_extraction_insn): Likewise.
22920 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22921 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
22922 (expand_mult_highpart, valid_multiword_target_p): Likewise.
22923 * recog.c (offsettable_address_addr_space_p): Likewise.
22924 * regcprop.c (maybe_mode_change): Likewise.
22925 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
22926 * regrename.c (build_def_use): Likewise.
22927 * regstat.c (dump_reg_info): Likewise.
22928 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
22929 (find_reloads, find_reloads_subreg_address): Likewise.
22930 * reload1.c (eliminate_regs_1): Likewise.
22931 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
22932 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
22933 (simplify_binary_operation_1, simplify_subreg): Likewise.
22934 * targhooks.c (default_function_arg_padding): Likewise.
22935 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
22936 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
22937 (verify_gimple_assign_ternary): Likewise.
22938 * tree-inline.c (estimate_move_cost): Likewise.
22939 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22940 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
22941 (get_address_cost_ainc): Likewise.
22942 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
22943 (vect_supportable_dr_alignment): Likewise.
22944 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22945 (vectorizable_reduction): Likewise.
22946 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
22947 (vectorizable_operation, vectorizable_load): Likewise.
22948 * tree.c (build_same_sized_truth_vector_type): Likewise.
22949 * valtrack.c (cleanup_auto_inc_dec): Likewise.
22950 * var-tracking.c (emit_note_insn_var_location): Likewise.
22951 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
22952 (ADDR_VEC_ALIGN): Likewise.
22954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22955 Alan Hayward <alan.hayward@arm.com>
22956 David Sherwood <david.sherwood@arm.com>
22958 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22960 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22961 or if measurement_type is polynomial.
22962 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22963 * combine.c (make_extraction): Likewise.
22964 * dse.c (find_shift_sequence): Likewise.
22965 * dwarf2out.c (mem_loc_descriptor): Likewise.
22966 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22967 (extract_bit_field, extract_low_bits): Likewise.
22968 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22969 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22970 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22971 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22972 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22973 * reload.c (find_reloads): Likewise.
22974 * reload1.c (alter_reg): Likewise.
22975 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22976 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22977 * tree-if-conv.c (predicate_mem_writes): Likewise.
22978 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22979 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22980 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22981 * valtrack.c (dead_debug_insert_temp): Likewise.
22982 * varasm.c (mergeable_constant_section): Likewise.
22983 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22986 Alan Hayward <alan.hayward@arm.com>
22987 David Sherwood <david.sherwood@arm.com>
22989 * expr.c (expand_assignment): Cope with polynomial mode sizes
22990 when assigning to a CONCAT.
22992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22993 Alan Hayward <alan.hayward@arm.com>
22994 David Sherwood <david.sherwood@arm.com>
22996 * machmode.h (mode_precision): Change from unsigned short to
22998 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23000 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23001 or if measurement_type is not polynomial.
23002 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23003 in which the mode is already known to be a scalar_int_mode.
23004 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23005 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23007 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23008 for GET_MODE_PRECISION.
23009 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23010 for GET_MODE_PRECISION.
23011 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23013 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23014 (expand_field_assignment, make_extraction): Likewise.
23015 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23016 (get_last_value): Likewise.
23017 * convert.c (convert_to_integer_1): Likewise.
23018 * cse.c (cse_insn): Likewise.
23019 * expr.c (expand_expr_real_1): Likewise.
23020 * lra-constraints.c (simplify_operand_subreg): Likewise.
23021 * optabs-query.c (can_atomic_load_p): Likewise.
23022 * optabs.c (expand_atomic_load): Likewise.
23023 (expand_atomic_store): Likewise.
23024 * ree.c (combine_reaching_defs): Likewise.
23025 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23026 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23027 * tree.h (type_has_mode_precision_p): Likewise.
23028 * ubsan.c (instrument_si_overflow): Likewise.
23030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23031 Alan Hayward <alan.hayward@arm.com>
23032 David Sherwood <david.sherwood@arm.com>
23034 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23035 polynomial numbers of units.
23036 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23037 (valid_vector_subparts_p): New function.
23038 (build_vector_type): Remove temporary shim and take the number
23039 of units as a poly_uint64 rather than an int.
23040 (build_opaque_vector_type): Take the number of units as a
23041 poly_uint64 rather than an int.
23042 * tree.c (build_vector_from_ctor): Handle polynomial
23043 TYPE_VECTOR_SUBPARTS.
23044 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23045 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23046 (build_vector_from_val): If the number of units is variable,
23047 use build_vec_duplicate_cst for constant operands and
23048 VEC_DUPLICATE_EXPR otherwise.
23049 (make_vector_type): Remove temporary is_constant ().
23050 (build_vector_type, build_opaque_vector_type): Take the number of
23051 units as a poly_uint64 rather than an int.
23052 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23054 * cfgexpand.c (expand_debug_expr): Likewise.
23055 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23056 (store_constructor, expand_expr_real_1): Likewise.
23057 (const_scalar_mask_from_tree): Likewise.
23058 * fold-const-call.c (fold_const_reduction): Likewise.
23059 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23060 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23061 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23062 (fold_relational_const): Likewise.
23063 (native_interpret_vector): Likewise. Change the size from an
23064 int to an unsigned int.
23065 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23066 TYPE_VECTOR_SUBPARTS.
23067 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23068 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23069 duplicating a non-constant operand into a variable-length vector.
23070 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23071 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23072 * ipa-icf.c (sem_variable::equals): Likewise.
23073 * match.pd: Likewise.
23074 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23075 * print-tree.c (print_node): Likewise.
23076 * stor-layout.c (layout_type): Likewise.
23077 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23078 * tree-cfg.c (verify_gimple_comparison): Likewise.
23079 (verify_gimple_assign_binary): Likewise.
23080 (verify_gimple_assign_ternary): Likewise.
23081 (verify_gimple_assign_single): Likewise.
23082 * tree-pretty-print.c (dump_generic_node): Likewise.
23083 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23084 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23085 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23086 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23087 (vect_shift_permute_load_chain): Likewise.
23088 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23089 (expand_vector_condition, optimize_vector_constructor): Likewise.
23090 (lower_vec_perm, get_compute_type): Likewise.
23091 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23092 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23093 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23094 (vect_recog_mask_conversion_pattern): Likewise.
23095 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23096 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23097 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23098 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23099 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23100 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23101 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23102 (supportable_widening_operation): Likewise.
23103 (supportable_narrowing_operation): Likewise.
23104 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23106 * varasm.c (output_constant): Likewise.
23108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23109 Alan Hayward <alan.hayward@arm.com>
23110 David Sherwood <david.sherwood@arm.com>
23112 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23113 so that both the length == 3 and length != 3 cases set up their
23114 own permute vectors. Add comments explaining why we know the
23115 number of elements is constant.
23116 (vect_permute_load_chain): Likewise.
23118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23119 Alan Hayward <alan.hayward@arm.com>
23120 David Sherwood <david.sherwood@arm.com>
23122 * machmode.h (mode_nunits): Change from unsigned char to
23124 (ONLY_FIXED_SIZE_MODES): New macro.
23125 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23126 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23127 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23129 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23130 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23131 or if measurement_type is not polynomial.
23132 * genmodes.c (ZERO_COEFFS): New macro.
23133 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23135 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23136 Use ZERO_COEFFS when emitting initializers.
23137 * data-streamer.h (bp_pack_poly_value): New function.
23138 (bp_unpack_poly_value): Likewise.
23139 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23140 for GET_MODE_NUNITS.
23141 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23142 for GET_MODE_NUNITS.
23143 * tree.c (make_vector_type): Remove temporary shim and make
23144 the real function take the number of units as a poly_uint64
23145 rather than an int.
23146 (build_vector_type_for_mode): Handle polynomial nunits.
23147 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23148 * emit-rtl.c (const_vec_series_p_1): Likewise.
23149 (gen_rtx_CONST_VECTOR): Likewise.
23150 * fold-const.c (test_vec_duplicate_folding): Likewise.
23151 * genrecog.c (validate_pattern): Likewise.
23152 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23153 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23154 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23155 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23156 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23157 * rtlanal.c (subreg_get_info): Likewise.
23158 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23159 (vect_grouped_load_supported): Likewise.
23160 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23161 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23162 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23163 (simplify_const_unary_operation, simplify_binary_operation_1)
23164 (simplify_const_binary_operation, simplify_ternary_operation)
23165 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23166 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23167 instead of CONST_VECTOR_NUNITS.
23168 * varasm.c (output_constant_pool_2): Likewise.
23169 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23170 explicit-encoded elements in the XVEC for variable-length vectors.
23172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23174 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23177 Alan Hayward <alan.hayward@arm.com>
23178 David Sherwood <david.sherwood@arm.com>
23180 * coretypes.h (fixed_size_mode): Declare.
23181 (fixed_size_mode_pod): New typedef.
23182 * builtins.h (target_builtins::x_apply_args_mode)
23183 (target_builtins::x_apply_result_mode): Change type to
23184 fixed_size_mode_pod.
23185 * builtins.c (apply_args_size, apply_result_size, result_vector)
23186 (expand_builtin_apply_args_1, expand_builtin_apply)
23187 (expand_builtin_return): Update accordingly.
23189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23191 * cse.c (hash_rtx_cb): Hash only the encoded elements.
23192 * cselib.c (cselib_hash_rtx): Likewise.
23193 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
23194 CONST_VECTOR encoding.
23196 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23197 Jeff Law <law@redhat.com>
23200 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
23201 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
23202 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
23203 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
23206 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
23207 explicitly probe *sp in a noreturn function if there were any callee
23208 register saves or frame pointer is needed.
23210 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23213 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
23214 BLKmode for ternary, binary or unary expressions.
23217 * var-tracking.c (delete_vta_debug_insn): New inline function.
23218 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
23219 insns from get_insns () to NULL instead of each bb separately.
23220 Use delete_vta_debug_insn. No longer static.
23221 (vt_debug_insns_local, variable_tracking_main_1): Adjust
23222 delete_vta_debug_insns callers.
23223 * rtl.h (delete_vta_debug_insns): Declare.
23224 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
23225 instead of variable_tracking_main.
23227 2018-01-03 Martin Sebor <msebor@redhat.com>
23229 PR tree-optimization/83603
23230 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
23231 arguments past the endof the argument list in functions declared
23232 without a prototype.
23233 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
23234 Avoid checking when arguments are null.
23236 2018-01-03 Martin Sebor <msebor@redhat.com>
23239 * doc/extend.texi (attribute const): Fix a typo.
23240 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
23241 issuing -Wsuggest-attribute for void functions.
23243 2018-01-03 Martin Sebor <msebor@redhat.com>
23245 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
23246 offset_int::from instead of wide_int::to_shwi.
23247 (maybe_diag_overlap): Remove assertion.
23248 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
23249 * gimple-ssa-sprintf.c (format_directive): Same.
23250 (parse_directive): Same.
23251 (sprintf_dom_walker::compute_format_length): Same.
23252 (try_substitute_return_value): Same.
23254 2018-01-03 Jeff Law <law@redhat.com>
23256 PR middle-end/83654
23257 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
23258 non-constant residual for zero at runtime and avoid probing in
23259 that case. Reorganize code for trailing problem to mirror handling
23262 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23264 PR tree-optimization/83501
23265 * tree-ssa-strlen.c (get_string_cst): New.
23266 (handle_char_store): Call get_string_cst.
23268 2018-01-03 Martin Liska <mliska@suse.cz>
23270 PR tree-optimization/83593
23271 * tree-ssa-strlen.c: Include tree-cfg.h.
23272 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
23273 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
23274 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
23276 (strlen_dom_walker::before_dom_children): Call
23277 gimple_purge_dead_eh_edges. Dump tranformation with details
23279 (strlen_dom_walker::before_dom_children): Update call by adding
23280 new argument cleanup_eh.
23281 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
23283 2018-01-03 Martin Liska <mliska@suse.cz>
23286 * cif-code.def (VARIADIC_THUNK): New enum value.
23287 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
23290 2018-01-03 Jan Beulich <jbeulich@suse.com>
23292 * sse.md (mov<mode>_internal): Tighten condition for when to use
23293 vmovdqu<ssescalarsize> for TI and OI modes.
23295 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23297 Update copyright years.
23299 2018-01-03 Martin Liska <mliska@suse.cz>
23302 * ipa-visibility.c (function_and_variable_visibility): Skip
23303 functions with noipa attribure.
23305 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23307 * gcc.c (process_command): Update copyright notice dates.
23308 * gcov-dump.c (print_version): Ditto.
23309 * gcov.c (print_version): Ditto.
23310 * gcov-tool.c (print_version): Ditto.
23311 * gengtype.c (create_file): Ditto.
23312 * doc/cpp.texi: Bump @copying's copyright year.
23313 * doc/cppinternals.texi: Ditto.
23314 * doc/gcc.texi: Ditto.
23315 * doc/gccint.texi: Ditto.
23316 * doc/gcov.texi: Ditto.
23317 * doc/install.texi: Ditto.
23318 * doc/invoke.texi: Ditto.
23320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23322 * vector-builder.h (vector_builder::m_full_nelts): Change from
23323 unsigned int to poly_uint64.
23324 (vector_builder::full_nelts): Update prototype accordingly.
23325 (vector_builder::new_vector): Likewise.
23326 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
23327 (vector_builder::operator ==): Likewise.
23328 (vector_builder::finalize): Likewise.
23329 * int-vector-builder.h (int_vector_builder::int_vector_builder):
23330 Take the number of elements as a poly_uint64 rather than an
23332 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
23333 from unsigned int to poly_uint64.
23334 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
23335 (vec_perm_indices::new_vector): Likewise.
23336 (vec_perm_indices::length): Likewise.
23337 (vec_perm_indices::nelts_per_input): Likewise.
23338 (vec_perm_indices::input_nelts): Likewise.
23339 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
23340 number of elements per input as a poly_uint64 rather than an
23341 unsigned int. Use the original encoding for variable-length
23342 vectors, rather than clamping each individual element.
23343 For the second and subsequent elements in each pattern,
23344 clamp the step and base before clamping their sum.
23345 (vec_perm_indices::series_p): Handle polynomial element counts.
23346 (vec_perm_indices::all_in_range_p): Likewise.
23347 (vec_perm_indices_to_tree): Likewise.
23348 (vec_perm_indices_to_rtx): Likewise.
23349 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
23350 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
23351 (tree_vector_builder::new_binary_operation): Handle polynomial
23352 element counts. Return false if we need to know the number
23353 of elements at compile time.
23354 * fold-const.c (fold_vec_perm): Punt if the number of elements
23355 isn't known at compile time.
23357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23359 * vec-perm-indices.h (vec_perm_builder): Change element type
23360 from HOST_WIDE_INT to poly_int64.
23361 (vec_perm_indices::element_type): Update accordingly.
23362 (vec_perm_indices::clamp): Handle polynomial element_types.
23363 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23364 (vec_perm_indices::all_in_range_p): Likewise.
23365 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
23367 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
23368 polynomial vec_perm_indices element types.
23369 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
23370 * fold-const.c (fold_vec_perm): Likewise.
23371 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
23372 * tree-vect-generic.c (lower_vec_perm): Likewise.
23373 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23374 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
23375 element type to HOST_WIDE_INT.
23377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23378 Alan Hayward <alan.hayward@arm.com>
23379 David Sherwood <david.sherwood@arm.com>
23381 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
23382 rather than an int. Use plus_constant.
23383 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
23384 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
23386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23387 Alan Hayward <alan.hayward@arm.com>
23388 David Sherwood <david.sherwood@arm.com>
23390 * calls.c (emit_call_1, expand_call): Change struct_value_size from
23391 a HOST_WIDE_INT to a poly_int64.
23393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23394 Alan Hayward <alan.hayward@arm.com>
23395 David Sherwood <david.sherwood@arm.com>
23397 * calls.c (load_register_parameters): Cope with polynomial
23398 mode sizes. Require a constant size for BLKmode parameters
23399 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
23400 forces a parameter to be padded at the lsb end in order to
23401 fill a complete number of words, require the parameter size
23402 to be ordered wrt UNITS_PER_WORD.
23404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23405 Alan Hayward <alan.hayward@arm.com>
23406 David Sherwood <david.sherwood@arm.com>
23408 * reload1.c (spill_stack_slot_width): Change element type
23409 from unsigned int to poly_uint64_pod.
23410 (alter_reg): Treat mode sizes as polynomial.
23412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23413 Alan Hayward <alan.hayward@arm.com>
23414 David Sherwood <david.sherwood@arm.com>
23416 * reload.c (complex_word_subreg_p): New function.
23417 (reload_inner_reg_of_subreg, push_reload): Use it.
23419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23420 Alan Hayward <alan.hayward@arm.com>
23421 David Sherwood <david.sherwood@arm.com>
23423 * lra-constraints.c (process_alt_operands): Reject matched
23424 operands whose sizes aren't ordered.
23425 (match_reload): Refer to this check here.
23427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23428 Alan Hayward <alan.hayward@arm.com>
23429 David Sherwood <david.sherwood@arm.com>
23431 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
23432 that the mode size is in the set {1, 2, 4, 8, 16}.
23434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23435 Alan Hayward <alan.hayward@arm.com>
23436 David Sherwood <david.sherwood@arm.com>
23438 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
23439 Use plus_constant instead of gen_rtx_PLUS.
23441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23442 Alan Hayward <alan.hayward@arm.com>
23443 David Sherwood <david.sherwood@arm.com>
23445 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
23446 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
23447 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
23448 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
23449 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
23450 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
23451 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
23452 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
23453 * config/i386/i386.c (ix86_push_rounding): ...this new function.
23454 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
23456 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
23457 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
23458 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
23459 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
23460 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
23461 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
23462 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
23463 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
23464 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
23465 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
23467 * expr.c (emit_move_resolve_push): Treat the input and result
23468 of PUSH_ROUNDING as a poly_int64.
23469 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
23470 (emit_push_insn): Likewise.
23471 * lra-eliminations.c (mark_not_eliminable): Likewise.
23472 * recog.c (push_operand): Likewise.
23473 * reload1.c (elimination_effects): Likewise.
23474 * rtlanal.c (nonzero_bits1): Likewise.
23475 * calls.c (store_one_arg): Likewise. Require the padding to be
23476 known at compile time.
23478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23479 Alan Hayward <alan.hayward@arm.com>
23480 David Sherwood <david.sherwood@arm.com>
23482 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
23483 Use plus_constant instead of gen_rtx_PLUS.
23485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23486 Alan Hayward <alan.hayward@arm.com>
23487 David Sherwood <david.sherwood@arm.com>
23489 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
23490 rather than an int.
23492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23493 Alan Hayward <alan.hayward@arm.com>
23494 David Sherwood <david.sherwood@arm.com>
23496 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
23497 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
23498 via stack temporaries. Treat the mode size as polynomial too.
23500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23501 Alan Hayward <alan.hayward@arm.com>
23502 David Sherwood <david.sherwood@arm.com>
23504 * expr.c (expand_expr_real_2): When handling conversions involving
23505 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
23506 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
23507 as a poly_uint64 too.
23509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23510 Alan Hayward <alan.hayward@arm.com>
23511 David Sherwood <david.sherwood@arm.com>
23513 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
23515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23516 Alan Hayward <alan.hayward@arm.com>
23517 David Sherwood <david.sherwood@arm.com>
23519 * combine.c (can_change_dest_mode): Handle polynomial
23520 REGMODE_NATURAL_SIZE.
23521 * expmed.c (store_bit_field_1): Likewise.
23522 * expr.c (store_constructor): Likewise.
23523 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
23524 and polynomial REGMODE_NATURAL_SIZE.
23525 (gen_lowpart_common): Likewise.
23526 * reginfo.c (record_subregs_of_mode): Likewise.
23527 * rtlanal.c (read_modify_subreg_p): Likewise.
23529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23530 Alan Hayward <alan.hayward@arm.com>
23531 David Sherwood <david.sherwood@arm.com>
23533 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
23534 numbers of elements.
23536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23537 Alan Hayward <alan.hayward@arm.com>
23538 David Sherwood <david.sherwood@arm.com>
23540 * match.pd: Cope with polynomial numbers of vector elements.
23542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23543 Alan Hayward <alan.hayward@arm.com>
23544 David Sherwood <david.sherwood@arm.com>
23546 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
23547 in a POINTER_PLUS_EXPR.
23549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23550 Alan Hayward <alan.hayward@arm.com>
23551 David Sherwood <david.sherwood@arm.com>
23553 * omp-simd-clone.c (simd_clone_subparts): New function.
23554 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
23555 (ipa_simd_modify_function_body): Likewise.
23557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23558 Alan Hayward <alan.hayward@arm.com>
23559 David Sherwood <david.sherwood@arm.com>
23561 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
23562 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
23563 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
23564 (expand_vector_condition, vector_element): Likewise.
23565 (subparts_gt): New function.
23566 (get_compute_type): Use subparts_gt.
23567 (count_type_subparts): Delete.
23568 (expand_vector_operations_1): Use subparts_gt instead of
23569 count_type_subparts.
23571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23572 Alan Hayward <alan.hayward@arm.com>
23573 David Sherwood <david.sherwood@arm.com>
23575 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
23576 (vect_compile_time_alias): ...this new function. Do the calculation
23577 on poly_ints rather than trees.
23578 (vect_prune_runtime_alias_test_list): Update call accordingly.
23580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23581 Alan Hayward <alan.hayward@arm.com>
23582 David Sherwood <david.sherwood@arm.com>
23584 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
23586 (vect_schedule_slp_instance): Likewise.
23588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23589 Alan Hayward <alan.hayward@arm.com>
23590 David Sherwood <david.sherwood@arm.com>
23592 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
23593 constant and extern definitions for variable-length vectors.
23594 (vect_get_constant_vectors): Note that the number of units
23595 is known to be constant.
23597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23598 Alan Hayward <alan.hayward@arm.com>
23599 David Sherwood <david.sherwood@arm.com>
23601 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
23602 of units as polynomial. Choose between WIDE and NARROW based
23605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23606 Alan Hayward <alan.hayward@arm.com>
23607 David Sherwood <david.sherwood@arm.com>
23609 * tree-vect-stmts.c (simd_clone_subparts): New function.
23610 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
23612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23613 Alan Hayward <alan.hayward@arm.com>
23614 David Sherwood <david.sherwood@arm.com>
23616 * tree-vect-stmts.c (vectorizable_call): Treat the number of
23617 vectors as polynomial. Use build_index_vector for
23618 IFN_GOMP_SIMD_LANE.
23620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23621 Alan Hayward <alan.hayward@arm.com>
23622 David Sherwood <david.sherwood@arm.com>
23624 * tree-vect-stmts.c (get_load_store_type): Treat the number of
23625 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
23626 for variable-length vectors.
23627 (vectorizable_mask_load_store): Treat the number of units as
23628 polynomial, asserting that it is constant if the condition has
23629 already been enforced.
23630 (vectorizable_store, vectorizable_load): Likewise.
23632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23633 Alan Hayward <alan.hayward@arm.com>
23634 David Sherwood <david.sherwood@arm.com>
23636 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
23637 of units as polynomial. Punt if we can't tell at compile time
23638 which vector contains the final result.
23640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23641 Alan Hayward <alan.hayward@arm.com>
23642 David Sherwood <david.sherwood@arm.com>
23644 * tree-vect-loop.c (vectorizable_induction): Treat the number
23645 of units as polynomial. Punt on SLP inductions. Use an integer
23646 VEC_SERIES_EXPR for variable-length integer reductions. Use a
23647 cast of such a series for variable-length floating-point
23650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23651 Alan Hayward <alan.hayward@arm.com>
23652 David Sherwood <david.sherwood@arm.com>
23654 * tree.h (build_index_vector): Declare.
23655 * tree.c (build_index_vector): New function.
23656 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
23657 of units as polynomial, forcibly converting it to a constant if
23658 vectorizable_reduction has already enforced the condition.
23659 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
23660 to create a {1,2,3,...} vector.
23661 (vectorizable_reduction): Treat the number of units as polynomial.
23662 Choose vectype_in based on the largest scalar element size rather
23663 than the smallest number of units. Enforce the restrictions
23666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23667 Alan Hayward <alan.hayward@arm.com>
23668 David Sherwood <david.sherwood@arm.com>
23670 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
23671 number of units as polynomial.
23673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23674 Alan Hayward <alan.hayward@arm.com>
23675 David Sherwood <david.sherwood@arm.com>
23677 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
23678 * target.def (autovectorize_vector_sizes): Return the vector sizes
23679 by pointer, using vector_sizes rather than a bitmask.
23680 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
23681 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
23682 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
23684 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
23685 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
23686 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
23687 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
23688 * omp-general.c (omp_max_vf): Likewise.
23689 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23690 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
23691 * tree-vect-loop.c (vect_analyze_loop): Likewise.
23692 * tree-vect-slp.c (vect_slp_bb): Likewise.
23693 * doc/tm.texi: Regenerate.
23694 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23696 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
23697 the vector size as a poly_uint64 rather than an unsigned int.
23698 (current_vector_size): Change from an unsigned int to a poly_uint64.
23699 (get_vectype_for_scalar_type): Update accordingly.
23700 * tree.h (build_truth_vector_type): Take the size and number of
23701 units as a poly_uint64 rather than an unsigned int.
23702 (build_vector_type): Add a temporary overload that takes
23703 the number of units as a poly_uint64 rather than an unsigned int.
23704 * tree.c (make_vector_type): Likewise.
23705 (build_truth_vector_type): Take the number of units as a poly_uint64
23706 rather than an unsigned int.
23708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23709 Alan Hayward <alan.hayward@arm.com>
23710 David Sherwood <david.sherwood@arm.com>
23712 * target.def (get_mask_mode): Take the number of units and length
23713 as poly_uint64s rather than unsigned ints.
23714 * targhooks.h (default_get_mask_mode): Update accordingly.
23715 * targhooks.c (default_get_mask_mode): Likewise.
23716 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
23717 * doc/tm.texi: Regenerate.
23719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23720 Alan Hayward <alan.hayward@arm.com>
23721 David Sherwood <david.sherwood@arm.com>
23723 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
23724 * omp-general.c (omp_max_vf): Likewise.
23725 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
23726 (expand_omp_simd): Handle polynomial safelen.
23727 * omp-low.c (omplow_simd_context): Add a default constructor.
23728 (omplow_simd_context::max_vf): Change from int to poly_uint64.
23729 (lower_rec_simd_input_clauses): Update accordingly.
23730 (lower_rec_input_clauses): Likewise.
23732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23733 Alan Hayward <alan.hayward@arm.com>
23734 David Sherwood <david.sherwood@arm.com>
23736 * tree-vectorizer.h (vect_nunits_for_cost): New function.
23737 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
23738 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
23739 (vect_analyze_slp_cost): Likewise.
23740 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
23741 (vect_model_load_cost): Likewise.
23743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23744 Alan Hayward <alan.hayward@arm.com>
23745 David Sherwood <david.sherwood@arm.com>
23747 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
23748 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
23749 from an unsigned int * to a poly_uint64_pod *.
23750 (calculate_unrolling_factor): New function.
23751 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
23753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23754 Alan Hayward <alan.hayward@arm.com>
23755 David Sherwood <david.sherwood@arm.com>
23757 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
23758 from an unsigned int to a poly_uint64.
23759 (_loop_vec_info::slp_unrolling_factor): Likewise.
23760 (_loop_vec_info::vectorization_factor): Change from an int
23762 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
23763 (vect_get_num_vectors): New function.
23764 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
23765 (vect_get_num_copies): Use vect_get_num_vectors.
23766 (vect_analyze_data_ref_dependences): Change max_vf from an int *
23767 to an unsigned int *.
23768 (vect_analyze_data_refs): Change min_vf from an int * to a
23770 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23771 than an unsigned HOST_WIDE_INT.
23772 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
23773 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23774 to an unsigned int *.
23775 (vect_analyze_data_ref_dependences): Likewise.
23776 (vect_compute_data_ref_alignment): Handle polynomial vf.
23777 (vect_enhance_data_refs_alignment): Likewise.
23778 (vect_prune_runtime_alias_test_list): Likewise.
23779 (vect_shift_permute_load_chain): Likewise.
23780 (vect_supportable_dr_alignment): Likewise.
23781 (dependence_distance_ge_vf): Take the vectorization factor as a
23782 poly_uint64 rather than an unsigned HOST_WIDE_INT.
23783 (vect_analyze_data_refs): Change min_vf from an int * to a
23785 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
23786 vfm1 as a poly_uint64 rather than an int. Make the same change
23787 for the returned bound_scalar.
23788 (vect_gen_vector_loop_niters): Handle polynomial vf.
23789 (vect_do_peeling): Likewise. Update call to
23790 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
23791 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
23793 * tree-vect-loop.c (vect_determine_vectorization_factor)
23794 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
23795 (vect_get_known_peeling_cost): Likewise.
23796 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
23797 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
23798 (vect_transform_loop): Likewise. Use the lowest possible VF when
23799 updating the upper bounds of the loop.
23800 (vect_min_worthwhile_factor): Make static. Return an unsigned int
23801 rather than an int.
23802 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
23803 polynomial unroll factors.
23804 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
23805 (vect_make_slp_decision): Likewise.
23806 (vect_supported_load_permutation_p): Likewise, and polynomial
23808 (vect_analyze_slp_cost): Handle polynomial vf.
23809 (vect_slp_analyze_node_operations): Likewise.
23810 (vect_slp_analyze_bb_1): Likewise.
23811 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23812 than an unsigned HOST_WIDE_INT.
23813 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
23814 (vectorizable_load): Handle polynomial vf.
23815 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
23817 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
23819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23820 Alan Hayward <alan.hayward@arm.com>
23821 David Sherwood <david.sherwood@arm.com>
23823 * match.pd: Handle bit operations involving three constants
23824 and try to fold one pair.
23826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23828 * tree-vect-loop-manip.c: Include gimple-fold.h.
23829 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
23830 niters_maybe_zero parameters. Handle other cases besides a step of 1.
23831 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
23832 Add a path that uses a step of VF instead of 1, but disable it
23834 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
23835 and niters_no_overflow parameters. Update calls to
23836 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
23837 Create a new SSA name if the latter choses to use a ste other
23838 than zero, and return it via niters_vector_mult_vf_var.
23839 * tree-vect-loop.c (vect_transform_loop): Update calls to
23840 vect_do_peeling, vect_gen_vector_loop_niters and
23841 slpeel_make_loop_iterate_ntimes.
23842 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
23843 (vect_gen_vector_loop_niters): Update declarations after above changes.
23845 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
23847 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
23848 128-bit round to integer instructions.
23849 (ceil<mode>2): Likewise.
23850 (btrunc<mode>2): Likewise.
23851 (round<mode>2): Likewise.
23853 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23855 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
23856 unaligned VSX load/store on P8/P9.
23857 (expand_block_clear): Allow the use of unaligned VSX
23858 load/store on P8/P9.
23860 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23862 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
23864 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
23865 swap associated with both a load and a store.
23867 2018-01-02 Andrew Waterman <andrew@sifive.com>
23869 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
23870 * config/riscv/riscv.md (clear_cache): Use it.
23872 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
23874 * web.c: Remove out-of-date comment.
23876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23878 * expr.c (fixup_args_size_notes): Check that any existing
23879 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
23880 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
23881 (emit_single_push_insn): ...here.
23883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23885 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
23886 (const_vector_encoded_nelts): New function.
23887 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
23888 (const_vector_int_elt, const_vector_elt): Declare.
23889 * emit-rtl.c (const_vector_int_elt_1): New function.
23890 (const_vector_elt): Likewise.
23891 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
23892 of CONST_VECTOR_ELT.
23894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23896 * expr.c: Include rtx-vector-builder.h.
23897 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
23898 directly on the tree encoding.
23899 (const_vector_from_tree): Likewise.
23900 * optabs.c: Include rtx-vector-builder.h.
23901 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
23902 sequence of "u" values.
23903 * vec-perm-indices.c: Include rtx-vector-builder.h.
23904 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
23905 directly on the vec_perm_indices encoding.
23907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23909 * doc/rtl.texi (const_vector): Describe new encoding scheme.
23910 * Makefile.in (OBJS): Add rtx-vector-builder.o.
23911 * rtx-vector-builder.h: New file.
23912 * rtx-vector-builder.c: Likewise.
23913 * rtl.h (rtx_def::u2): Add a const_vector field.
23914 (CONST_VECTOR_NPATTERNS): New macro.
23915 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
23916 (CONST_VECTOR_DUPLICATE_P): Likewise.
23917 (CONST_VECTOR_STEPPED_P): Likewise.
23918 (CONST_VECTOR_ENCODED_ELT): Likewise.
23919 (const_vec_duplicate_p): Check for a duplicated vector encoding.
23920 (unwrap_const_vec_duplicate): Likewise.
23921 (const_vec_series_p): Check for a non-duplicated vector encoding.
23922 Say that the function only returns true for integer vectors.
23923 * emit-rtl.c: Include rtx-vector-builder.h.
23924 (gen_const_vec_duplicate_1): Delete.
23925 (gen_const_vector): Call gen_const_vec_duplicate instead of
23926 gen_const_vec_duplicate_1.
23927 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
23928 (gen_const_vec_duplicate): Use rtx_vector_builder.
23929 (gen_const_vec_series): Likewise.
23930 (gen_rtx_CONST_VECTOR): Likewise.
23931 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
23932 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23933 Build a new vector rather than modifying a CONST_VECTOR in-place.
23934 (handle_special_swappables): Update call accordingly.
23935 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
23936 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23937 Build a new vector rather than modifying a CONST_VECTOR in-place.
23938 (handle_special_swappables): Update call accordingly.
23940 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23942 * simplify-rtx.c (simplify_const_binary_operation): Use
23943 CONST_VECTOR_ELT instead of XVECEXP.
23945 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23947 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
23948 the selector elements to be different from the data elements
23949 if the selector is a VECTOR_CST.
23950 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
23951 ssizetype for the selector.
23953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23955 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23956 before testing each element individually.
23957 * tree-vect-generic.c (lower_vec_perm): Likewise.
23959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23961 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23962 * selftest-run-tests.c (selftest::run_tests): Call it.
23963 * vector-builder.h (vector_builder::operator ==): New function.
23964 (vector_builder::operator !=): Likewise.
23965 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23966 (vec_perm_indices::all_from_input_p): New function.
23967 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23968 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23969 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23970 instead of reading the VECTOR_CST directly. Detect whether both
23971 vector inputs are the same before constructing the vec_perm_indices,
23972 and update the number of inputs argument accordingly. Use the
23973 utility functions added above. Only construct sel2 if we need to.
23975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23977 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23978 the broadcast of the low byte.
23979 (expand_mult_highpart): Use an explicit encoding for the permutes.
23980 * optabs-query.c (can_mult_highpart_p): Likewise.
23981 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23982 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23983 (vectorizable_bswap): Likewise.
23984 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23985 explicit encoding for the power-of-2 permutes.
23986 (vect_permute_store_chain): Likewise.
23987 (vect_grouped_load_supported): Likewise.
23988 (vect_permute_load_chain): Likewise.
23990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23992 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23993 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23994 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23995 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23996 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23997 (vect_gen_perm_mask_any): Likewise.
23999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24001 * int-vector-builder.h: New file.
24002 * vec-perm-indices.h: Include int-vector-builder.h.
24003 (vec_perm_indices): Redefine as an int_vector_builder.
24004 (auto_vec_perm_indices): Delete.
24005 (vec_perm_builder): Redefine as a stand-alone class.
24006 (vec_perm_indices::vec_perm_indices): New function.
24007 (vec_perm_indices::clamp): Likewise.
24008 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24009 (vec_perm_indices::new_vector): New function.
24010 (vec_perm_indices::new_expanded_vector): Update for new
24011 vec_perm_indices class.
24012 (vec_perm_indices::rotate_inputs): New function.
24013 (vec_perm_indices::all_in_range_p): Operate directly on the
24014 encoded form, without computing elided elements.
24015 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24016 encoding. Update for new vec_perm_indices class.
24017 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24018 the given vec_perm_builder.
24019 (expand_vec_perm_var): Update vec_perm_builder constructor.
24020 (expand_mult_highpart): Use vec_perm_builder instead of
24021 auto_vec_perm_indices.
24022 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24023 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24024 or double series encoding as appropriate.
24025 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24026 vec_perm_indices instead of auto_vec_perm_indices.
24027 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24028 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24029 (vect_permute_store_chain): Likewise.
24030 (vect_grouped_load_supported): Likewise.
24031 (vect_permute_load_chain): Likewise.
24032 (vect_shift_permute_load_chain): Likewise.
24033 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24034 (vect_transform_slp_perm_load): Likewise.
24035 (vect_schedule_slp_instance): Likewise.
24036 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24037 (vectorizable_mask_load_store): Likewise.
24038 (vectorizable_bswap): Likewise.
24039 (vectorizable_store): Likewise.
24040 (vectorizable_load): Likewise.
24041 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24042 vec_perm_indices instead of auto_vec_perm_indices. Use
24043 tree_to_vec_perm_builder to read the vector from a tree.
24044 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24045 vec_perm_builder instead of a vec_perm_indices.
24046 (have_whole_vector_shift): Use vec_perm_builder and
24047 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24048 truncation to calc_vec_perm_mask_for_shift.
24049 (vect_create_epilog_for_reduction): Likewise.
24050 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24051 from auto_vec_perm_indices to vec_perm_indices.
24052 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24053 instead of changing individual elements.
24054 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24055 the vector in d.perm.
24056 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24057 from auto_vec_perm_indices to vec_perm_indices.
24058 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24059 instead of changing individual elements.
24060 (arm_vectorize_vec_perm_const): Use new_vector to install
24061 the vector in d.perm.
24062 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24063 Update vec_perm_builder constructor.
24064 (rs6000_expand_interleave): Likewise.
24065 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24066 (rs6000_expand_interleave): Likewise.
24068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24070 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24071 to qimode could truncate the indices.
24072 * optabs.c (expand_vec_perm_var): Likewise.
24074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24076 * Makefile.in (OBJS): Add vec-perm-indices.o.
24077 * vec-perm-indices.h: New file.
24078 * vec-perm-indices.c: Likewise.
24079 * target.h (vec_perm_indices): Replace with a forward class
24081 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24082 * optabs.h: Include vec-perm-indices.h.
24083 (expand_vec_perm): Delete.
24084 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24085 (expand_vec_perm_const): Declare.
24086 * target.def (vec_perm_const_ok): Replace with...
24087 (vec_perm_const): ...this new hook.
24088 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24089 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24090 * doc/tm.texi: Regenerate.
24091 * optabs.def (vec_perm_const): Delete.
24092 * doc/md.texi (vec_perm_const): Likewise.
24093 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24094 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24095 expand_vec_perm for constant permutation vectors. Assert that
24096 the mode of variable permutation vectors is the integer equivalent
24097 of the mode that is being permuted.
24098 * optabs-query.h (selector_fits_mode_p): Declare.
24099 * optabs-query.c: Include vec-perm-indices.h.
24100 (selector_fits_mode_p): New function.
24101 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24102 is defined, instead of checking whether the vec_perm_const_optab
24103 exists. Use targetm.vectorize.vec_perm_const instead of
24104 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24105 fit in the vector mode before using a variable permute.
24106 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24107 vec_perm_indices instead of an rtx.
24108 (expand_vec_perm): Replace with...
24109 (expand_vec_perm_const): ...this new function. Take the selector
24110 as a vec_perm_indices rather than an rtx. Also take the mode of
24111 the selector. Update call to shift_amt_for_vec_perm_mask.
24112 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24113 Use vec_perm_indices::new_expanded_vector to expand the original
24114 selector into bytes. Check whether the indices fit in the vector
24115 mode before using a variable permute.
24116 (expand_vec_perm_var): Make global.
24117 (expand_mult_highpart): Use expand_vec_perm_const.
24118 * fold-const.c: Includes vec-perm-indices.h.
24119 * tree-ssa-forwprop.c: Likewise.
24120 * tree-vect-data-refs.c: Likewise.
24121 * tree-vect-generic.c: Likewise.
24122 * tree-vect-loop.c: Likewise.
24123 * tree-vect-slp.c: Likewise.
24124 * tree-vect-stmts.c: Likewise.
24125 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24127 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24128 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24129 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24130 (aarch64_vectorize_vec_perm_const): ...this new function.
24131 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24132 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24133 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24134 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24135 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24136 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24137 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24139 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24140 check for NEON modes.
24141 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24142 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24143 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24144 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24146 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24147 the old VEC_PERM_CONST conditions.
24148 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24149 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24150 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24151 (ia64_vectorize_vec_perm_const_ok): Merge into...
24152 (ia64_vectorize_vec_perm_const): ...this new function.
24153 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24154 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24155 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24156 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24157 * config/mips/mips.c (mips_expand_vec_perm_const)
24158 (mips_vectorize_vec_perm_const_ok): Merge into...
24159 (mips_vectorize_vec_perm_const): ...this new function.
24160 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24161 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24162 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24163 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24164 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24165 (rs6000_expand_vec_perm_const): Delete.
24166 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24168 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24169 (altivec_expand_vec_perm_const_le): Take each operand individually.
24170 Operate on constant selectors rather than rtxes.
24171 (altivec_expand_vec_perm_const): Likewise. Update call to
24172 altivec_expand_vec_perm_const_le.
24173 (rs6000_expand_vec_perm_const): Delete.
24174 (rs6000_vectorize_vec_perm_const_ok): Delete.
24175 (rs6000_vectorize_vec_perm_const): New function.
24176 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24177 an element count and rtx array.
24178 (rs6000_expand_extract_even): Update call accordingly.
24179 (rs6000_expand_interleave): Likewise.
24180 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24181 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24182 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24183 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24184 (rs6000_expand_vec_perm_const): Delete.
24185 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24186 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24187 (altivec_expand_vec_perm_const_le): Take each operand individually.
24188 Operate on constant selectors rather than rtxes.
24189 (altivec_expand_vec_perm_const): Likewise. Update call to
24190 altivec_expand_vec_perm_const_le.
24191 (rs6000_expand_vec_perm_const): Delete.
24192 (rs6000_vectorize_vec_perm_const_ok): Delete.
24193 (rs6000_vectorize_vec_perm_const): New function. Remove stray
24194 reference to the SPE evmerge intructions.
24195 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24196 an element count and rtx array.
24197 (rs6000_expand_extract_even): Update call accordingly.
24198 (rs6000_expand_interleave): Likewise.
24199 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
24200 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
24202 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24206 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
24207 vector mode and that that mode matches the mode of the data
24209 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
24210 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
24211 directly using expand_vec_perm_1 when forcing selectors into
24213 (expand_vec_perm_var): New function, split out from expand_vec_perm.
24215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24217 * optabs-query.h (can_vec_perm_p): Delete.
24218 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
24219 * optabs-query.c (can_vec_perm_p): Split into...
24220 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
24221 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
24222 particular selector is valid.
24223 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24224 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24225 (vect_grouped_load_supported): Likewise.
24226 (vect_shift_permute_load_chain): Likewise.
24227 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24228 (vect_transform_slp_perm_load): Likewise.
24229 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24230 (vectorizable_bswap): Likewise.
24231 (vect_gen_perm_mask_checked): Likewise.
24232 * fold-const.c (fold_ternary_loc): Likewise. Don't take
24233 implementations of variable permutation vectors into account
24234 when deciding which selector to use.
24235 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
24236 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
24237 with a false third argument.
24238 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
24239 to test whether the constant selector is valid and can_vec_perm_var_p
24240 to test whether a variable selector is valid.
24242 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24244 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
24245 * optabs-query.c (can_vec_perm_p): Likewise.
24246 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
24247 instead of vec_perm_indices.
24248 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
24249 (vect_gen_perm_mask_checked): Likewise,
24250 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
24251 (vect_gen_perm_mask_checked): Likewise,
24253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24255 * optabs-query.h (qimode_for_vec_perm): Declare.
24256 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
24257 (qimode_for_vec_perm): ...this new function.
24258 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
24260 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24262 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
24263 does not have a conditional at the top.
24265 2018-01-02 Richard Biener <rguenther@suse.de>
24267 * ipa-inline.c (big_speedup_p): Fix expression.
24269 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24272 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
24275 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24279 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
24280 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
24281 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
24282 cond_taken_branch_cost 3->4.
24284 2018-01-01 Jakub Jelinek <jakub@redhat.com>
24286 PR tree-optimization/83581
24287 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
24288 TODO_cleanup_cfg if any changes have been made.
24290 PR middle-end/83608
24291 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
24292 convert_modes if target mode has the right side, but different mode
24295 PR middle-end/83609
24296 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
24297 last argument when extracting from CONCAT. If either from_real or
24298 from_imag is NULL, use expansion through memory. If result is not
24299 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
24300 the parts directly to inner mode, if even that fails, use expansion
24303 PR middle-end/83623
24304 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
24305 check for bswap in mode rather than HImode and use that in expand_unop
24308 Copyright (C) 2018 Free Software Foundation, Inc.
24310 Copying and distribution of this file, with or without modification,
24311 are permitted in any medium without royalty provided the copyright
24312 notice and this notice are preserved.