1 2018-08-29 Martin Liska <mliska@suse.cz>
3 * tree-switch-conversion.c (switch_conversion::expand):
4 Strenghten assumption about gswitch statements.
6 2018-08-29 Richard Biener <rguenther@suse.de>
8 PR tree-optimization/87117
9 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
10 re-value-number released SSA VDEFs.
12 2018-08-29 Richard Biener <rguenther@suse.de>
14 PR tree-optimization/87126
15 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
17 2018-08-28 Jim Wilson <jimw@sifive.com>
19 * config/riscv/pic.md: Rewrite.
20 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
22 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
23 (SOFTF, default_load, softload, softstore): New.
25 2018-08-28 Jeff Law <law@redhat.com>
27 * fold-const.c (fold_binary_loc): Remove recently added assert.
29 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
31 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
32 to OP parmeter of generated function.
34 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
36 PR tree-optimization/87009
37 * match.pd: Add boolean optimizations.
39 2018-08-28 Martin Sebor <msebor@redhat.com>
42 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
43 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
44 (pass_walloca::gate): Use it.
45 (alloca_call_type): Same.
46 (pass_walloca::execute): Same.
47 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
49 2018-08-28 David Malcolm <dmalcolm@redhat.com>
51 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
52 GCC_VERSION for usage of "__gcc_dump_printf__" format from
55 2018-08-28 Richard Biener <rguenther@suse.de>
57 PR tree-optimization/87124
58 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
59 constants before looking up avail.
61 2018-08-28 Jakub Jelinek <jakub@redhat.com>
64 * calls.c (maybe_warn_nonstring_arg): Punt early if
65 warn_stringop_overflow is zero. Don't call get_range_strlen
66 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
67 Swap comparison operands to have constants on rhs. Only use
68 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
71 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
73 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
74 use of tree_to_shwi. Remove duplicated test for the size being
75 a whole number of bytes.
77 2018-08-28 Richard Biener <rguenther@suse.de>
79 PR tree-optimization/87117
80 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
81 Handle removed stmt without LHS (GIMPLE_NOP).
83 2018-08-28 Richard Biener <rguenther@suse.de>
85 PR tree-optimization/87117
86 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
87 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
89 2018-08-28 Richard Biener <rguenther@suse.de>
91 PR tree-optimization/87117
92 * tree-ssa-pre.c (compute_avail): Do not make expressions
93 with predicated values available.
94 (get_expr_value_id): Assert we do not run into predicated value
97 2018-08-28 Richard Biener <rguenther@suse.de>
99 PR tree-optimization/87117
100 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
101 get virtual operands.
102 (get_expr_operands): Handle STRING_CST like other decls.
104 2018-08-28 Martin Liska <mliska@suse.cz>
106 * tree.h: Update documentation of fndecl_built_in_p
110 2018-08-27 Jeff Law <law@redhat.com>
111 PR tree-optimization/87110
112 * tree-ssa-dse.c (compute_trims): Handle non-constant
115 2018-08-27 Martin Sebor <msebor@redhat.com>
117 PR tree-optimization/86914
118 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
120 2018-08-27 Martin Sebor <msebor@redhat.com>
122 PR tree-optimization/87112
123 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
124 the type of the bound argument.
126 2018-08-27 Jeff Law <law@redhat.com>
128 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
129 type does not have a TYPE_SIZE_UNIT.
131 2018-08-27 Steve Ellcey <sellcey@cavium.com>
133 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
134 with include of backend.h.
136 2018-08-27 Richard Biener <rguenther@suse.de>
138 PR tree-optimization/86927
139 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
140 use const cond reduction code.
142 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
144 PR tree-optimization/85758
145 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
147 2018-08-27 David Malcolm <dmalcolm@redhat.com>
150 * diagnostic-show-locus.c (class layout_range): Update for
151 conversion of show_caret_p to a tri-state.
152 (layout_range::layout_range): Likewise.
153 (make_range): Likewise.
154 (layout::maybe_add_location_range): Likewise.
155 (layout::should_print_annotation_line_p): Don't show annotation
156 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
157 (layout::get_state_at_point): Update for conversion of
158 show_caret_p to a tri-state. Bail out early for
159 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
160 underlining or source colorization.
161 (gcc_rich_location::add_location_if_nearby): Update for conversion
162 of show_caret_p to a tri-state.
163 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
164 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
166 (selftest::test_one_liner_labels): Likewise.
167 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
168 conversion of show_caret_p to a tri-state.
169 * pretty-print.c (text_info::set_location): Likewise.
170 * pretty-print.h (text_info::set_location): Likewise.
171 * substring-locations.c (format_warning_n_va): Likewise.
172 * tree-diagnostic.c (default_tree_printer): Likewise.
173 * tree-pretty-print.c (newline_and_indent): Likewise.
175 2018-08-27 David Malcolm <dmalcolm@redhat.com>
178 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
179 line above for line-insertion fix-it hints.
180 (selftest::test_fixit_insert_containing_newline): Update the
181 expected results, and add a test with line-numbering enabled.
183 2018-08-27 Martin Liska <mliska@suse.cz>
186 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
187 params with DECL_HAS_VALUE_EXPR_P.
189 2018-08-27 Martin Liska <mliska@suse.cz>
191 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
192 selected expansion strategy.
194 2018-08-27 Martin Liska <mliska@suse.cz>
196 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
197 * builtins.c (is_builtin_fn): Likewise.
198 * attribs.c (diag_attr_exclusions): Use new function
199 fndecl_built_in_p and remove check for FUNCTION_DECL if
201 (builtin_mathfn_code): Likewise.
202 (fold_builtin_expect): Likewise.
203 (fold_call_expr): Likewise.
204 (fold_builtin_call_array): Likewise.
205 (fold_call_stmt): Likewise.
206 (set_builtin_user_assembler_name): Likewise.
207 (is_simple_builtin): Likewise.
208 * calls.c (gimple_alloca_call_p): Likewise.
209 (maybe_warn_nonstring_arg): Likewise.
210 * cfgexpand.c (expand_call_stmt): Likewise.
211 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
212 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
213 (cgraph_node::verify_node): Likewise.
214 * cgraphclones.c (build_function_decl_skip_args): Likewise.
215 (cgraph_node::create_clone): Likewise.
216 * config/arm/arm.c (arm_insert_attributes): Likewise.
217 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
218 * dse.c (scan_insn): Likewise.
219 * expr.c (expand_expr_real_1): Likewise.
220 * fold-const.c (operand_equal_p): Likewise.
221 (fold_binary_loc): Likewise.
222 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
223 * gimple-low.c (lower_stmt): Likewise.
224 * gimple-pretty-print.c (dump_gimple_call): Likewise.
225 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
226 * gimple.c (gimple_build_call_from_tree): Likewise.
227 (gimple_call_builtin_p): Likewise.
228 (gimple_call_combined_fn): Likewise.
229 * gimplify.c (gimplify_call_expr): Likewise.
230 (gimple_boolify): Likewise.
231 (gimplify_modify_expr): Likewise.
232 (gimplify_addr_expr): Likewise.
233 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
234 * ipa-cp.c (determine_versionability): Likewise.
235 * ipa-fnsummary.c (compute_fn_summary): Likewise.
236 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
237 * ipa-split.c (visit_bb): Likewise.
238 (split_function): Likewise.
239 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
240 * lto-cgraph.c (input_node): Likewise.
241 * lto-streamer-out.c (write_symbol): Likewise.
242 * omp-low.c (setjmp_or_longjmp_p): Likewise.
243 (lower_omp_1): Likewise.
244 * predict.c (strip_predict_hints): Likewise.
245 * print-tree.c (print_node): Likewise.
246 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
247 * trans-mem.c (is_tm_irrevocable): Likewise.
248 (is_tm_load): Likewise.
249 (is_tm_simple_load): Likewise.
250 (is_tm_store): Likewise.
251 (is_tm_simple_store): Likewise.
252 (is_tm_abort): Likewise.
253 (tm_region_init_1): Likewise.
254 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
255 * tree-cfg.c (verify_gimple_call): Likewise.
256 (move_stmt_r): Likewise.
257 (stmt_can_terminate_bb_p): Likewise.
258 * tree-eh.c (lower_eh_constructs_2): Likewise.
259 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
260 * tree-inline.c (remap_gimple_stmt): Likewise.
262 (estimate_num_insns): Likewise.
263 (fold_marked_statements): Likewise.
264 * tree-sra.c (scan_function): Likewise.
265 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
266 (optimize_stack_restore): Likewise.
267 (pass_fold_builtins::execute): Likewise.
268 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
269 (mark_all_reaching_defs_necessary_1): Likewise.
270 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
271 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
272 (pass_forwprop::execute): Likewise.
273 * tree-ssa-loop-im.c (stmt_cost): Likewise.
274 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
275 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
276 * tree-ssa-strlen.c (get_string_length): Likewise.
277 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
278 (find_func_aliases_for_call): Likewise.
279 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
280 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
281 * tree-tailcall.c (find_tail_calls): Likewise.
282 * tree.c (need_assembler_name_p): Likewise.
283 (free_lang_data_in_decl): Likewise.
284 (get_call_combined_fn): Likewise.
285 * ubsan.c (is_ubsan_builtin_p): Likewise.
286 * varasm.c (incorporeal_function_p): Likewise.
287 * tree.h (DECL_BUILT_IN): Remove and replace with
289 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
290 (fndecl_built_in_p): New.
292 2018-08-27 Martin Liska <mliska@suse.cz>
294 PR tree-optimization/86847
295 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
296 Dump also subtree probability.
297 (switch_decision_tree::do_jump_if_equal): New function.
298 (switch_decision_tree::emit_case_nodes): Handle special
299 situations in balanced tree that can be emitted much simpler.
300 Fix calculation of probabilities that happen in tree expansion.
301 * tree-switch-conversion.h (struct cluster): Add
303 (struct simple_cluster): Likewise.
304 (struct case_tree_node): Add new function has_child.
305 (do_jump_if_equal): New.
307 2018-08-27 Martin Liska <mliska@suse.cz>
309 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
310 Add new argument to bit_test_cluster constructor.
311 (bit_test_cluster::emit): Set bits really number of values
313 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
314 probability argument.
315 * tree-switch-conversion.h (struct bit_test_cluster):
316 Add m_handles_entire_switch.
318 2018-08-27 Martin Liska <mliska@suse.cz>
320 PR tree-optimization/86702
321 * tree-switch-conversion.c (jump_table_cluster::emit):
322 Make probabilities even for values in jump table
323 according to number of cases handled.
324 (switch_decision_tree::compute_cases_per_edge): Pass
325 argument to reset_out_edges_aux function.
326 (switch_decision_tree::analyze_switch_statement): Likewise.
327 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
330 2018-08-27 Martin Liska <mliska@suse.cz>
332 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
333 cfun argument explicitly.
334 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
335 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
336 function gimple_switch_default_bb.
337 (convert_switch_statements):
339 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
340 * stmt.c (label_to_block_fn): Use label_to_block and pass
341 cfun argument explicitly and use gimple_switch_label_bb.
342 (expand_case): Likewise.
343 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
344 cfun argument explicitly. Likewise.
345 (make_edges_bb): Likewise.
346 (make_cond_expr_edges): Likewise.
347 (get_cases_for_edge): Likewise.
348 (make_gimple_switch_edges): Likewise.
349 (label_to_block_fn): Likewise.
350 (label_to_block): Likewise.
351 (make_goto_expr_edges): Likewise.
352 (make_gimple_asm_edges): Likewise.
353 (main_block_label): Likewise.
354 (group_case_labels_stmt): Likewise.
355 (find_taken_edge_computed_goto): Likewise.
356 (find_taken_edge_switch_expr): Likewise.
357 (gimple_verify_flow_info): Likewise.
358 (gimple_redirect_edge_and_branch): Likewise.
359 (gimple_switch_label_bb): New function.
360 (gimple_switch_default_bb): Likewise.
361 (gimple_switch_edge): Likewise.
362 (gimple_switch_default_edge): Likewise.
363 * tree-cfg.h (label_to_block_fn): Remove and replace ...
364 (label_to_block): ... with this.
365 (gimple_switch_label_bb): New.
366 (gimple_switch_default_bb): Likewise.
367 (gimple_switch_edge): Likewise.
368 (gimple_switch_default_edge): Likewise.
369 * tree-cfgcleanup.c (convert_single_case_switch): Use
370 new gimple functions and pass new argument to label_to_block.
371 (cleanup_control_flow_bb):
372 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
373 cfun argument explicitly.
374 (make_eh_edges): Likewise.
375 (redirect_eh_dispatch_edge): Likewise.
376 (lower_resx): Likewise.
377 (lower_eh_dispatch): Likewise.
378 (maybe_remove_unreachable_handlers): Likewise.
379 (unsplit_eh): Likewise.
380 (cleanup_empty_eh): Likewise.
381 (verify_eh_edges): Likewise.
382 (verify_eh_dispatch_edge): Likewise.
383 * tree-ssa-dom.c (record_edge_info): Likewise.
384 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
385 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
386 (thread_through_normal_block): Likewise.
387 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
388 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
389 * tree-switch-conversion.c (switch_conversion::collect): Use new
391 (switch_conversion::check_final_bb): Likewise.
392 (switch_conversion::gather_default_values): Pass new argument
394 (switch_conversion::build_constructors): Likewise.
395 (switch_decision_tree::compute_cases_per_edge): Use new
396 gimple_switch_edge function.
397 (switch_decision_tree::analyze_switch_statement): Pass new argument
399 (switch_decision_tree::try_switch_expansion): Use
400 gimple_switch_default_edge.
401 * tree-vrp.c (find_switch_asserts): Pass new argument
403 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
404 (vr_values::simplify_switch_using_ranges): Likewise.
406 2018-08-27 Richard Biener <rguenther@suse.de>
408 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
409 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
411 * tree-ssa-sccvn.h (struct vn_pval): New structure.
412 (struct vn_nary_op_s): Add unwind_to member. Add
413 predicated_values flag and put result into a union together
414 with a linked list of vn_pval.
415 (struct vn_ssa_aux): Add name member to make maintaining
416 a map of SSA name to vn_ssa_aux possible. Remove no longer
417 needed info, dfsnum, low, visited, on_sccstack, use_processed
418 and range_info_anti_range_p members.
419 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
420 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
422 (vn_valueize): New global.
423 (vn_context_bb): Likewise.
424 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
425 VN_INFO_PTR_INFO): Remove.
426 * tree-ssa-sccvn.c: ... (rewrite)
427 (pass_fre::execute): For -O2+ initialize loops and run
428 RPO VN in optimistic mode (iterating). For -O1 and -Og
429 run RPO VN in non-optimistic mode.
430 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
431 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
432 * doc/invoke.texi (sccvn-max-scc-size): Remove.
433 (rpo-vn-max-loop-depth): Document.
434 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
435 when valuezing the VUSE signals we walked out of the region.
436 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
437 (phi_translate): Set VN context block to use for availability
439 (compute_avail): Likewise.
440 (pre_valueize): New function.
441 (pass_pre::execute): Adjust to the RPO VN API.
443 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
444 (propagate_constants_for_unrolling): Remove.
445 (tree_unroll_loops_completely): Perform value-numbering
446 on the unrolled bodies loop parent.
448 2018-08-27 Richard Biener <rguenther@suse.de>
450 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
451 for partial antic compute.
453 2018-08-27 Jakub Jelinek <jakub@redhat.com>
455 PR rtl-optimization/87065
456 * combine.c (simplify_if_then_else): Formatting fix.
457 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
459 (known_cond): Don't return const_true_rtx for vector modes. Use
460 CONST0_RTX instead of const0_rtx. Formatting fixes.
462 2018-08-27 Martin Liska <mliska@suse.cz>
464 PR gcov-profile/87069
465 * gcov.c (process_file): Record files already processed
466 and warn about a file being processed multiple times.
468 2018-08-27 Martin Liska <mliska@suse.cz>
471 * config/aarch64/aarch64.c (aarch64_override_options_internal):
472 Set default values for x_aarch64_*_string strings.
473 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
474 prefix. For -mabi do not print '=ABI' in help and use
475 <option_value> format for -msve-vector-bits and -moverride
478 2018-08-26 Jeff Law <law@redhat.com>
480 * config/mips/frame-header-opt.c: Include "backend.h" rather than
483 2018-08-26 Marek Polacek <polacek@redhat.com>
485 PR c++/87029, Implement -Wredundant-move.
486 * doc/invoke.texi: Document -Wredundant-move.
488 2018-08-25 Martin Sebor <msebor@redhat.com>
490 PR tree-optimization/87059
491 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
492 to the same type as the other.
493 * fold-const.c (fold_binary_loc): Assert expectation.
495 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
497 * config/darwin.c (machopic_legitimize_pic_address): Clean up
498 extraneous parentheses, dead code section and formatting.
500 2018-08-24 David Malcolm <dmalcolm@redhat.com>
503 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
504 wide enough for jumps in the line-numbering to be visible.
505 (layout::print_gap_in_line_numbering): New member function.
506 (layout::calculate_line_spans): When using line numbering, merge
507 line spans that are only 1 line apart.
508 (diagnostic_show_locus): When printing line numbers, show gaps in
509 line numbering directly, rather than printing headers.
510 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
511 line-numbering with multiple line spans.
512 (selftest::test_fixit_insert_containing_newline_2): Add test of
513 line-numbering, in which the spans are close enough to be merged.
515 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
517 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
518 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
519 * tree-vrp.c (range_is_nonnull): Remove.
520 (range_includes_zero_p): Accept value_range instead of min/max.
521 (extract_range_from_binary_expr_1): Do not early bail on
523 Use range_includes_zero_p instead of range_is_nonnull.
524 (extract_range_from_unary_expr): Use range_includes_zero_p instead
526 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
527 special case VR_ANTI_RANGE.
528 (vrp_finalize): Same.
529 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
531 (range_is_nonnull): Remove.
532 * vr-values.c (vrp_stmt_computes_nonzero): Use
533 range_includes_zero_p instead of range_is_nonnull.
534 (extract_range_basic): Pass value_range to range_includes_zero_p
535 instead of range_is_nonnull.
537 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
539 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
540 * emit-rtl.h (rtl_data): Remove return_bnd.
541 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
542 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
543 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
544 (POINTER_BOUNDS_MODE): Remove definition.
545 (make_pointer_bounds_mode): Remove.
546 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
547 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
548 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
549 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
550 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
551 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
552 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
554 * config/i386/i386-modes.def (BND32, BND64): Remove.
555 * config/i386/i386.c (dbx_register_map): Remove bound registers.
556 (dbx64_register_map): Ditto.
557 (svr4_dbx_register_map): Ditto.
558 (indirect_thunk_bnd_needed): Remove.
559 (indirect_thunks_bnd_used): Ditto.
560 (indirect_return_bnd_needed): Ditto.
561 (indirect_return_via_cx_bnd): Ditto.
562 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
563 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
564 (output_indirect_thunk): Ditto. Remove need_prefix argument.
565 (output_indirect_thunk_function): Remove handling of
566 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
567 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
568 (ix86_save_reg): Remove handling of crtl->return_bnd.
569 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
570 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
571 and UNSPEC_BNDLX_ADDR.
572 (ix86_output_indirect_branch_via_reg): Remove handling of
573 indirect_thunk_prefix_bnd.
574 (ix86_output_indirect_branch_via_push): Ditto.
575 (ix86_output_function_return): Ditto.
576 (ix86_output_indirect_function_return): Ditto.
577 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
578 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
579 (CALL_USED_REGISTERS): Ditto.
580 (REG_ALLOC_ORDER): Update for removal of bound registers.
581 (HI_REGISTER_NAMES): Ditto.
582 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
583 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
584 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
585 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
586 (FIRST_PSEUDO_REG): Update.
587 (BND): Remove mode iterator.
588 * config/i386/predicates.md (bnd_mem_operator): Remove.
590 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
592 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
595 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
597 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
598 the case in which the permute needs only a single element and
599 repeats for every vector of the result. Extend that case to
600 handle variable-length vectors.
601 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
603 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
606 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
607 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
609 2018-08-24 Richard Biener <rguenther@suse.de>
611 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
612 bb_flags_allocated members.
613 (auto_flag): New RAII class for allocating flags.
614 (auto_edge_flag): New RAII class for allocating edge flags.
615 (auto_bb_flag): New RAII class for allocating bb flags.
616 * cfgloop.c (verify_loop_structure): Allocate temporary edge
618 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
619 in favor of temporarily allocated BB flag.
620 * hsa-brig.c: Re-order includes.
621 * hsa-dump.c: Likewise.
622 * hsa-regalloc.c: Likewise.
623 * print-rtl.c: Likewise.
624 * profile-count.c: Likewise.
626 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
629 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
632 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
635 * wide-int-range.cc (wide_int_range_div): Do not ignore result
636 from wide_int_range_multiplicative_op.
638 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
640 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
643 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
645 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
648 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
650 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
651 full GENERIC expression used for replacement.
653 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
655 * tree-vrp.c (abs_extent_range): Remove.
656 (extract_range_into_wide_ints): Pass wide ints by reference.
657 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
658 Pass wide ints by reference in all calls to
659 extract_range_into_wide_ints.
660 * wide-int-range.cc (wide_int_range_div): New.
661 * wide-int-range.h (wide_int_range_div): New.
662 (wide_int_range_includes_zero_p): New.
663 (wide_int_range_zero_p): New.
665 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
667 * config/aarch64/aarch64.md (arches): New enum.
668 (arch): New enum attr.
669 (arch_enabled): New attr.
670 (enabled): Now uses arch_enabled only.
671 (simd, sve, fp16): Removed attribute.
672 (fp): Attr now defined in terms of 'arch'.
673 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
674 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
675 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
676 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
677 attributes into 'arch'.
678 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
679 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
680 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
681 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
682 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
683 'simd' attribute into 'arch'.
684 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
685 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
686 Convert use of 'fp' attribute to 'arch'.
687 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
688 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
690 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
691 (different modes) Merge 'fp' and 'simd' into 'arch'.
692 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
695 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
697 PR rtl-optimization/87026
698 * expmed.c (canonicalize_comparison): If we can no longer create
699 pseudoregisters, don't.
701 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
704 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
706 * config/arm/arm.c (speculation_barrier_libfunc): New static
708 (arm_init_libfuncs): Initialize it.
709 (arm_emit_speculation_barrier): New function.
710 * config/arm/arm.md (speculation_barrier): Call
711 arm_emit_speculation_barrier for architectures that do not have
713 (speculation_barrier_insn): Only match on Armv7 or later.
715 2018-08-23 Richard Biener <rguenther@suse.de>
718 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
721 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
723 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
724 of single-vector TBLs.
725 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
728 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
731 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
732 aarch64_evpc_tbl guard.
734 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
736 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
739 2018-08-22 Martin Sebor <msebor@redhat.com>
742 * tree-pretty-print.c (pretty_print_string): Add argument.
743 (dump_generic_node): Call to pretty_print_string with string size.
745 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
747 PR rtl-optimization/86771
748 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
749 of two SETs into those two SETs, one to be placed at i2, if that SETs
750 destination is modified between i2 and i3.
752 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
754 PR tree-optimization/86725
755 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
757 (vect_analyze_scalar_cycles_1): Check it.
759 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
761 PR tree-optimization/86725
762 * tree-vect-loop.c (vect_is_simple_reduction): When treating
763 an outer loop phi as a double reduction, make sure that the
764 single user of the phi result is an inner loop phi.
766 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
768 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
769 grouped stores with gaps to a strided group.
771 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
773 * tree-vect-stmts.c (get_group_load_store_type)
774 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
775 first statement in a group.
777 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
779 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
780 the sequence used in gcc/gcc.c.
782 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
785 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
788 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
790 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
791 Darwin10-specific unwinder-shim.
792 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
793 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
794 New to cater for Darwin10 Rosetta.
796 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
798 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
799 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
801 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
806 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
807 (dwarf2out_switch_text_section): Generate a local label for the
808 second function sub-section and apply it as the second FDE start
810 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
811 second sub-section start.
813 2018-08-22 Richard Biener <rguenther@suse.de>
815 PR tree-optimization/86988
816 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
818 2018-08-22 Richard Biener <rguenther@suse.de>
820 PR tree-optimization/86945
821 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
823 2018-08-22 Alexandre Oliva <oliva@adacore.com>
825 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
826 a comment about how uses of r2 for .sdata2 come about.
828 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
830 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
832 2018-08-21 Marek Polacek <polacek@redhat.com>
834 PR c++/86981, Implement -Wpessimizing-move.
835 * doc/invoke.texi: Document -Wpessimizing-move.
837 2018-08-21 Jan Hubicka <jh@suse.cz>
839 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
840 * tree.h (is_redundant_typedef): Remove.
841 * dwarf2out.c (is_redundant_typedef): Turn into static function.
843 2018-08-21 Jan Hubicka <jh@suse.cz>
845 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
848 2018-08-21 Tamar Christina <tamar.christina@arm.com>
850 * expmed.c (extract_low_bits): Reject invalid subregs early.
852 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
855 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
858 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
860 * config/vxworks.h: Guard vxworks_asm_out_constructor and
861 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
862 * config/vxworks.c: Likewise.
864 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
866 * config/vxworks.c: Set targetm.have_ctors_dtors
867 if HAVE_INITFINI_ARRAY_SUPPORT.
868 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
869 if HAVE_INITFINI_ARRAY_SUPPORT.
871 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
873 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
874 default search path for VxWorks < 7.
876 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
878 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
879 (get_format_string): Refer to c_getstr.
881 2018-08-21 Tom de Vries <tdevries@suse.de>
883 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
884 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
885 (debuginfo_early_stop): Declare.
886 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
887 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
888 (debuginfo_early_stop): New function.
889 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
890 and debuginfo_early_stop.
891 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
892 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
893 (general_init): Call debuginfo_early_init.
894 (finalize): Call debuginfo_fini.
895 (do_compile): Call debuginfo_init.
896 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
898 (@item -fdump-debug, @item -fdump-earlydebug): Add.
900 2018-08-21 Tom de Vries <tdevries@suse.de>
902 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
903 flag_dump_noaddr and flag_dump_unnumbered.
905 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
907 * wide-int-range.cc (wide_int_range_abs): New.
908 (wide_int_range_order_set): Rename from wide_int_range_min_max.
909 * wide-int-range.h (wide_int_range_abs): New.
910 (wide_int_range_min_max): New.
911 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
912 case to call wide_int_range_abs.
913 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
914 (extract_range_from_abs_expr): Delete.
916 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
919 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
920 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
923 2018-08-20 Nathan Sidwell <nathan@acm.org>
924 Jeff Law <law@redhat.com>
926 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
927 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
929 2018-08-20 David Malcolm <dmalcolm@redhat.com>
932 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
933 (decl_attributes): Likewise.
934 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
936 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
937 * diagnostic-core.h (class auto_diagnostic_group): New class.
938 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
939 (diagnostic_report_diagnostic): Handle the first diagnostics within
941 (emit_diagnostic): Add auto_diagnostic_group instance.
943 (inform_n): Likewise.
945 (warning_at): Likewise.
946 (warning_n): Likewise.
948 (permerror): Likewise.
951 (error_at): Likewise.
953 (fatal_error): Likewise.
954 (internal_error): Likewise.
955 (internal_error_no_backtrace): Likewise.
956 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
957 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
958 * diagnostic.h (struct diagnostic_context): Add fields
959 "diagnostic_group_nesting_depth",
960 "diagnostic_group_emission_count", "begin_group_cb",
962 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
963 Add auto_diagnostic_group instance(s).
964 (find_explicit_erroneous_behavior): Likewise.
965 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
966 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
967 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
968 (gimplify_va_arg_expr): Likewise.
969 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
970 (HSA_SORRY_AT): Likewise.
971 * ipa-devirt.c (compare_virtual_tables): Likewise.
972 (warn_odr): Likewise.
973 * multiple_target.c (expand_target_clones): Likewise.
974 * opts-common.c (cmdline_handle_error): Likewise.
975 * reginfo.c (globalize_reg): Likewise.
976 * substring-locations.c (format_warning_n_va): Likewise.
977 * tree-inline.c (expand_call_inline): Likewise.
978 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
979 * tree-ssa-loop-niter.c
980 (do_warn_aggressive_loop_optimizations): Likewise.
981 * tree-ssa-uninit.c (warn_uninit): Likewise.
982 * tree.c (warn_deprecated_use): Likewise.
984 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
987 * config/i386/i386.md (eh_return): Always update EH return
988 address in word_mode.
990 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
992 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
993 TARGET_SPLIT_COMPLEX_ARG is defined.
995 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
997 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
999 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1002 * expr.c (expand_assignment): Assert that bitpos is positive.
1003 (store_field): Likewise
1004 (expand_expr_real_1): Make sure that bitpos is positive.
1005 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
1008 2018-08-20 Nathan Sidwell <nathan@acm.org>
1010 * Makefile.in (CPP_ID_DATA_H): Delete.
1011 (CPP_INTERNAL_H): Don't add it.
1012 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
1013 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
1015 2018-08-20 Richard Biener <rguenther@suse.de>
1017 PR tree-optimization/78655
1018 * tree-vrp.c (extract_range_from_binary_expr_1): Make
1019 pointer + offset nonnull if either operand is nonnull work.
1021 2018-08-20 Tom de Vries <tdevries@suse.de>
1023 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
1024 unless the referenced die describes the added property using
1025 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
1026 Otherwise, add a DW_AT_location to the referenced die.
1028 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
1031 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
1032 register_operand when calling ix86_set_reg_reg_cost.
1033 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
1034 Set *total to 0 for operands that satisfy x86_64_immediate_operand
1035 predicate and to 1 otherwise.
1037 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1039 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1040 emit a diagnostic that it is not supported and reset the option.
1041 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1042 supported and consume the option. (ASM_FINAL_SPEC): New.
1044 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1046 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
1049 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
1051 C-SKY port: Documentation
1053 * doc/extend.texi (C-SKY Function Attributes): New section.
1054 * doc/invoke.texi (Option Summary): Add C-SKY options.
1055 (C-SKY Options): New section.
1056 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
1058 2018-08-17 Jojo <jijie_rong@c-sky.com>
1059 Huibin Wang <huibin_wang@c-sky.com>
1060 Sandra Loosemore <sandra@codesourcery.com>
1061 Chung-Lin Tang <cltang@codesourcery.com>
1063 C-SKY port: Backend implementation
1065 * config/csky/*: New.
1066 * common/config/csky/*: New.
1068 2018-08-17 Jojo <jijie_rong@c-sky.com>
1069 Huibin Wang <huibin_wang@c-sky.com>
1070 Sandra Loosemore <sandra@codesourcery.com>
1071 Chung-Lin Tang <cltang@codesourcery.com>
1072 Andrew Jenner <andrew@codesourcery.com>
1074 C-SKY port: Configury
1076 * config.gcc (csky-*-*): New.
1077 * configure.ac: Add csky to targets for dwarf2 debug_line support.
1078 * configure: Regenerated.
1080 2018-08-17 David Malcolm <dmalcolm@redhat.com>
1082 * dump-context.h: Include "dumpfile.h".
1083 (dump_context::dump_printf_va): Convert final param from va_list
1084 to va_list *. Convert from ATTRIBUTE_PRINTF to
1085 ATTRIBUTE_GCC_DUMP_PRINTF.
1086 (dump_context::dump_printf_loc_va): Likewise.
1087 * dumpfile.c: Include "stringpool.h".
1088 (make_item_for_dump_printf_va): Delete.
1089 (make_item_for_dump_printf): Delete.
1090 (class dump_pretty_printer): New class.
1091 (dump_pretty_printer::dump_pretty_printer): New ctor.
1092 (dump_pretty_printer::emit_items): New member function.
1093 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
1095 (dump_pretty_printer::emit_item): New member function.
1096 (dump_pretty_printer::stash_item): New member function.
1097 (dump_pretty_printer::format_decoder_cb): New member function.
1098 (dump_pretty_printer::decode_format): New member function.
1099 (dump_context::dump_printf_va): Reimplement in terms of
1100 dump_pretty_printer.
1101 (dump_context::dump_printf_loc_va): Convert final param from va_list
1103 (dump_context::begin_scope): Reimplement call to
1104 make_item_for_dump_printf.
1105 (dump_printf): Update for change to dump_printf_va.
1106 (dump_printf_loc): Likewise.
1107 (selftest::test_capture_of_dump_calls): Convert "stmt" from
1108 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
1109 with %T, %E, and %G.
1110 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
1111 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
1112 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
1113 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
1114 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
1115 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
1116 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
1117 within a dump_printf_loc call to "%wu".
1118 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
1119 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
1120 missing space after "=".
1121 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
1122 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
1123 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
1124 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
1125 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
1126 duplicate "vectorized" from message.
1128 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
1130 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
1131 polyNxK_t element's TYPE_STRING_FLAG.
1133 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1135 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
1136 (they were unnamed before). Fix comments.
1138 2018-08-17 Nathan Sidwell <nathan@acm.org>
1140 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
1142 2018-08-17 Richard Biener <rguenther@suse.de>
1144 PR tree-optimization/86841
1145 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
1147 2018-08-17 Martin Liska <mliska@suse.cz>
1149 * common.opt: Remove Warn, Init and Report for options with
1150 Ignore/Deprecated flag. Warning is done automatically for
1152 * config/i386/i386.opt: Likewise.
1153 * config/ia64/ia64.opt: Likewise.
1154 * config/rs6000/rs6000.opt: Likewise.
1155 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1156 Remove usage of flag_check_pointer_bounds.
1157 * lto-wrapper.c (merge_and_complain): Do not handle
1158 OPT_fcheck_pointer_bounds.
1159 (append_compiler_options): Likewise.
1160 * opt-functions.awk: Do not handle Deprecated.
1161 * optc-gen.awk: Check that Var, Report and Init are not
1162 used for an option with Ignore/Deprecated flag.
1163 * opts-common.c (decode_cmdline_option): Do not report
1165 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
1167 * opts.h (struct cl_option): Remove cl_deprecated flag.
1168 (CL_ERR_DEPRECATED): Remove error enum value.
1170 2018-08-17 Richard Biener <rguenther@suse.de>
1173 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
1174 across a va-arg-pack using call adjust its return value accordingly.
1176 2018-08-16 Martin Sebor <msebor@redhat.com>
1178 PR tree-optimization/86853
1179 * gimple-ssa-sprintf.c (struct format_result): Rename member.
1180 (struct fmtresult): Add member and initialize it in ctors.
1181 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
1182 (format_string): Handle %S the same as %ls. Set MAYFAIL.
1183 (format_directive): Set POSUNDER4K when MAYFAIL is set.
1184 (parse_directive): Handle %C same as %c.
1185 (sprintf_dom_walker::compute_format_length): Adjust.
1186 (is_call_safe): Adjust.
1188 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1190 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1191 for determining how to count the elements.
1192 * builtins.h (c_strlen): Adjust prototype.
1193 * expr.c (string_constant): Add new parameter mem_size.
1194 Set *mem_size appropriately.
1195 * expr.h (string_constant): Adjust protoype.
1196 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1197 * gimple-fold.h (get_range_strlen): Adjust prototype.
1198 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1199 (format_string): Call get_string_length with eltsize.
1201 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1203 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1204 to emit the span, rather than setting it as the prefix.
1206 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1208 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1209 "margin_char" parameter, defaulting to space. Use it in place
1210 of pp_space for the initial part of the margin.
1211 (layout::print_leading_fixits): Use '+' when filling the margin
1212 of line-insertion fix-it hints.
1214 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1216 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1219 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1221 * config/rs6000/altivec.md: Don't set length attribute to the default
1223 * config/rs6000/darwin.md: Ditto.
1224 * config/rs6000/dfp.md: Ditto.
1225 * config/rs6000/htm.md: Ditto.
1226 * config/rs6000/rs6000.md: Ditto.
1227 * config/rs6000/sync.md: Ditto.
1228 * config/rs6000/vsx.md: Ditto.
1230 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1232 * config/rs6000/altivec.md: Don't set length attribute to the default
1233 value, for branch instructions.
1234 * config/rs6000/darwin.md: Ditto.
1235 * config/rs6000/rs6000.md: Ditto.
1237 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1239 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1240 (unnamed conditional branch define_insn): Set length to 4 or 8
1241 depending on offset.
1242 (<bd>_<mode>): Similar, for alternative 0.
1243 (<bd>tf_<mode>): Ditto.
1245 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1247 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1249 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1251 * doc/rtl.texi: Replace old RTX class names with new names.
1254 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1256 * expmed.h (canonicalize_comparison): New declaration.
1257 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1258 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1259 * optabs.c (prepare_cmp_insn): Likewise.
1260 * rtl.h (unsigned_condition_p): New function which checks if a
1261 comparison operator is unsigned.
1263 2018-08-16 Nathan Sidwell <nathan@acm.org>
1265 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1266 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1268 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1271 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1272 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1273 (mov<mov>): ..this and enable unconditionally.
1275 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1277 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1279 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1281 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1282 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1283 "Common" with "Target".
1285 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1287 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1288 * doc/invoke.texi (mmitigate-rop): Remove.
1289 * config/i386/i386.c: Do not include "regrename.h".
1290 (ix86_rop_should_change_byte_p, reg_encoded_number)
1291 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1293 (ix86_reorg): Remove call to ix86_mitigate_rop.
1294 * config/i386/i386.md (attr "modrm_class"): Remove.
1295 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1296 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1297 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1299 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1301 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1302 allow folding of mergeh() and mergel() for the float and double types.
1303 (fold_mergehl_helper): Rework to handle building a permute tree
1306 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1308 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1311 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1313 * common.opt (fdiagnostics-show-labels): New option.
1314 * diagnostic-show-locus.c (class layout_range): Add field
1316 (class layout): Add field "m_show_labels_p".
1317 (layout_range::layout_range): Add param "label" and use it to
1319 (make_range): Pass in NULL for new "label" param of layout_range's
1321 (layout::layout): Initialize m_show_labels_p.
1322 (layout::maybe_add_location_range): Pass in loc_range->m_label
1323 when constructing layout_range instances.
1324 (struct line_label): New struct.
1325 (layout::print_any_labels): New member function.
1326 (layout::print_line): Call it if label-printing is enabled.
1327 (selftest::test_one_liner_labels): New test.
1328 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1329 * diagnostic.c (diagnostic_initialize): Initialize
1330 context->show_labels_p.
1331 * diagnostic.h (struct diagnostic_context): Add field
1333 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1334 -fno-diagnostics-show-labels.
1335 * dwarf2out.c (gen_producer_string): Add
1336 OPT_fdiagnostics_show_labels to the ignored options.
1337 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1339 (gcc_rich_location::maybe_add_expr): Likewise.
1340 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1341 label" param, defaulting to NULL.
1342 (gcc_rich_location::add_expr): Add "label" param.
1343 (gcc_rich_location::maybe_add_expr): Likewise.
1344 (class text_range_label): New class.
1345 (class range_label_for_type_mismatch): New class.
1346 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1347 of format_warning_va.
1348 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1349 * lto-wrapper.c (merge_and_complain): Add
1350 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1351 (append_compiler_options): Likewise to the dropped options.
1352 (append_diag_options): Likewise to the passed-on options.
1353 * opts.c (common_handle_option): Handle the new option.
1354 * selftest-diagnostic.c
1355 (test_diagnostic_context::test_diagnostic_context): Enable
1357 * substring-locations.c: Include "gcc-rich-location.h".
1358 (format_warning_n_va): Add "fmt_label" and "param_label" params
1359 and use them as appropriate.
1360 (format_warning_va): Add "fmt_label" and "param_label" params,
1361 passing them on to format_warning_n_va.
1362 (format_warning_at_substring): Likewise.
1363 (format_warning_at_substring_n): Likewise.
1364 * substring-locations.h (format_warning_va): Add "fmt_label" and
1365 "param_label" params.
1366 (format_warning_n_va): Likewise.
1367 (format_warning_at_substring): Likewise.
1368 (format_warning_at_substring_n): Likewise.
1369 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1371 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1374 * builtins.c (expand_builtin_memcmp): Do not expand the call
1375 when overflow is detected.
1377 2018-08-15 Martin Sebor <msebor@redhat.com>
1379 PR tree-optimization/71625
1380 * config/aarch64/aarch64-builtins.c
1381 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
1383 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
1385 * config/s390/s390.c (s390_reorg): Remove loop.
1387 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1390 (darwin_function_switched_text_sections): Delete.
1391 * gcc/config/darwin.h
1392 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
1394 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1397 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1398 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1399 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1401 2018-08-15 Martin Liska <mliska@suse.cz>
1403 PR tree-optimization/86925
1404 * predict.c (expr_expected_value_1): When taking
1405 later predictor, assign also probability.
1406 Use fold_build2_initializer_loc in order to fold
1407 the expression in -frounding-math.
1409 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
1411 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
1413 (expand_vec_perm_1): Try the new method.
1415 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
1418 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
1419 Check whether lra_live_max_point is 0 before dividing.
1421 2018-08-14 Martin Sebor <msebor@redhat.com>
1423 PR tree-optimization/86650
1424 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
1425 (vrp_prop::check_mem_ref): Same.
1427 2018-08-13 Liu Hao <lh_mouse@126.com>
1429 * pretty-print.c (eat_esc_sequence): Swap the foreground and
1430 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
1431 and clear it thereafter, as it only works for DBCS.
1433 2018-08-13 Liu Hao <lh_mouse@126.com>
1435 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
1436 handle returned by _get_osf_handle().
1438 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1440 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
1441 for folding vec_perm.
1443 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1445 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
1446 Add support for gimple-folding of vec_pack() and vec_unpack()
1449 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1451 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
1452 vec_xst variants to the list.
1453 (rs6000_gimple_fold_builtin): Add support for folding unaligned
1454 vector loads and stores.
1456 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
1458 * config.gcc (rs6000-ibm-aix4.x): Delete.
1459 (rs6000-ibm-aix5.1): Delete.
1460 (rs6000-ibm-aix5.2): Delete.
1461 (rs6000-ibm-aix5.3): Delete.
1462 * config/rs6000/aix43.h: Delete.
1463 * config/rs6000/aix51.h: Delete.
1464 * config/rs6000/aix52.h: Delete.
1465 * config/rs6000/t-aix43: Delete.
1467 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
1469 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1471 (s390_decompose_address): Factor out constant pool ref
1474 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config/nds32/nds32-predicates.c
1477 (nds32_can_use_bclr_p): Change return type as bool.
1478 (nds32_can_use_bset_p): Ditto.
1479 (nds32_can_use_btgl_p): Ditto.
1480 (nds32_can_use_bitci_p): Ditto.
1481 * config/nds32/nds32-protos.h
1482 (nds32_can_use_bclr_p): Change declaration.
1483 (nds32_can_use_bset_p): Ditto.
1484 (nds32_can_use_btgl_p): Ditto.
1485 (nds32_can_use_bitci_p): Ditto.
1487 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1489 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
1490 Support -msched-prolog-epilog option.
1491 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
1493 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1495 * common/config/nds32/nds32-common.c
1496 (nds32_option_optimization_table): Enalbe -malways-align.
1498 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1500 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
1502 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
1503 OPT_misr_secure_ case.
1504 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
1505 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
1506 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
1508 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
1509 (nds32_isr_info): New field security_level.
1510 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
1511 * config/nds32/nds32.md (return_internal): Consider critical attribute.
1512 * config/nds32/nds32.opt (misr-secure): New option.
1513 * config/nds32/nds32_init.inc: New file.
1514 * config/nds32/nds32_isr.h: New file.
1516 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
1518 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
1519 Update comment for atomic instructions.
1520 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
1521 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
1523 (atomic_loaddi): Revise fence expansion to only emit fence prior to
1524 load for __ATOMIC_SEQ_CST model.
1525 (atomic_loaddi_1): Remove float register target.
1526 (atomic_storedi): Handle CONST_INT values.
1527 (atomic_storedi_1): Remove float register source. Add special case
1529 (memory_barrier): New expander and insn.
1531 2018-08-11 Jakub Jelinek <jakub@redhat.com>
1533 PR tree-optimization/86835
1534 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
1535 new_stmt after def_gsi, make sure to insert new_square_stmt after
1536 that stmt, not 2 stmts before it.
1538 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
1541 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
1544 2018-08-10 Martin Liska <mliska@suse.cz>
1547 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
1549 * builtins.c (expand_builtin_expect_with_probability):
1551 (expand_builtin_expect_with_probability): New function.
1552 (build_builtin_expect_predicate): Add new argumnet probability
1553 for BUILT_IN_EXPECT_WITH_PROBABILITY.
1554 (fold_builtin_expect):
1557 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
1558 * builtins.h (fold_builtin_expect): Set new argument.
1559 * doc/extend.texi: Document __builtin_expect_with_probability.
1560 * doc/invoke.texi: Likewise.
1561 * gimple-fold.c (gimple_fold_call): Pass new argument.
1562 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
1563 also BUILT_IN_EXPECT_WITH_PROBABILITY.
1564 * predict.c (get_predictor_value): New function.
1565 (expr_expected_value): Add new argument probability. Assume
1566 that predictor and probability are always non-null.
1567 (expr_expected_value_1): Likewise. For __builtin_expect and
1568 __builtin_expect_with_probability set probability. Handle
1569 combination in binary expressions.
1570 (tree_predict_by_opcode): Simplify code by simply calling
1571 get_predictor_value.
1572 (pass_strip_predict_hints::execute): Add handling of
1573 BUILT_IN_EXPECT_WITH_PROBABILITY.
1574 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
1576 * tree.h (DECL_BUILT_IN_P): New function.
1578 2018-08-10 Martin Liska <mliska@suse.cz>
1580 PR tree-optimization/85799
1581 * passes.def: Add argument for pass_strip_predict_hints.
1582 * predict.c (class pass_strip_predict_hints): Add new argument
1584 (strip_predictor_early): New function.
1585 (pass_strip_predict_hints::execute): Call the function to
1587 (strip_predict_hints): New function.
1588 * predict.def: Fix comment.
1590 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
1592 * Makefile.in: Clarify which tm.texi to copy over to assert the
1593 right to grant a GFDL license for all.
1595 2018-08-09 Jeff Law <law@redhat.com>
1597 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
1600 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
1602 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
1605 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1607 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
1608 reductions for variable-length vectors.
1610 2018-08-09 David Malcolm <dmalcolm@redhat.com>
1613 * common.opt (fdiagnostics-show-line-numbers): New option.
1614 * diagnostic-show-locus.c (class layout): Add fields
1615 "m_show_line_numbers_p" and "m_linenum_width";
1616 (num_digits): New function.
1617 (test_num_digits): New function.
1618 (layout::layout): Initialize new fields. Update m_x_offset
1619 logic to handle any left margin.
1620 (layout::print_source_line): Print line number when requested.
1621 (layout::start_annotation_line): New member function.
1622 (layout::print_annotation_line): Call it.
1623 (layout::print_leading_fixits): Likewise.
1624 (layout::print_trailing_fixits): Likewise. Update calls to
1625 move_to_column for new parameter.
1626 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
1627 it to potentially call start_annotation_line.
1628 (layout::show_ruler): Call start_annotation_line.
1629 (selftest::test_line_numbers_multiline_range): New selftest.
1630 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
1631 and selftest::test_line_numbers_multiline_range.
1632 * diagnostic.c (diagnostic_initialize): Initialize
1633 show_line_numbers_p.
1634 * diagnostic.h (struct diagnostic_context): Add field
1635 "show_line_numbers_p".
1636 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1637 -fno-diagnostics-show-line-numbers.
1638 * dwarf2out.c (gen_producer_string): Add
1639 OPT_fdiagnostics_show_line_numbers to the ignored options.
1640 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
1641 one setting" options.
1642 (append_compiler_options): Likewise to the dropped options.
1643 (append_diag_options): Likewise to the passed-on options.
1644 * opts.c (common_handle_option): Handle the new option.
1645 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
1647 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
1649 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
1650 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
1651 third argument of type "const signed char" to descriptions of
1652 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
1653 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
1654 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
1655 __builtin_bcdsub_ov functions.
1657 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1659 PR tree-optimization/86858
1660 * tree-vect-loop.c (vect_is_simple_reduction): Restore
1661 flow_bb_inside_loop_p calls.
1663 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1665 PR tree-optimization/86871
1666 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
1667 instead of gimple_assign_lhs.
1669 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
1672 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
1673 register constraint to operand 0.
1674 (add<mode>3_carryinC): Likewise.
1675 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
1677 2018-08-09 Martin Liska <mliska@suse.cz>
1680 * common.opt: Remove extra line.
1682 2018-08-09 Martin Liska <mliska@suse.cz>
1684 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
1685 at the end of a line, make first letter capital and end up
1686 a sentence with a dot.
1687 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
1688 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
1689 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
1690 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
1691 (PARAM_MAX_ISL_OPERATIONS): Likewise.
1692 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
1693 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
1694 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
1695 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
1696 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
1697 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
1698 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
1699 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
1700 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
1701 (PARAM_TREE_REASSOC_WIDTH): Likewise.
1702 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
1703 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
1704 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
1706 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
1709 * config/s390/s390.c (s390_option_override_internal): Reduce the
1710 stack-clash-protection-probe-interval param if it would be too big
1713 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
1716 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
1717 * config/m68k/m68k.c (handle_move_double): Don't call
1718 m68k_final_prescan_insn.
1719 (m68k_adjust_decorated_operand): Renamed from
1720 m68k_final_prescan_insn, remove first and third operand and
1722 (print_operand): Call it.
1723 (print_operand_address): Call it.
1725 2018-08-08 Nathan Sidwell <nathan@acm.org>
1727 * diagnostic.c (diagnostic_report_current_module): Use
1728 linemap_included_from & linemap_included_from_linemap.
1730 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
1732 * config/aarch64/aarch64-cores.def: Add phecda core.
1733 * config/aarch64/aarch64-tune.md: Regenerate.
1734 * doc/invoke.texi: Add phecda core.
1736 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1739 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
1741 * config/s390/s390.md ("movti"): Add more alternatives for
1742 constant to GPR copies.
1744 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1746 * config/s390/s390.c: Fix whitespace damage throughout the file.
1747 * config/s390/s390.h: Likewise.
1748 * config/s390/tpf.h: Likewise.
1750 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1752 * config/s390/s390.c (s390_loadrelative_operand_p):
1753 Remove TARGET_CPU_ZARCH usages.
1754 (s390_rtx_costs): Likewise.
1755 (s390_legitimate_constant_p): Likewise.
1756 (s390_cannot_force_const_mem): Likewise.
1757 (legitimate_reload_constant_p): Likewise.
1758 (s390_preferred_reload_class): Likewise.
1759 (legitimize_pic_address): Likewise.
1760 (legitimize_tls_address): Likewise.
1761 (s390_split_branches): Removed.
1762 (s390_add_execute): Removed.
1763 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
1764 (s390_mainpool_start): Likewise.
1765 (s390_mainpool_finish): Likewise.
1766 (s390_mainpool_cancel): Removed.
1767 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
1768 (s390_chunkify_cancel): Likewise.
1769 (s390_return_addr_rtx): Likewise.
1770 (s390_register_info): Remove split_branches_pending_p uages.
1771 (s390_optimize_register_info): Likewise.
1772 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
1773 split_branches_pending_p usages.
1774 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
1775 (s390_load_got): Likewise.
1776 (s390_expand_split_stack_prologue): Likewise.
1777 (output_asm_nops): Likewise.
1778 (s390_function_profiler): Likewise.
1779 (s390_emit_call): Likewise.
1780 (s390_conditional_register_usage): Likewise.
1781 (s390_optimize_prologue): Likewise.
1782 (s390_reorg): Remove TARGET_CPU_ZARCH and
1783 split_branches_pending_p usages.
1784 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
1786 (s390_output_indirect_thunk_function): Likewise.
1787 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
1788 (TARGET_CPU_ZARCH_P): Removed.
1789 (struct machine_function): Remove split_branches_pending_p.
1790 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
1792 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1794 * common/config/s390/s390-common.c (processor_flags_table):
1796 * config.gcc: Remove with_arch/with_tune support.
1797 * config/s390/2064.md: Remove cpu attribute comparisons.
1798 * config/s390/driver-native.c (s390_host_detect_local_cpu):
1800 * config/s390/linux.h (ASM_SPEC):
1801 Remove -march support.
1802 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
1803 Use a table to get an arch level.
1804 * config/s390/s390-opts.h (enum processor_type):
1806 * config/s390/s390.c
1807 (processor_table): Remove entries, add arch_level values.
1808 (s390_issue_rate): Remove cases.
1809 (s390_option_override): Adjust
1810 s390_option_override_internal() call.
1811 (s390_option_override_internal): Remove deprecation warning.
1812 (s390_valid_target_attribute_tree): Adjust
1813 s390_option_override_internal() call.
1814 * config/s390/s390.h (struct s390_processor):
1815 Share with s390-c.c, add arch_level field.
1816 * config/s390/s390.md:
1817 Remove occurrences in cpu attribute.
1818 * config/s390/s390.opt: Remove -march/-mtune support.
1819 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
1820 * doc/invoke.texi: Remove deprecation warning.
1822 2018-08-08 Luis Machado <luis.machado@linaro.org>
1824 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
1826 (qdf24xx_tunings): Set vector cost structure to
1827 qdf24xx_vector_cost.
1829 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
1830 <register_sextend>: Set to 3.
1832 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
1835 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
1836 * config/aarch64/aarch64-simd.md
1837 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
1838 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
1839 (aarch64_frecpx<mode>): ...this new pattern.
1840 * config/aarch64/aarch64-simd-builtins.def: Remove comment
1841 about aarch64_frecp<FRECP:frecp_suffix><mode>.
1843 2018-08-07 Martin Liska <mliska@suse.cz>
1846 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
1847 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
1848 * predict.def (PRED_MALLOC_NONNULL): New predictor.
1849 * doc/extend.texi: Document that malloc attribute adds
1852 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
1855 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1856 Define to speculation_safe_value_not_needed.
1858 2018-08-06 Jeff Law <law@redhat.com>
1860 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
1861 the vr_values instance to cprop_into_stmt.
1862 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
1863 (cprop_operand): Also query EVRP to determine if OP is a constant.
1865 2018-08-06 Nathan Sidwell <nathan@acm.org>
1867 * diagnostic.c (diagnostic_report_current_module): Reroll
1868 included-at loop. Translate text.
1870 2018-08-06 David Malcolm <dmalcolm@redhat.com>
1872 * function-tests.c (selftest::test_expansion_to_rtl): Call
1873 free_after_compilation.
1875 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1877 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
1879 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
1881 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
1882 loops with memory block operations from getting unrolled.
1884 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
1887 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1888 Define to speculation_safe_value_not_needed.
1890 2018-08-06 Jeff Law <law@redhat.com>
1892 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
1895 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1898 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
1899 with all enabled __intN types.
1901 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
1903 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1905 * alias.c (record_set): Check for clobber high.
1906 * cfgexpand.c (expand_gimple_stmt): Likewise.
1907 * combine-stack-adj.c (single_set_for_csa): Likewise.
1908 * combine.c (find_single_use_1): Likewise.
1909 (set_nonzero_bits_and_sign_copies): Likewise.
1910 (get_combine_src_dest): Likewise.
1911 (is_parallel_of_n_reg_sets): Likewise.
1912 (try_combine): Likewise.
1913 (record_dead_and_set_regs_1): Likewise.
1914 (reg_dead_at_p_1): Likewise.
1915 (reg_dead_at_p): Likewise.
1916 * dce.c (deletable_insn_p): Likewise.
1917 (mark_nonreg_stores_1): Likewise.
1918 (mark_nonreg_stores_2): Likewise.
1919 * df-scan.c (df_find_hard_reg_defs): Likewise.
1920 (df_uses_record): Likewise.
1921 (df_get_call_refs): Likewise.
1922 * dwarf2out.c (mem_loc_descriptor): Likewise.
1923 * haifa-sched.c (haifa_classify_rtx): Likewise.
1924 * ira-build.c (create_insn_allocnos): Likewise.
1925 * ira-costs.c (scan_one_insn): Likewise.
1926 * ira.c (equiv_init_movable_p): Likewise.
1927 (rtx_moveable_p): Likewise.
1928 (interesting_dest_for_shprep): Likewise.
1929 * jump.c (mark_jump_label_1): Likewise.
1930 * postreload-gcse.c (record_opr_changes): Likewise.
1931 * postreload.c (reload_cse_simplify): Likewise.
1932 (struct reg_use): Add source expr.
1933 (reload_combine): Check for clobber high.
1934 (reload_combine_note_use): Likewise.
1935 (reload_cse_move2add): Likewise.
1936 (move2add_note_store): Likewise.
1937 * print-rtl.c (print_pattern): Likewise.
1938 * recog.c (decode_asm_operands): Likewise.
1939 (store_data_bypass_p): Likewise.
1940 (if_test_bypass_p): Likewise.
1941 * regcprop.c (kill_clobbered_value): Likewise.
1942 (kill_set_value): Likewise.
1943 * reginfo.c (reg_scan_mark_refs): Likewise.
1944 * reload1.c (maybe_fix_stack_asms): Likewise.
1945 (eliminate_regs_1): Likewise.
1946 (elimination_effects): Likewise.
1947 (mark_not_eliminable): Likewise.
1948 (scan_paradoxical_subregs): Likewise.
1949 (forget_old_reloads_1): Likewise.
1950 * reorg.c (find_end_label): Likewise.
1951 (try_merge_delay_insns): Likewise.
1952 (redundant_insn): Likewise.
1953 (own_thread_p): Likewise.
1954 (fill_simple_delay_slots): Likewise.
1955 (fill_slots_from_thread): Likewise.
1956 (dbr_schedule): Likewise.
1957 * resource.c (update_live_status): Likewise.
1958 (mark_referenced_resources): Likewise.
1959 (mark_set_resources): Likewise.
1960 * rtl.c (copy_rtx): Likewise.
1961 * rtlanal.c (reg_referenced_p): Likewise.
1962 (single_set_2): Likewise.
1963 (noop_move_p): Likewise.
1964 (note_stores): Likewise.
1965 * sched-deps.c (sched_analyze_reg): Likewise.
1966 (sched_analyze_insn): Likewise.
1968 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1970 * cse.c (invalidate_reg): New function extracted from...
1971 (invalidate): ...here.
1972 (canonicalize_insn): Check for clobber high.
1973 (invalidate_from_clobbers): invalidate clobber highs.
1974 (invalidate_from_sets_and_clobbers): Likewise.
1975 (count_reg_usage): Check for clobber high.
1976 (insn_live_p): Likewise.
1977 * cselib.c (cselib_expand_value_rtx_1):Likewise.
1978 (cselib_invalidate_regno): Check for clobber in setter.
1979 (cselib_invalidate_rtx): Pass through setter.
1980 (cselib_invalidate_rtx_note_stores):
1981 (cselib_process_insn): Check for clobber high.
1982 * cselib.h (cselib_invalidate_rtx): Add operand.
1984 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1986 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
1987 (mark_not_eliminable): Likewise.
1988 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
1989 * lra-lives.c (process_bb_lives): Check for clobber high.
1990 * lra.c (new_insn_reg): Remember clobber highs.
1991 (collect_non_operand_hard_regs): Check for clobber high.
1992 (lra_set_insn_recog_data): Likewise.
1993 (add_regs_to_insn_regno_info): Likewise.
1994 (lra_update_insn_regno_info): Likewise.
1996 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1998 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
1999 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
2001 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2003 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
2004 (copy_insn_1): Likewise.
2005 (gen_hard_reg_clobber_high): New gen function.
2006 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
2007 * genemit.c (gen_exp): Likewise.
2008 (gen_emit_seq): Pass through info.
2009 (gen_insn): Check for CLOBBER_HIGH.
2010 (gen_expand): Pass through info.
2011 (gen_split): Likewise.
2012 (output_add_clobbers): Likewise.
2013 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
2014 (remove_clobbers): Likewise.
2015 * rtl.h (gen_hard_reg_clobber_high): New declaration.
2017 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2019 * doc/rtl.texi (clobber_high): Add.
2020 (parallel): Add in clobber high
2021 * rtl.c (rtl_check_failed_code3): Add function.
2022 * rtl.def (CLOBBER_HIGH): Add expression.
2023 * rtl.h (RTL_CHECKC3): Add macro.
2024 (rtl_check_failed_code3): Add declaration.
2025 (XC3EXP): Add macro.
2027 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2030 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2031 cfun->machine->max_used_stack_alignment if needed.
2033 2018-08-04 Martin Sebor <msebor@redhat.com>
2035 PR tree-optimization/86571
2036 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
2039 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
2042 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2045 2018-08-03 Jeff Law <law@redhat.com>
2048 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2049 Define to speculation_safe_value_not_needed.
2051 2018-08-03 David Malcolm <dmalcolm@redhat.com>
2053 * doc/gcov.texi (-x): Remove duplicate "to".
2054 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
2055 (-Wif-not-aligned): Remove duplicate "is".
2056 (-flto): Remove duplicate "the".
2057 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
2058 duplicate "v5.00.b".
2059 (MSP430 Options): Remove duplicate "and" from the description
2060 of "-mgprel-sec=regexp".
2061 (x86 Options): Remove duplicate copies of "vmldLog102" and
2062 vmlsLog104 from description of "-mveclibabi=type".
2064 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
2066 * internal-fn.h (first_commutative_argument): Declare.
2067 * internal-fn.c (first_commutative_argument): New function.
2068 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
2069 restrictions for pattern statements. Use first_commutative_argument
2070 to look for commutative operands in calls to internal functions.
2072 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
2074 * Makefile.in (wide-int-range.o): New.
2075 * tree-vrp.c: Move all the wide_int_* functions to...
2076 * wide-int-range.cc: ...here.
2077 * tree-vrp.h: Move all the wide_int_* prototypes to...
2078 * wide-int-range.h: ...here.
2080 2018-08-03 Tom de Vries <tdevries@suse.de>
2082 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
2084 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
2085 * except.c (output_function_exception_table): Do early exit if
2086 targetm_common.except_unwind_info (&global_options) == UI_NONE.
2088 2018-08-03 Martin Liska <mliska@suse.cz>
2090 * predict.c (dump_prediction): Change to 2 digits
2093 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2095 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2097 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2099 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
2101 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
2102 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
2103 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
2104 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
2105 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
2107 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2109 * diagnostic-show-locus.c (diagnostic_show_locus): Use
2110 pp_take_prefix when saving the existing prefix.
2111 * diagnostic.c (diagnostic_append_note): Likewise.
2112 * langhooks.c (lhd_print_error_function): Likewise.
2113 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
2114 param's type. Free the existing prefix.
2115 (pp_take_prefix): New function.
2116 (pretty_printer::pretty_printer): Drop the prefix parameter.
2117 Rename the length parameter to match the comment.
2118 (pretty_printer::~pretty_printer): Free the prefix.
2119 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
2121 (struct pretty_printer): Drop the "const" from "prefix" field's
2122 type and clarify memory management.
2123 (pp_set_prefix): Drop the "const" from the 2nd param.
2124 (pp_take_prefix): New decl.
2126 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2128 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
2129 for word_mode_ok here instead of passing as argument.
2130 (expand_block_compare): Change select_block_compare_mode() call.
2131 (expand_strncmp_gpr_sequence): New function.
2132 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
2134 2018-08-02 Jeff Law <law@redhat.com>
2137 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2138 Define to speculation_safe_value_not_needed.
2141 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2142 Define to speculation_safe_value_not_needed.
2144 2018-08-02 Tom de Vries <tdevries@suse.de>
2147 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
2148 function. Return UI_TARGET unconditionally.
2149 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
2150 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
2152 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2154 * genemit.c (print_overload_test): Fix typo.
2156 2018-08-02 Richard Biener <rguenther@suse.de>
2158 PR tree-optimization/86816
2159 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2160 which checks for value availability before querying it.
2161 (gvn_uses_equal): Use it.
2162 (same_succ_hash): Likewise.
2163 (gimple_equal_p): Likewise.
2165 2018-08-02 Nick Clifton <nickc@redhat.com>
2168 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2169 Define to speculation_safe_value_not_needed.
2172 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2173 Define to speculation_safe_value_not_needed.
2176 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2177 Define to speculation_safe_value_not_needed.
2180 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2181 Define to speculation_safe_value_not_needed.
2184 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2185 Define to speculation_safe_value_not_needed.
2188 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2189 Define to speculation_safe_value_not_needed.
2192 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2193 Define to speculation_safe_value_not_needed.
2196 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2197 Define to speculation_safe_value_not_needed.
2200 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2201 speculation_safe_value_not_needed.
2204 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2205 to speculation_safe_value_not_needed.
2207 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2209 * doc/md.texi: Expand the documentation of instruction names
2210 to mention port-local uses. Document '@' in pattern names.
2211 * read-md.h (overloaded_instance, overloaded_name): New structs.
2213 (md_reader::handle_overloaded_name): New member function.
2214 (md_reader::get_overloads): Likewise.
2215 (md_reader::m_first_overload): New member variable.
2216 (md_reader::m_next_overload_ptr): Likewise.
2217 (md_reader::m_overloads_htab): Likewise.
2218 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2219 m_next_overload_ptr and m_overloads_htab.
2220 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2221 (get_mode_token, get_code_token, get_int_token): New functions.
2222 (map_attr_string): Add an optional argument that passes back
2223 the associated iterator.
2224 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2225 (md_reader::handle_overloaded_name, add_overload_instance): New
2227 (apply_iterators): Handle '@' names. Report an error if '@'
2228 is used without iterators.
2229 (initialize_iterators): Initialize the new iterator_group fields.
2230 * genopinit.c (handle_overloaded_code_for)
2231 (handle_overloaded_gen): New functions.
2232 (main): Use them to print declarations of maybe_code_for_* and
2233 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2234 * genemit.c (print_overload_arguments, print_overload_test)
2235 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2236 (main): Use it to print definitions of maybe_code_for_* and
2237 maybe_gen_* functions.
2238 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2239 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2240 instead of explicit mode checks.
2241 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2242 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2243 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2244 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2245 (aarch64_expand_compare_and_swap): Likewise
2246 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2247 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2248 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2249 (aarch64_constant_pool_reload_icode): Delete.
2250 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2251 instead of aarch64_constant_pool_reload_icode. Use
2252 code_for_aarch64_reload_mov instead of explicit mode checks.
2253 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2254 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2255 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2256 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2257 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2258 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2259 (aarch64_atomic_load_op_code): Delete.
2260 (aarch64_emit_atomic_load_op): Likewise.
2261 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2262 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2263 instead of aarch64_emit_atomic_load_op.
2264 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2265 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2266 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2267 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2268 character before the pattern name.
2269 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2270 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2271 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2272 (aarch64_frecps<mode>): Likewise.
2273 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2274 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2275 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2276 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2277 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2279 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2281 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2282 Allow HFmode constants if TARGET_FP_F16INST.
2284 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2287 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2288 No longer check last store for clobber of address register.
2290 2018-08-02 Martin Liska <mliska@suse.cz>
2292 PR gcov-profile/86817
2293 * gcov.c (process_all_functions): New function.
2295 (process_file): Move functions processing to
2296 process_all_functions.
2298 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2300 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2301 "const" to the "gimple *" and "rtx_insn *" parameters.
2302 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2304 (dump_location_t::dump_location_t): Likewise.
2306 2018-08-01 Martin Sebor <msebor@redhat.com>
2308 PR tree-optimization/86650
2309 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2310 rather than a "gcall *". Directly pass the data of interest
2311 to percent_K_format, rather than building a temporary CALL_EXPR
2313 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2314 (gimple_fold_builtin_strncat): Adjust.
2315 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2316 gcall* argument with gimple*.
2317 * gimple-ssa-warn-restrict.c (check_call): Same.
2318 (wrestrict_dom_walker::before_dom_children): Same.
2319 (builtin_access::builtin_access): Same.
2320 (check_bounds_or_overlap): Same
2321 (maybe_diag_overlap): Same.
2322 (maybe_diag_offset_bounds): Same.
2323 * tree-diagnostic.c (default_tree_printer): Move usage of
2324 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2326 * tree-pretty-print.c (percent_K_format): Add argument.
2327 * tree-pretty-print.h: Add argument.
2328 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2329 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2330 (maybe_diag_stxncpy_trunc): Same.
2331 (handle_builtin_stxncpy): Same.
2332 (handle_builtin_strcat): Same.
2334 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2336 * match.pd: Optimise pointer range checks.
2338 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2340 PR tree-optimization/86758
2341 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2342 to remove pattern statements.
2344 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2346 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2347 result of dfs_enumerate_from when constructing stmt_vec_infos,
2348 instead of additionally calling get_loop_body.
2350 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2352 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2354 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2355 When creating an iv, assert that the step is not known to be zero.
2356 (vect_setup_realignment): Update call accordingly.
2357 * tree-vect-stmts.c (vectorizable_store): Likewise.
2358 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2360 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2362 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2363 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2364 (vectorizable_reduction): Likewise.
2365 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2366 (vect_detect_hybrid_slp_stmts): Likewise.
2367 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2369 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2371 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2372 (wide_int_set_zero_nonzero_bits): ...this.
2373 (zero_nonzero_bits_from_vr): Rename to...
2374 (vrp_set_zero_nonzero_bits): ...this.
2375 (extract_range_from_multiplicative_op_1): Abstract wide int
2377 (wide_int_range_multiplicative_op): ...here.
2378 (extract_range_from_binary_expr_1): Extract wide int binary
2379 operations into their own functions.
2380 (wide_int_range_lshift): New.
2381 (wide_int_range_can_optimize_bit_op): New.
2382 (wide_int_range_shift_undefined_p): New.
2383 (wide_int_range_bit_xor): New.
2384 (wide_int_range_bit_ior): New.
2385 (wide_int_range_bit_and): New.
2386 (wide_int_range_trunc_mod): New.
2387 (extract_range_into_wide_ints): New.
2388 (vrp_shift_undefined_p): New.
2389 (extract_range_from_multiplicative_op): New.
2390 (vrp_can_optimize_bit_op): New.
2391 * tree-vrp.h (value_range::dump): New.
2392 (wide_int_range_multiplicative_op): New.
2393 (wide_int_range_lshift):New.
2394 (wide_int_range_shift_undefined_p): New.
2395 (wide_int_range_bit_xor): New.
2396 (wide_int_range_bit_ior): New.
2397 (wide_int_range_bit_and): New.
2398 (wide_int_range_trunc_mod): New.
2399 (zero_nonzero_bits_from_bounds): Rename to...
2400 (wide_int_set_zero_nonzero_bits): ...this.
2401 (zero_nonzero_bits_from_vr): Rename to...
2402 (vrp_set_zero_nonzero_bits): ...this.
2403 (range_easy_mask_min_max): Rename to...
2404 (wide_int_range_can_optimize_bit_op): this.
2405 * vr-values.c (simplify_bit_ops_using_ranges): Rename
2406 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
2408 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2410 * tree-vectorizer.h (vect_orig_stmt): New function.
2411 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
2412 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
2413 (vect_create_epilog_for_reduction): Likewise.
2414 (vectorizable_live_operation): Likewise.
2415 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2416 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
2417 * tree-vect-stmts.c (vectorizable_call): Likewise.
2418 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
2420 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2422 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
2424 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
2425 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
2426 (vect_transform_loop): Likewise.
2427 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2429 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2431 * tree-vectorizer.h (vect_schedule_slp): Return void.
2432 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2433 (vect_schedule_slp): Likewise.
2435 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2437 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
2439 (vect_transform_loop): Update calls accordingly. Schedule SLP
2440 instances before the main loop, if any exist.
2442 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2444 PR tree-optimization/86749
2445 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
2446 If the lhs is used in a COND_EXPR, check that it is being used
2447 as the "then" or "else" value.
2449 2018-08-01 Tom de Vries <tdevries@suse.de>
2452 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2453 speculation_safe_value_not_needed.
2455 2018-08-01 Richard Biener <rguenther@suse.de>
2457 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
2460 2018-08-01 Martin Liska <mliska@suse.cz>
2462 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
2463 format how successful transformation is dumped.
2464 (gimple_mod_pow2_value_transform): Likewise.
2465 (gimple_mod_subtract_transform): Likewise.
2466 (gimple_stringops_transform): Likewise.
2468 2018-08-01 Martin Liska <mliska@suse.cz>
2471 * value-prof.c (interesting_stringop_to_profile_p):
2472 Simplify the code and add BUILT_IN_MEMMOVE.
2473 (gimple_stringops_transform): Likewise.
2475 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
2477 * config/aarch64/aarch64-simd.md
2478 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
2479 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
2480 use GPI iterator instead of SI mode.
2482 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2484 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
2485 rs6000_speculation_barrier.
2486 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
2487 new barrier pattern name.
2489 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2491 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
2492 (speculation_barrier): New insn.
2494 2018-08-01 Richard Biener <rguenther@suse.de>
2497 * graphite.h: Include isl/id.h and isl/space.h to allow build
2500 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
2503 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2504 mode for COFF targets.
2505 * defaults.h (TARGET_COFF): Define.
2506 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
2507 TARGET_COFF): Define.
2508 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
2509 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
2510 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
2512 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2514 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
2515 * gimple-streamer-out.c (output_bb): Save it.
2516 * lto-streamer-in.c (input_struct_function_base): Restore
2517 instance discriminator if available. Create map on demand.
2518 * lto-streamer-out.c (output_struct_function_base): Save it if
2520 * final.c (decl_to_instance_map): Document LTO strategy.
2522 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2523 Olivier Hainque <hainque@adacore.com>
2525 * debug.h (decl_to_instance_map_t): New type.
2526 (decl_to_instance_map): Declare.
2527 (maybe_create_decl_to_instance_map): New inline function.
2528 * final.c (bb_discriminator, last_bb_discriminator): New statics,
2529 to track basic block discriminators.
2530 (final_start_function_1): Initialize them.
2531 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
2533 (decl_to_instance_map): New variable.
2534 (map_decl_to_instance, maybe_set_discriminator): New functions.
2535 (notice_source_line): Set discriminator.
2537 2018-07-31 Ian Lance Taylor <iant@golang.org>
2539 * targhooks.c (default_have_speculation_safe_value): Add
2542 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2544 * dump-context.h: Include "pretty-print.h".
2545 (dump_context::refresh_dumps_are_enabled): New decl.
2546 (dump_context::emit_item): New decl.
2547 (class dump_context): Add fields "m_test_pp" and
2549 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2550 (temp_dump_context::get_dumped_text): New decl.
2551 (class temp_dump_context): Add field "m_pp".
2552 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
2553 (dump_context::refresh_dumps_are_enabled): ...and add a test for
2555 (set_dump_file): Update for above change.
2556 (set_alt_dump_file): Likewise.
2557 (dump_loc): New overload, taking a pretty_printer *.
2558 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
2559 to any test pretty-printer.
2560 (make_item_for_dump_gimple_stmt): New function, adapted from
2561 optinfo::add_gimple_stmt.
2562 (dump_context::dump_gimple_stmt): Call it, and use the result,
2563 eliminating the direct usage of dump_file and alt_dump_file in
2564 favor of indirectly using them via emit_item.
2565 (make_item_for_dump_gimple_expr): New function, adapted from
2566 optinfo::add_gimple_expr.
2567 (dump_context::dump_gimple_expr): Call it, and use the result,
2568 eliminating the direct usage of dump_file and alt_dump_file in
2569 favor of indirectly using them via emit_item.
2570 (make_item_for_dump_generic_expr): New function, adapted from
2572 (dump_context::dump_generic_expr): Call it, and use the result,
2573 eliminating the direct usage of dump_file and alt_dump_file in
2574 favor of indirectly using them via emit_item.
2575 (make_item_for_dump_printf_va): New function, adapted from
2576 optinfo::add_printf_va.
2577 (make_item_for_dump_printf): New function.
2578 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
2579 and use the result, eliminating the direct usage of dump_file and
2580 alt_dump_file in favor of indirectly using them via emit_item.
2581 (make_item_for_dump_dec): New function.
2582 (dump_context::dump_dec): Call it, and use the result,
2583 eliminating the direct usage of dump_file and alt_dump_file in
2584 favor of indirectly using them via emit_item.
2585 (make_item_for_dump_symtab_node): New function, adapted from
2586 optinfo::add_symtab_node.
2587 (dump_context::dump_symtab_node): Call it, and use the result,
2588 eliminating the direct usage of dump_file and alt_dump_file in
2589 favor of indirectly using them via emit_item.
2590 (dump_context::begin_scope): Reimplement, avoiding direct usage
2591 of dump_file and alt_dump_file in favor of indirectly using them
2593 (dump_context::emit_item): New member function.
2594 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2595 Set up test pretty-printer on the underlying context. Call
2596 refresh_dumps_are_enabled.
2597 (temp_dump_context::~temp_dump_context): Call
2598 refresh_dumps_are_enabled.
2599 (temp_dump_context::get_dumped_text): New member function.
2600 (selftest::verify_dumped_text): New function.
2601 (ASSERT_DUMPED_TEXT_EQ): New macro.
2602 (selftest::test_capture_of_dump_calls): Run all tests twice, with
2603 and then without optinfo enabled. Add uses of
2604 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
2605 * dumpfile.h: Update comment for the dump_* API.
2606 * optinfo-emit-json.cc
2607 (selftest::test_building_json_from_dump_calls): Update for new
2608 param for temp_dump_context ctor.
2609 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
2610 and "m_owned" field.
2611 (optinfo_item::~optinfo_item): Likewise.
2612 (optinfo::add_item): New member function.
2613 (optinfo::emit): Update comment.
2614 (optinfo::add_string): Delete.
2615 (optinfo::add_printf): Delete.
2616 (optinfo::add_printf_va): Delete.
2617 (optinfo::add_gimple_stmt): Delete.
2618 (optinfo::add_gimple_expr): Delete.
2619 (optinfo::add_tree): Delete.
2620 (optinfo::add_symtab_node): Delete.
2621 (optinfo::add_dec): Delete.
2622 * optinfo.h (class dump_context): New forward decl.
2623 (optinfo::add_item): New decl.
2624 (optinfo::add_string): Delete.
2625 (optinfo::add_printf): Delete.
2626 (optinfo::add_printf_va): Delete.
2627 (optinfo::add_gimple_stmt): Delete.
2628 (optinfo::add_gimple_expr): Delete.
2629 (optinfo::add_tree): Delete.
2630 (optinfo::add_symtab_node): Delete.
2631 (optinfo::add_dec): Delete.
2632 (optinfo::add_poly_int): Delete.
2633 (optinfo_item::optinfo_item): Remove "owned" param.
2634 (class optinfo_item): Remove field "m_owned".
2636 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2639 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
2640 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
2641 requested variable alignment.
2642 (expand_one_ssa_partition): Likewise.
2643 (expand_one_var): Likewise.
2645 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2647 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2648 to speculation_safe_value_not_needed.
2650 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2652 * targhooks.h (speculation_safe_value_not_needed): New prototype.
2653 * targhooks.c (speculation_safe_value_not_needed): New function.
2654 * target.def (have_speculation_safe_value): Update documentation.
2655 * doc/tm.texi: Regenerated.
2657 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2659 * config/aarch64/iterators.md (ALLI_TI): New iterator.
2660 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
2662 (despeculate_copy<ALLI:mode>_insn): New insn.
2663 (despeculate_copyti_insn): New insn.
2664 (despeculate_simple<ALLI:mode>): New insn
2665 (despeculate_simpleti): New insn.
2666 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
2668 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
2669 aarch64_speculation_safe_value.
2670 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
2672 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2674 * config/aarch64/aarch64-speculation.cc: New file.
2675 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
2676 before pass_reorder_blocks.
2677 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
2679 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
2680 X14 and X15 when tracking speculation.
2681 * config/aarch64/aarch64.md (register name constants): Add
2682 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
2683 (unspec): Add UNSPEC_SPECULATION_TRACKER.
2684 (speculation_barrier): New insn attribute.
2685 (cmp<mode>): Allow SP in comparisons.
2686 (speculation_tracker): New insn.
2687 (speculation_barrier): Add speculation_barrier attribute.
2688 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
2689 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
2690 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
2692 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2694 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
2695 aarch64_track_speculation is true.
2696 (tb<optab><mode>1): Likewise.
2697 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
2698 generate CB[N]Z when tracking speculation.
2699 (aarch64_split_compare_and_swap): Likewise.
2700 (aarch64_split_atomic_op): Likewise.
2702 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2704 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
2706 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2708 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
2709 (speculation_barrier): New insn.
2711 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2713 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
2714 * config/arm/arm.md (speculation_barrier): New expand.
2715 (speculation_barrier_insn): New pattern.
2717 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2719 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
2720 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
2721 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
2722 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
2724 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
2725 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
2726 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
2727 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
2728 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
2729 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
2730 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
2731 * builtins.c (expand_speculation_safe_value): New function.
2732 (expand_builtin): Call it.
2733 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
2734 * doc/extend.texi: Document __builtin_speculation_safe_value.
2735 * doc/md.texi: Document "speculation_barrier" pattern.
2736 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
2737 TARGET_HAVE_SPECULATION_SAFE_VALUE.
2738 * doc/tm.texi: Regenerated.
2739 * target.def (have_speculation_safe_value, speculation_safe_value): New
2741 * targhooks.c (default_have_speculation_safe_value): New function.
2742 (default_speculation_safe_value): New function.
2743 * targhooks.h (default_have_speculation_safe_value): Add prototype.
2744 (default_speculation_safe_value): Add prototype.
2746 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2748 * dump-context.h (dump_context::dump_loc): New decl.
2749 * dumpfile.c (dump_context::dump_loc): New member function.
2750 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
2751 and dump_gimple_stmt.
2752 (dump_context::dump_gimple_expr_loc): Likewise, using
2754 (dump_context::dump_generic_expr_loc): Likewise, using
2756 (dump_context::dump_printf_loc_va): Likewise, using
2758 (dump_context::begin_scope): Explicitly using the global function
2759 "dump_loc", rather than the member function.
2761 2018-07-31 Martin Sebor <msebor@redhat.com>
2763 PR tree-optimization/86741
2764 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
2766 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
2768 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
2769 depend on whether prefetch instructions will be emitted or not.
2770 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
2771 will be emitted or not.
2772 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
2773 (TARGET_SETMEM_PFD): New macros.
2775 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2777 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
2778 (NULL_STMT_VEC_INFO): Delete.
2779 (stmt_vec_info::operator*): Likewise.
2780 (stmt_vec_info::operator gimple *): Likewise.
2781 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
2782 of NULL_STMT_VEC_INFO.
2783 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2784 (vect_reassociating_reduction_p): Likewise.
2785 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2786 (vectorizable_store): Likewise.
2787 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
2788 (vec_info::free_stmt_vec_infos): Likewise.
2790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2792 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
2793 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
2795 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2797 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
2798 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2799 (vec_info::free_stmt_vec_info): New private member functions.
2800 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
2801 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
2802 * tree-parloops.c (gather_scalar_reductions): Remove calls to
2803 set_stmt_vec_info_vec and free_stmt_vec_infos.
2804 * tree-vect-loop.c (_loop_vec_info): Remove call to
2805 set_stmt_vec_info_vec.
2806 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
2807 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
2808 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
2809 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2810 (vec_info::free_stmt_vec_info): ...these new functions. Remove
2811 assignments in {vec_info::,}new_stmt_vec_info that are redundant
2812 with the clearing in the xcalloc.
2813 (stmt_vec_info_vec): Delete.
2814 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
2815 (vectorize_loops): Likewise.
2816 (vec_info::~vec_info): Remove argument from call to
2817 free_stmt_vec_infos.
2818 (vec_info::add_stmt): Remove vinfo argument from call to
2821 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2823 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
2824 rather than a gimple stmt.
2825 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
2826 information for pattern statements when passed the original
2827 statement; instead wait to be passed the pattern statement itself.
2828 Don't call set_vinfo_for_stmt here.
2829 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
2830 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
2831 stmt_vec_infos here.
2832 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
2833 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
2834 stmt_vec_infos entry.
2836 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2838 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
2839 * tree-vectorizer.c (vec_info::replace_stmt): New function.
2840 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
2841 * tree-vect-stmts.c (vectorizable_call): Likewise.
2842 (vectorizable_simd_clone_call): Likewise.
2844 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2846 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
2847 * tree-vectorizer.c (vec_info::remove_stmt): New function.
2848 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
2849 * tree-vect-loop.c (vect_transform_loop): Likewise.
2850 * tree-vect-slp.c (vect_schedule_slp): Likewise.
2851 * tree-vect-stmts.c (vect_remove_stores): Likewise.
2853 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2855 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
2856 (vect_dr_stmt): Delete.
2857 * tree-vectorizer.c (vec_info::lookup_dr): New function.
2858 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
2860 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
2861 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
2862 (vect_verify_datarefs_alignment, vect_peeling_supportable)
2863 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2864 (vect_analyze_data_refs): Likewise.
2865 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
2867 (vect_find_same_alignment_drs): Likewise.
2868 (vect_slp_analyze_node_dependences): Update calls accordingly.
2869 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2870 instead of DR_VECT_AUX.
2871 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
2872 of a vector data references. Use vec_info::lookup_dr instead of
2874 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
2875 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2876 instead of DR_VECT_AUX.
2878 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2880 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
2882 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2884 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2885 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2886 (vect_gen_prolog_loop_niters): Likewise.
2888 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2890 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
2891 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
2892 (vect_known_alignment_in_bytes, vect_dr_behavior)
2893 (vect_get_scalar_dr_size): Take references as dr_vec_infos
2894 instead of data_references. Update calls to other routines for
2895 which the same change has been made.
2896 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
2897 dr_vec_infos instead of stmt_vec_infos.
2898 (vect_analyze_data_ref_dependence): Update call accordingly.
2899 (vect_slp_analyze_data_ref_dependence)
2900 (vect_record_base_alignments): Use DR_VECT_AUX.
2901 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
2902 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
2903 (vector_alignment_reachable_p, vect_get_data_access_cost)
2904 (vect_peeling_supportable, vect_analyze_group_access_1)
2905 (vect_analyze_group_access, vect_analyze_data_ref_access)
2906 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
2907 (vect_compile_time_alias, vect_small_gap_p)
2908 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
2909 (vect_supportable_dr_alignment): Take references as dr_vec_infos
2910 instead of data_references. Update calls to other routines for
2911 which the same change has been made.
2912 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
2913 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2914 (vect_slp_analyze_and_verify_node_alignment)
2915 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2916 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2917 (vect_setup_realignment): Use dr_vec_infos. Update calls after
2919 (_vect_peel_info::dr): Replace with...
2920 (_vect_peel_info::dr_info): ...this new field.
2921 (vect_peeling_hash_get_most_frequent)
2922 (vect_peeling_hash_choose_best_peeling): Update accordingly.
2923 (vect_peeling_hash_get_lowest_cost):
2924 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
2925 routines for which the same change has been made.
2926 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
2928 * tree-vect-loop-manip.c (get_misalign_in_elems)
2929 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
2931 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2932 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
2933 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
2934 (get_group_load_store_type, get_negative_load_store_type)
2935 (vect_get_data_ptr_increment, vectorizable_store)
2936 (vectorizable_load): Likewise.
2937 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
2938 Update calls to other routines for which the same change has been made.
2940 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2942 * tree-vectorizer.h (vec_info::move_dr): New member function.
2943 (dataref_aux): Rename to...
2944 (dr_vec_info): ...this and add "dr" and "stmt" fields.
2945 (_stmt_vec_info::dr_aux): Update accordingly.
2946 (_stmt_vec_info::data_ref_info): Delete.
2947 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
2948 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
2949 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
2950 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
2952 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
2953 (STMT_VINFO_DR_INFO): New macro.
2954 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
2955 (set_dr_misalignment): Update after rename of dataref_aux.
2956 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
2957 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
2958 initialization of STMT_VINFO_DATA_REF.
2959 * tree-vectorizer.c (vec_info::move_dr): New function.
2960 * tree-vect-patterns.c (vect_recog_bool_pattern)
2961 (vect_recog_mask_conversion_pattern)
2962 (vect_recog_gather_scatter_pattern): Use it.
2963 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
2964 the "dr" and "stmt" fields of dr_vec_info instead of
2965 STMT_VINFO_DATA_REF.
2967 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2969 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
2970 (is_pattern_stmt_p): Use it.
2971 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
2972 on pattern statements.
2974 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2976 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
2977 original stmt as a stmt_vec_info rather than a gimple stmt.
2978 (vect_pattern_recog_1): Take the statement directly as a
2979 stmt_vec_info, rather than via a gimple_stmt_iterator.
2980 Update call to vect_mark_pattern_stmts.
2981 (vect_pattern_recog): Update calls accordingly.
2983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2985 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
2986 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
2987 a vect_def_type for the first argument.
2988 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
2989 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
2990 operand if it isn't defined by a vectorized statement.
2991 (vect_build_gather_load_calls): Remove the mask_dt argument and
2992 update calls to vect_get_vec_def_for_stmt_copy.
2993 (vectorizable_bswap): Likewise the dt argument.
2994 (vectorizable_call): Update calls to vectorizable_bswap and
2995 vect_get_vec_def_for_stmt_copy.
2996 (vectorizable_simd_clone_call, vectorizable_assignment)
2997 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
2998 (vectorizable_comparison): Update calls to
2999 vect_get_vec_def_for_stmt_copy.
3000 (vectorizable_store): Likewise. Remove now-unnecessary calls to
3002 (vect_get_loop_based_defs): Remove dt argument and update call
3003 to vect_get_vec_def_for_stmt_copy.
3004 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
3005 and vect_get_vec_def_for_stmt_copy.
3006 (vectorizable_load): Update calls to vect_build_gather_load_calls
3007 and vect_get_vec_def_for_stmt_copy.
3008 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3009 (vectorizable_reduction, vectorizable_live_operation): Update calls
3010 to vect_get_vec_def_for_stmt_copy.
3012 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3014 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
3015 and gimple stmt arguments with a stmt_vec_info.
3016 (vect_record_base_alignments): Update calls accordingly.
3017 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
3018 and gimple stmt arguments with a stmt_vec_info.
3019 (vect_build_slp_tree_1): Remove vinfo argument and update call
3020 to vect_record_max_nunits.
3021 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
3022 and vect_record_max_nunits.
3024 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3026 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
3027 file and take a stmt_vec_info instead of a gimple stmt.
3028 (supportable_widening_operation, vect_finish_replace_stmt)
3029 (vect_finish_stmt_generation, vect_get_store_rhs)
3030 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
3031 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
3032 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
3033 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
3034 (vect_create_data_ref_ptr, bump_vector_ptr)
3035 (vect_permute_store_chain, vect_setup_realignment)
3036 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
3037 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
3038 (vectorizable_reduction, vectorizable_induction)
3039 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
3040 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
3041 than gimple stmts as arguments.
3042 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
3043 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
3044 (can_group_stmts_p, vect_check_gather_scatter)
3045 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3046 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
3047 (vect_permute_load_chain, vect_shift_permute_load_chain)
3048 (vect_transform_grouped_load)
3049 (vect_record_grouped_load_vectors): Likewise.
3050 * tree-vect-loop.c (vect_fixup_reduc_chain)
3051 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
3052 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
3053 (vectorizable_reduction, vectorizable_induction)
3054 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
3055 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
3056 (vect_get_load_store_mask): Likewise.
3057 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
3058 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
3059 * tree-vect-stmts.c (vect_mark_relevant)
3060 (is_simple_and_all_uses_invariant)
3061 (exist_non_indexing_operands_for_use_p, process_use)
3062 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
3063 (vect_get_vec_def_for_operand, vect_get_vec_defs)
3064 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
3065 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
3066 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
3067 (get_negative_load_store_type, get_load_store_type)
3068 (vect_check_load_store_mask, vect_check_store_rhs)
3069 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
3070 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3071 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3072 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3073 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
3074 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
3075 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
3076 (supportable_widening_operation): Likewise.
3078 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3080 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
3081 a stmt_vec_info instead of a gcall.
3082 (vect_check_gather_scatter): Update call accordingly.
3083 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
3085 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
3086 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
3087 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
3088 instead of a gimple stmt.
3089 (vect_transform_loop): Update calls accordingly.
3090 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
3091 stmt_vec_infos instead of gimple stmts.
3092 (vect_analyze_slp_instance): Update use accordingly.
3093 * tree-vect-stmts.c (read_vector_array, write_vector_array)
3094 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
3095 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
3096 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
3097 (vect_gen_widened_results_half, vect_get_loop_based_defs)
3098 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
3099 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
3100 down to subroutines.
3102 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3104 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
3105 of the worklist from a vector of gimple stmts to a vector of
3107 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3108 (vect_mark_stmts_to_be_vectorized): Likewise
3110 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3112 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
3113 statement before passing it to vect_analyze_stmt.
3114 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
3115 the chain of phi vector definitions. Track the exit phi via its
3117 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
3118 STMT_VINFO_REDUC_DEF.
3119 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
3120 stmt_vec_infos to handle the statement chains.
3121 (vect_get_slp_defs): Record the first statement in the node
3122 using a stmt_vec_info.
3123 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
3124 statements here and pass their stmt_vec_info down to subroutines.
3125 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
3126 down to vect_finish_stmt_generation.
3127 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
3128 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
3129 stmt_vec_infos to subroutines.
3130 (vect_remove_stores): Use stmt_vec_infos to handle the statement
3133 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3135 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
3136 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
3137 (vect_permute_store_chain, vect_setup_realignment)
3138 (vect_permute_load_chain, vect_shift_permute_load_chain)
3139 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
3140 stmts internally, and when passing values to other vectorizer routines.
3141 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3142 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
3143 (vect_analyze_loop_operations, get_initial_def_for_reduction)
3144 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
3145 (vectorizable_reduction, vectorizable_induction)
3146 (vectorizable_live_operation, vect_transform_loop_stmt)
3147 (vect_transform_loop): Likewise.
3148 * tree-vect-patterns.c (vect_reassociating_reduction_p)
3149 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
3150 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
3151 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3152 (vect_slp_analyze_node_operations_1): Likewise.
3153 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3154 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
3155 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
3156 (vect_finish_stmt_generation_1, get_group_load_store_type)
3157 (get_load_store_type, vect_build_gather_load_calls)
3158 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3159 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3160 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3161 (vectorizable_store, vectorizable_load, vectorizable_condition)
3162 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
3163 (supportable_widening_operation): Likewise.
3164 (vect_get_vector_types_for_stmt): Likewise.
3165 * tree-vectorizer.h (vect_dr_behavior): Likewise.
3167 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3169 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3170 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
3171 (vect_permute_store_chain, vect_permute_load_chain)
3172 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
3173 repeated stmt_vec_info lookups.
3174 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3175 (vect_update_ivs_after_vectorizer): Likewise.
3176 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
3177 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3178 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
3179 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3180 (vect_bb_slp_scalar_cost): Likewise.
3181 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
3183 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3185 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
3186 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
3187 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3188 to get gassigns and gcalls, rather than operating on generc gimple
3190 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3191 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3192 (vectorizable_load, vect_analyze_stmt): Likewise.
3193 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3195 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3197 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3198 return stmt_vec_infos rather than gimple stmts. Do not accept
3200 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3202 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3203 Update use of get_later_stmt.
3204 (vect_get_constant_vectors): Update call accordingly.
3205 (vect_schedule_slp_instance): Likewise
3206 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3207 (vect_slp_analyze_instance_dependence): Likewise.
3208 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3210 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3212 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3213 (stmt_info_for_cost::stmt_info): ...this new field.
3214 (add_stmt_costs): Update accordingly.
3215 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3216 (vect_get_known_peeling_cost): Likewise.
3217 (vect_estimate_min_profitable_iters): Likewise.
3218 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3220 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3222 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3223 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3224 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3226 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3228 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3230 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3231 a gimple stmt to a stmt_vec_info.
3232 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3234 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3236 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3237 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3238 (_loop_vec_info::reduction_chains): Likewise.
3239 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3241 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3243 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3245 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3246 a gimple stmt to a stmt_vec_info.
3247 (_stmt_vec_info::next_element): Likewise.
3248 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3249 (vect_slp_analyze_and_verify_node_alignment)
3250 (vect_analyze_group_access_1, vect_analyze_group_access)
3251 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3252 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3253 (vect_supportable_dr_alignment): Update accordingly.
3254 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3255 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3256 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3257 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3258 * tree-vect-slp.c (vect_build_slp_tree_1)
3259 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3260 (vect_split_slp_store_group, vect_analyze_slp_instance)
3261 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3262 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3263 (get_group_load_store_type, get_load_store_type)
3264 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3265 (vect_transform_stmt, vect_remove_stores): Likewise.
3267 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3269 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3271 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3272 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3273 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3274 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3275 (vector_alignment_reachable_p, vect_get_data_access_cost)
3276 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3277 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3278 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3279 (vect_analyze_group_access_1, vect_analyze_group_access)
3280 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3281 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3282 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3283 result of vect_dr_stmt and use the stmt_vec_info instead of
3284 the associated gimple stmt.
3285 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3286 (vect_gen_prolog_loop_niters): Likewise.
3287 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3289 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3291 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3292 to a vec<stmt_vec_info>.
3293 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3294 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3296 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3297 to a vec<stmt_vec_info>.
3298 (bst_traits::value_type, bst_traits::value_type): Likewise.
3299 (bst_traits::hash): Update accordingly.
3300 (vect_get_and_check_slp_defs): Change the stmts parameter from
3301 a vec<gimple *> to a vec<stmt_vec_info>.
3302 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3303 (vect_build_slp_tree): Likewise.
3304 (vect_build_slp_tree_2): Likewise. Update uses of
3305 SLP_TREE_SCALAR_STMTS.
3306 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3307 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3308 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3309 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3310 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3311 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3312 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3313 (vect_get_constant_vectors, vect_get_slp_defs)
3314 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3315 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3316 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3317 instead of gimple stmts.
3318 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3319 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3320 (vect_slp_analyze_instance_dependence): Update uses of
3321 SLP_TREE_SCALAR_STMTS.
3322 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3323 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3324 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3325 (get_initial_defs_for_reduction): Likewise.
3326 (vect_create_epilog_for_reduction): Likewise.
3327 (vectorize_fold_left_reduction): Likewise.
3328 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3329 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3330 (can_vectorize_live_stmts): Likewise.
3332 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3334 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3335 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3336 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3338 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3340 (gather_scalar_reductions): Update after above interface changes.
3341 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3342 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3344 (vect_force_simple_reduction): Likewise.
3345 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3346 LOOP_VINFO_REDUCTIONS.
3347 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3349 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3351 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3352 a gimple stmt to a stmt_vec_info.
3353 * tree-vect-loop.c (vect_active_double_reduction_p)
3354 (vect_force_simple_reduction, vectorizable_reduction): Update
3357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3359 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3360 vec<gimple *> to a vec<stmt_vec_info>.
3361 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3362 the reduction_phis argument from a vec<gimple *> to a
3364 (vectorizable_reduction): Likewise the phis local variable that
3365 is passed to vect_create_epilog_for_reduction. Update for new type
3366 of SLP_TREE_VEC_STMTS.
3367 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3368 (vectorizable_live_operation): Likewise.
3369 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3370 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3372 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3374 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3375 a gimple stmt to a stmt_vec_info.
3376 (vectorizable_condition, vectorizable_live_operation)
3377 (vectorizable_reduction, vectorizable_induction): Pass back the
3378 vectorized statement as a stmt_vec_info.
3379 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
3380 use of STMT_VINFO_VEC_STMT.
3381 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
3382 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
3383 as stmt_vec_infos rather than gimple stmts.
3384 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
3386 (vectorizable_live_operation): Likewise.
3387 (vectorizable_reduction, vectorizable_induction): Likewise,
3388 updating use of STMT_VINFO_VEC_STMT.
3389 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
3390 of STMT_VINFO_VEC_STMT.
3391 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
3392 (vectorizable_simd_clone_call, vectorizable_conversion)
3393 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3394 (vectorizable_store, vectorizable_load, vectorizable_condition)
3395 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
3396 from a gimple stmt to a stmt_vec_info.
3397 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
3398 pointer to a stmt_vec_info to the vectorizable_* routines.
3400 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3402 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
3403 a gimple stmt to a stmt_vec_info.
3404 (is_pattern_stmt_p): Update accordingly.
3405 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
3406 (vect_record_grouped_load_vectors): Likewise.
3407 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3408 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
3409 (vect_model_reduction_cost): Likewise.
3410 (vect_create_epilog_for_reduction): Likewise.
3411 (vectorizable_reduction, vectorizable_induction): Likewise.
3412 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3413 Return the stmt_vec_info for the pattern statement.
3414 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
3415 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
3416 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3417 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
3418 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3419 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
3420 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
3421 (free_stmt_vec_info, vect_is_simple_use): Likewise.
3423 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3425 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
3426 (vect_finish_stmt_generation): Likewise.
3427 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3428 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
3429 (vect_build_gather_load_calls): Use the return value of the above
3430 functions instead of a separate call to vinfo_for_stmt. Use narrow
3431 scopes for the input gimple stmt and wider scopes for the associated
3432 stmt_vec_info. Use vec_info::lookup_def when setting these
3433 stmt_vec_infos from an SSA_NAME definition.
3434 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3435 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3436 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3437 (vectorizable_store, vectorizable_load, vectorizable_condition)
3438 (vectorizable_comparison): Likewise.
3439 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
3440 (vectorizable_reduction): Likewise.
3442 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3444 * tree-vectorizer.h (vect_is_simple_use): Add an optional
3445 stmt_vec_info * parameter before the optional gimple **.
3446 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3447 (process_use, vect_get_vec_def_for_operand_1): Update callers.
3448 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
3449 * tree-vect-loop.c (vectorizable_reduction): Likewise.
3450 (vectorizable_live_operation): Likewise.
3451 * tree-vect-patterns.c (type_conversion_p): Likewise.
3452 (vect_look_through_possible_promotion): Likewise.
3453 (vect_recog_rotate_pattern): Likewise.
3454 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3456 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3458 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
3459 a typedef to a wrapper class.
3460 (NULL_STMT_VEC_INFO): New macro.
3461 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
3462 (stmt_vec_info::operator*): New function.
3463 (stmt_vec_info::operator gimple *): Likewise.
3464 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
3465 (add_stmt_costs): Likewise.
3466 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
3467 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3468 (vect_get_known_peeling_cost): Likewise.
3469 (vect_estimate_min_profitable_iters): Likewise.
3470 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3471 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
3472 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3473 (vectorizable_store, free_stmt_vec_infos): Likewise.
3474 (new_stmt_vec_info): Change return type of xcalloc to
3477 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3479 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
3480 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
3481 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
3482 a single_imm_use-based sequence.
3483 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
3485 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3487 * tree-vectorizer.h (vec_info::lookup_def): Declare.
3488 * tree-vectorizer.c (vec_info::lookup_def): New function.
3489 * tree-vect-patterns.c (vect_get_internal_def): Use it.
3490 (vect_widened_op_tree): Likewise.
3491 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3492 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
3493 (vectorizable_reduction): Likewise.
3494 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
3496 (vect_is_slp_reduction): Update calls accordingly. Use
3497 vec_info::lookup_def.
3498 (vect_is_simple_reduction): Likewise
3499 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
3501 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3503 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
3504 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
3505 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
3507 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
3508 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
3509 (vect_update_vf_for_slp, vect_analyze_loop_operations)
3510 (vect_is_slp_reduction, vectorizable_induction)
3511 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
3512 * tree-vect-patterns.c (vect_init_pattern_stmt):
3513 (vect_determine_min_output_precision_1, vect_determine_precisions)
3514 (vect_pattern_recog): Likewise.
3515 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
3516 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
3517 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
3518 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3519 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
3520 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
3521 info field from a loop to a loop_vec_info.
3523 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3525 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
3526 (vec_info::add_stmt): Declare.
3527 * tree-vectorizer.c (vec_info::add_stmt): New function.
3528 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
3529 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
3530 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3531 (vectorizable_induction): Likewise.
3532 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
3533 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3534 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
3535 (vectorizable_load): Likewise.
3536 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3537 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
3538 (vect_recog_gather_scatter_pattern): Likewise.
3539 (append_pattern_def_seq): Likewise. Remove a check that is
3540 performed by add_stmt itself.
3542 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3544 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
3545 which make_ssa_name was called with new_stmt before new_stmt
3548 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3550 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
3552 (vect_is_slp_reduction): ...here...
3553 (vect_is_simple_reduction): ...and here. Remove repetition of tests
3554 that are already known to be false.
3556 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3558 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
3559 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
3560 STMT_VINFO_NUM_SLP_USES when it's true.
3561 (vect_free_slp_instance): Add a final_p parameter and pass it to
3563 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
3564 (vect_analyze_slp_instance): Likewise.
3565 (vect_slp_analyze_operations): Likewise.
3566 (vect_slp_analyze_bb_1): Likewise.
3567 * tree-vectorizer.c (vec_info): Likewise.
3568 * tree-vect-loop.c (vect_transform_loop): Likewise.
3570 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3572 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3573 function is not called for second and subsequent members of
3576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3578 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
3579 cases for nested loops from here to ...
3580 (vect_create_epilog_for_reduction): ...here. Only call
3581 vect_is_simple_use for inner-loop reductions.
3583 2018-07-31 Martin Liska <mliska@suse.cz>
3585 PR gcov-profile/85338
3586 PR gcov-profile/85350
3587 PR gcov-profile/85372
3588 * profile.c (struct location_triplet): New.
3589 (struct location_triplet_hash): Likewise.
3590 (output_location): Do not output a BB that
3591 is already recorded for a line.
3592 (branch_prob): Use streamed_locations.
3594 2018-07-31 Martin Liska <mliska@suse.cz>
3596 PR gcov-profile/85370
3597 * coverage.c (coverage_begin_function): Do not mark target
3598 clones as artificial functions.
3600 2018-07-31 Martin Liska <mliska@suse.cz>
3602 PR gcov-profile/83813
3603 PR gcov-profile/84758
3604 PR gcov-profile/85217
3605 PR gcov-profile/85332
3606 * profile.c (branch_prob): Do not record GOTO expressions
3607 for GIMPLE statements which locations are already streamed.
3609 2018-07-31 Olivier Hainque <hainque@adacore.com>
3611 * gcc.c (handle_spec_function): Accept a soft_matched_part
3612 argument, as do_spec_1. Pass it down to ...
3613 (eval_spec_function): Accept a soft_matched_part argument,
3614 and pass it down to ...
3615 (do_spec_2): Accept a soft_matched_part argument, and pass
3616 it down to do_spec_1.
3617 (do_spec_1): Pass soft_matched_part to handle_spec_function.
3618 (handle_braces): Update call to handle_spec_function.
3619 (driver::set_up_specs): Update calls to do_spec_2.
3620 (compare_debug_dump_opt_spec_function): Likewise.
3621 (compare_debug_self_opt_spec_function): Likewise.
3623 2018-07-31 Olivier Hainque <hainque@adacore.com>
3625 * common.opt (nolibc): New option.
3626 * doc/invoke.texi (Link Options): Document it.
3627 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
3628 * config/alpha/linux.h: Likewise.
3629 * config/arc/elf.h: Likewise.
3630 * config/arm/uclinux-elf.h: Likewise.
3631 * config/arm/unknown-elf.h: Likewise.
3632 * config/avr/avrlibc.h: Likewise.
3633 * config/bfin/bfin.h: Likewise.
3634 * config/bfin/linux.h: Likewise.
3635 * config/bfin/uclinux.h: Likewise.
3636 * config/darwin.h: Likewise.
3637 * config/darwin10.h: Likewise.
3638 * config/darwin12.h: Likewise.
3639 * config/gnu-user.h: Likewise.
3640 * config/lm32/uclinux-elf.h: Likewise.
3641 * config/pa/pa-hpux11.h: Likewise.
3642 * config/pa/pa64-hpux.h: Likewise.
3643 * config/sparc/sparc.h: Likewise.
3645 2018-07-31 Olivier Hainque <hainque@adacore.com>
3647 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
3648 undefined variables.
3650 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3653 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
3656 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3658 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3659 terminated string literal.
3661 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3663 PR rtl-optimization/85160
3664 * combine.c (is_just_move): New function.
3665 (try_combine): Allow combining two instructions into two if neither of
3666 the original instructions was a move.
3668 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
3671 * doc/extend.texi (Global Register Variables): Discourage use of type
3673 (Local Register Variables): Likewise.
3675 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
3677 PR tree-optimization/86506
3678 * hwint.h (ceil_log2): Resync with hwint.c implementation.
3680 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
3683 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
3684 hard_regno, make sure no insn between `from` and `to` clobbers it.
3686 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
3687 Tom de Vries <tdevries@suse.de>
3689 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
3690 (PTX_DEFAULT_RUNTIME_DIM): ... this.
3691 (nvptx_goacc_validate_dims): Set default worker and gang dims to
3692 PTX_DEFAULT_RUNTIME_DIM.
3693 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
3695 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
3697 * config/pa/pa.c (pa_output_addr_vec): Align address table.
3698 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
3699 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
3701 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
3703 * config/rs6000/constraints.md (wG constraint): Delete, no longer
3705 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
3706 predicate to reflect toc fusion has been deleted.
3707 (toc_fusion_mem_raw): Delete, no longer used.
3708 (toc_fusion_mem_wrapped): Likewise.
3709 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
3711 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
3712 Delete, no longer used.
3713 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
3714 meant to be used for toc fusion.
3715 (rs6000_debug_print_mode): Delete toc fusion debugging.
3716 (rs6000_debug_reg_global): Likewise.
3717 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
3718 fusion and secondary reload support that were never used.
3719 (rs6000_option_override_internal): Delete TOC fusion, that was only
3720 partially defined, and it did not work unless you also used the
3722 (rs6000_legitimate_address_p): Delete TOC fusion support.
3723 (rs6000_opt_masks): Likewise.
3724 (fusion_wrap_memory_address): Delete function, no longer used.
3725 (fusion_split_address); Delete TOC fusion support.
3726 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
3727 longer used with toc fusion being deleted.
3728 (TARGET_TOC_FUSION_FP): Likewise.
3729 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
3731 (toc fusion spliter): Delete TOC fusion support.
3732 (toc_fusionload_<mode>): Likewise.
3733 (toc_fusionload_di): Likewise.
3734 (fusion_gpr_load_<mode>): Delete generator function, this insn no
3735 longer needs to be named. Rename predicate to delete TOC fusion.
3736 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3737 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3738 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3739 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3740 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
3742 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3744 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
3745 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
3746 __int128 in built-in function prototypes.
3747 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
3748 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
3750 2018-07-27 Martin Sebor <msebor@redhat.com>
3752 PR tree-optimization/86696
3753 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
3754 types, including enums.
3755 (handle_char_store): Be prepared for the above function to fail.
3757 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
3759 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
3760 when optimization level is lower than 2 or optimize for size.
3762 2018-07-26 Martin Sebor <msebor@redhat.com>
3764 PR tree-optimization/86043
3765 PR tree-optimization/86042
3766 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
3767 (get_string_cst_length): Rename...
3768 (get_min_string_length): ...to this. Add argument.
3769 (handle_char_store): Extend to handle multi-character stores by
3771 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
3772 * tree.h (initializer_zerop): Add argument.
3774 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3777 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
3778 declare target to variables if they have always,{to,from,tofrom} map
3781 2018-07-26 Martin Liska <mliska@suse.cz>
3784 * lto-wrapper.c: Add linker_output as prefix
3785 for ltrans_output_file.
3787 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
3789 PR rtl-optimization/85805
3790 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
3791 value for hard registers if that was written in the same mode.
3793 2018-07-26 Martin Liska <mliska@suse.cz>
3795 PR gcov-profile/86536
3796 * gcov.c (format_gcov): Use printf format %.*f directly
3797 and do not handle special values.
3799 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3801 * common/config/arc/arc-common.c (arc_option_optimization_table):
3802 Update default optimizations for size.
3804 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3806 * config/arc/arc.md (movsf_insn): Add short instruction selection.
3807 * config/arc/constraints.md (CfZ): New constraint.
3808 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
3809 (subsf3_fpu): Likewise.
3810 (cmpsf_fpu): Likewise.
3811 (cmpsf_fpu_uneq): Likewise.
3813 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3815 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
3817 (arc_is_uncached_mem_p): uncached applies to both the variable and
3820 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3822 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
3825 2018-07-25 David Malcolm <dmalcolm@redhat.com>
3827 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
3828 field "m_scopes" from vec to auto_vec.
3830 2018-07-25 Martin Liska <mliska@suse.cz>
3832 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
3835 2018-07-25 Richard Biener <rguenther@suse.de>
3838 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
3839 special wrt context_die late.
3840 (gen_subprogram_die): Re-use DIEs in local scope.
3842 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
3844 PR tree-optimization/86644
3845 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
3847 2018-07-25 Martin Liska <mliska@suse.cz>
3850 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
3851 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
3853 2018-07-25 Martin Liska <mliska@suse.cz>
3856 * params.def: Explain ASan abbreviation and provide
3857 a documentation link.
3859 2018-07-24 Martin Sebor <msebor@redhat.com>
3861 PR tree-optimization/86622
3862 PR tree-optimization/86532
3863 * builtins.h (string_length): Declare.
3864 * builtins.c (c_strlen): Correct handling of non-constant offsets.
3865 (check_access): Be prepared for non-constant length ranges.
3866 (string_length): Make extern.
3867 * expr.c (string_constant): Only handle the minor non-constant
3868 array index. Use string_constant to compute the length of
3869 a generic string constant.
3871 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
3873 PR tree-optimization/86618
3874 * tree-vect-stmts.c (vectorizable_call): Don't take the address
3875 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
3877 2018-07-24 David Malcolm <dmalcolm@redhat.com>
3879 PR tree-optimization/86636
3880 * json.cc (json::object::set): Fix comment. Add assertions.
3881 (json::array::append): Move here from json.h. Add comment and an
3883 (json::string::string): Likewise.
3884 * json.h (json::array::append): Move to json.cc.
3885 (json::string::string): Likewise.
3886 * optinfo-emit-json.cc
3887 (optrecord_json_writer::impl_location_to_json): Assert that we
3888 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
3889 wrapper around it. Expand the location once, rather than three
3891 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
3892 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
3894 (optrecord_json_writer::optinfo_to_json): Likewise, in four
3895 places. Fix some overlong lines.
3897 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3899 * config/aarch64/aarch64-simd.md
3900 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
3901 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
3902 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
3903 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
3904 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
3905 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
3906 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
3907 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
3908 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
3910 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3913 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3914 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3915 and abs_d == d, do the power of two handling if profitable.
3917 2018-07-24 Richard Biener <rguenther@suse.de>
3919 * match.pd: Add BIT_FIELD_REF canonicalizations.
3921 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3924 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3926 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3928 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
3929 terminated STRING_CST object.
3931 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3933 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
3935 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3937 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3938 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
3939 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
3940 the elements into a register.
3941 (rs6000_split_v4si_init_di_reg): Delete.
3942 (rs6000_split_v4si_init): Delete.
3943 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
3944 (vsx_init_v4si): Rewrite as a define_expand.
3946 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3948 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
3949 zero_extend argument from memory): New.
3951 2018-07-22 Martin Sebor <msebor@redhat.com>
3954 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
3955 diagnosing calls with unknown arguments unless -Walloca-larger-than
3956 is restricted to less than PTRDIFF_MAX bytes.
3958 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
3960 * doc/gcov.texi (Invoking Gcov): Editorial changes.
3962 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3964 * pretty-print.c (text_info::set_location): Remove redundant
3965 "line_table" parameter from call to rich_location::set_range.
3967 2018-07-20 Martin Sebor <msebor@redhat.com>
3970 * builtins.c (expand_builtin_alloca): Adjust.
3971 * calls.c (alloc_max_size): Simplify.
3972 * cgraphunit.c (cgraph_node::expand): Adjust.
3973 * common.opt (larger_than_size, warn_frame_larger_than): Remove
3975 (frame_larger_than_size): Same.
3976 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
3977 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
3979 * doc/invoke.texi (GCC Command Options): Document option arguments.
3980 Explain byte-size arguments and suffixes.
3981 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
3982 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
3983 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
3984 * doc/options.texi (UInteger): Expand.
3985 (Host_Wide_Int, ByteSize): Document new properties.
3986 * final.c (final_start_function_1): Include sizes in an error message.
3987 * function.c (frame_offset_overflow): Same.
3988 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
3989 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
3990 Diagnose unbounded alloca calls only for limits of less than
3992 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
3993 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
3995 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
3996 only for limits of less than PTRDIFF_MAX.
3997 * langhooks-def.h (lhd_handle_option): Change function argument
3999 * langhooks.c (lhd_handle_option): Same.
4000 * langhooks.h (handle_option): Same.
4001 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
4003 (var_type, var_type_struct): Same.
4004 (var_set): Handle ByteSize flag.
4005 * optc-gen.awk: Add comments to output to ease debugging. Make
4006 use of HOST_WIDE_INT where appropriate.
4007 * opts-gen-save.awk: Use %lx to format unsigned long.
4008 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
4009 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
4010 arguments. Parse bytes-size suffixes.
4011 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
4012 (enum_value_to_arg): Same.
4013 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
4014 (handle_option): Adjust.
4015 (generate_option): Change function argument to HOST_WIDE_INT.
4016 (cmdline_handle_error): Adjust.
4017 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
4018 (set_option): Change function argument to HOST_WIDE_INT.
4019 (option_enabled): Handle cl_host_wide_int.
4020 (get_option_state): Handle CLVC_SIZE.
4021 (control_warning_option): Same.
4022 * opts.c (common_handle_option): Change function argument to
4023 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
4024 OPT_Wvla_larger_than_.
4025 * opts.h (enum cl_var_type): Add an enumerator.
4026 * stor-layout.c (layout_decl): Print a more meaningful warning.
4027 * toplev.c (output_stack_usage): Adjust.
4029 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
4031 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
4032 call to inline_expand_builtin_string_cmp.
4033 (expand_builtin_strcmp): Likewise.
4034 (expand_builtin_strncmp): Likewise.
4035 (inline_string_cmp): Delete the last parameter, change char_type_node
4036 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
4038 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
4039 the inlining expansion on target where the type of the call has same or
4040 narrower precision than unsigned char.
4042 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4044 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
4045 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
4046 * common.opt (fsave-optimization-record): New option.
4047 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
4048 * doc/invoke.texi (-fsave-optimization-record): New option.
4049 * dumpfile.c: Include "optinfo-emit-json.h".
4050 (struct kv_pair): Move to coretypes.h.
4051 (optgroup_options): Make non-static.
4052 (dump_context::end_scope): Call
4053 optimization_records_maybe_pop_dump_scope.
4054 * dumpfile.h (optgroup_options): New decl.
4055 * json.cc: New file.
4057 * optinfo-emit-json.cc: New file.
4058 * optinfo-emit-json.h: New file.
4059 * optinfo.cc: Include "optinfo-emit-json.h".
4060 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
4061 (optinfo_enabled_p): Check optimization_records_enabled_p.
4062 (optinfo_wants_inlining_info_p): Likewise.
4063 * optinfo.h: Update comment.
4064 * profile-count.c (profile_quality_as_string): New function.
4065 * profile-count.h (profile_quality_as_string): New decl.
4066 (profile_count::quality): New accessor.
4067 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
4068 and optinfo_emit_json_cc_tests.
4069 * selftest.h (selftest::json_cc_tests): New decl.
4070 (selftest::optinfo_emit_json_cc_tests): New decl.
4071 * toplev.c: Include "optinfo-emit-json.h".
4072 (compile_file): Call optimization_records_finish.
4073 (do_compile): Call optimization_records_start.
4074 * tree-ssa-live.c: Include optinfo.h.
4075 (remove_unused_scope_block_p): Retain inlining information if
4076 optinfo_wants_inlining_info_p returns true.
4078 2018-07-20 Richard Biener <rguenther@suse.de>
4081 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
4082 to cover -flto-partition=none.
4084 2018-07-20 Martin Liska <mliska@suse.cz>
4086 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
4087 (get_decl_source_range): Remove unused function.
4089 2018-07-20 Richard Biener <rguenther@suse.de>
4091 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
4092 (struct vn_phi_s): Likewise.
4093 (struct vn_reference_s): Likewise.
4094 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
4095 for searching the slot of an entry known to be in the hash itself.
4096 (vn_phi_hasher::equal): Likewise.
4097 (vn_reference_hasher::equal): Likewise.
4098 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
4100 (optimistic_info, current_info): Remove, keeping only valid_info.
4101 (vn_reference_lookup_1): Remove fallback lookup.
4102 (vn_reference_lookup_2): Likewise.
4103 (vn_nary_op_lookup_1): Likewise.
4104 (vn_phi_lookup): Likewise.
4105 (vn_nary_build_or_lookup_1): Make sure to not chain the built
4107 (vn_reference_insert): Adjust, chain the inserted hash element
4108 at last_inserted_ref.
4109 (vn_reference_insert_pieces): Likewise.
4110 (visit_reference_op_call): Likewise.
4111 (vn_nary_op_insert_into): Chain the inserted hash element at
4113 (vn_nary_op_insert_pieces): Adjust.
4114 (vn_nary_op_insert): Likewise.
4115 (vn_nary_op_insert_stmt): Likewise.
4116 (vn_phi_insert): Adjust, chain the inserted hash element at
4118 (process_scc): Remove clearing and copying the optimistic
4119 table. Instead remove elements inserted during an optimistic
4120 iteration from the single table we maintain.
4121 (init_scc_vn): Adjust.
4122 (free_scc_vn): Likewise.
4123 (sccvn_dom_walker::record_cond): Likewise.
4124 (sccvn_dom_walker::after_dom_children): Likewise.
4126 2018-07-19 Martin Sebor <msebor@redhat.com>
4128 PR tree-optimization/84047
4129 PR tree-optimization/83776
4130 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
4131 (check_array_bounds): Call it.
4133 2018-07-19 Martin Sebor <msebor@redhat.com>
4135 * align.h (align_flags): Use member initialization.
4137 2018-07-19 David Malcolm <dmalcolm@redhat.com>
4139 * Makefile.in (OBJS): Add optinfo.o.
4140 * coretypes.h (class symtab_node): New forward decl.
4141 (struct cgraph_node): New forward decl.
4142 (class varpool_node): New forward decl.
4143 * dump-context.h: New file.
4144 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
4146 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
4147 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
4148 (set_alt_dump_file): Likewise.
4149 (dump_context::~dump_context): New dtor.
4150 (dump_gimple_stmt): Move implementation to...
4151 (dump_context::dump_gimple_stmt): ...this new member function.
4152 Add the stmt to any pending optinfo, creating one if need be.
4153 (dump_gimple_stmt_loc): Move implementation to...
4154 (dump_context::dump_gimple_stmt_loc): ...this new member function.
4155 Start a new optinfo and add the stmt to it.
4156 (dump_gimple_expr): Move implementation to...
4157 (dump_context::dump_gimple_expr): ...this new member function.
4158 Add the stmt to any pending optinfo, creating one if need be.
4159 (dump_gimple_expr_loc): Move implementation to...
4160 (dump_context::dump_gimple_expr_loc): ...this new member function.
4161 Start a new optinfo and add the stmt to it.
4162 (dump_generic_expr): Move implementation to...
4163 (dump_context::dump_generic_expr): ...this new member function.
4164 Add the tree to any pending optinfo, creating one if need be.
4165 (dump_generic_expr_loc): Move implementation to...
4166 (dump_context::dump_generic_expr_loc): ...this new member
4167 function. Add the tree to any pending optinfo, creating one if
4169 (dump_printf): Move implementation to...
4170 (dump_context::dump_printf_va): ...this new member function. Add
4171 the text to any pending optinfo, creating one if need be.
4172 (dump_printf_loc): Move implementation to...
4173 (dump_context::dump_printf_loc_va): ...this new member function.
4174 Start a new optinfo and add the stmt to it.
4175 (dump_dec): Move implementation to...
4176 (dump_context::dump_dec): ...this new member function. Add the
4177 value to any pending optinfo, creating one if need be.
4178 (dump_context::dump_symtab_node): New member function.
4179 (dump_context::get_scope_depth): New member function.
4180 (dump_context::begin_scope): New member function.
4181 (dump_context::end_scope): New member function.
4182 (dump_context::ensure_pending_optinfo): New member function.
4183 (dump_context::begin_next_optinfo): New member function.
4184 (dump_context::end_any_optinfo): New member function.
4185 (dump_context::s_current): New global.
4186 (dump_context::s_default): New global.
4187 (dump_scope_depth): Delete global.
4188 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4189 (dump_symtab_node): New function.
4190 (get_dump_scope_depth): Reimplement in terms of dump_context.
4191 (dump_begin_scope): Likewise.
4192 (dump_end_scope): Likewise.
4193 (selftest::temp_dump_context::temp_dump_context): New ctor.
4194 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4195 (selftest::verify_item): New function.
4196 (ASSERT_IS_TEXT): New macro.
4197 (ASSERT_IS_TREE): New macro.
4198 (ASSERT_IS_GIMPLE): New macro.
4199 (selftest::test_capture_of_dump_calls): New test.
4200 (selftest::dumpfile_c_tests): Call it.
4201 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4202 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4203 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4204 descriptive comment.
4205 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4206 (dump_node, dump_bb): Move these unrelated decls.
4207 (class dump_manager): Add leading comment.
4208 * optinfo.cc: New file.
4209 * optinfo.h: New file.
4211 2018-07-19 Michael Collison <michael.collison@arm.com>
4212 Richard Henderson <rth@redhat.com>
4214 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4215 (subti): Handle op1 zero.
4216 (subvti4, usub4ti4): New.
4217 (*sub<GPI>3_compare1_imm): New.
4218 (sub<GPI>3_carryinCV): New.
4219 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4220 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4222 2018-07-19 Michael Collison <michael.collison@arm.com>
4223 Richard Henderson <rth@redhat.com>
4225 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4226 (addti3): Create simpler code if low part is already known to be 0.
4227 (addvti4, uaddvti4): New.
4228 (*add<GPI>3_compareC_cconly_imm): New.
4229 (*add<GPI>3_compareC_cconly): New.
4230 (*add<GPI>3_compareC_imm): New.
4231 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4232 handle constants within this pattern..
4233 (*add<GPI>3_compareV_cconly_imm): New.
4234 (*add<GPI>3_compareV_cconly): New.
4235 (*add<GPI>3_compareV_imm): New.
4236 (add<GPI>3_compareV): New.
4237 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4238 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4239 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4240 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4241 with 'comparison' operator.
4242 (*add<GPI>3_compareV_cconly_imm): Ditto.
4243 (*add<GPI>3_compareV_cconly): Ditto.
4244 (*add<GPI>3_compareV_imm): Ditto.
4245 (add<GPI>3_compareV): Ditto.
4246 (add<mode>3_carryinC): Ditto.
4247 (*add<mode>3_carryinC_zero): Ditto.
4248 (*add<mode>3_carryinC): Ditto.
4249 (add<mode>3_carryinV): Ditto.
4250 (*add<mode>3_carryinV_zero): Ditto.
4251 (*add<mode>3_carryinV): Ditto.
4253 2018-07-19 Michael Collison <michael.collison@arm.com>
4254 Richard Henderson <rth@redhat.com>
4256 * config/aarch64/aarch64-modes.def (CC_V): New.
4257 * config/aarch64/aarch64-protos.h
4258 (aarch64_addti_scratch_regs): Declare
4259 (aarch64_subvti_scratch_regs): Declare.
4260 (aarch64_expand_subvti): Declare.
4261 (aarch64_gen_unlikely_cbranch): Declare
4262 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4263 for signed overflow using CC_Vmode.
4264 (aarch64_get_condition_code_1): Handle CC_Vmode.
4265 (aarch64_gen_unlikely_cbranch): New function.
4266 (aarch64_addti_scratch_regs): New function.
4267 (aarch64_subvti_scratch_regs): New function.
4268 (aarch64_expand_subvti): New function.
4270 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4272 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4274 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4275 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4278 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4281 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4282 address check not strict.
4284 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4286 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4288 (store_pair_lanes<mode>): Likewise.
4289 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4290 enum value 'ADDR_QUERY_LDP_STP_N'.
4291 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4292 (aarch64_print_address_internal): Add declaration.
4293 (aarch64_print_ldpstp_address): Remove.
4294 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4295 (aarch64_print_operand): Change printing of 'y'.
4296 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4297 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4298 'true' rather than '1'.
4299 * config/aarch64/constraints.md (Uml): Likewise.
4300 (Uml): Rename to Umn.
4303 2018-07-19 Richard Biener <rguenther@suse.de>
4305 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4307 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4308 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4309 (vn_reference_hasher): Likewise.
4310 (struct vn_tables_s): Remove obstack and alloc-pool members.
4311 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4312 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4313 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4314 (vn_reference_insert_pieces): Likewise.
4315 (alloc_vn_nary_op_noinit): Adjust.
4316 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4317 (vn_phi_eq): Adjust.
4318 (shared_lookup_phiargs): Remove.
4319 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4320 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4321 (visit_reference_op_call): Likewise.
4322 (copy_nary, copy_phi, copy_reference): Remove.
4323 (process_scc): Rewind the obstack when iterating. Do not
4324 copy the elements to valid_info but just move them from one
4325 hashtable to the other.
4326 (allocate_vn_table): Adjust.
4327 (free_vn_table): Likewise.
4328 (init_scc_vn): Likewise.
4329 (free_scc_vn): Likewise.
4331 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4334 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4335 indirect_return as function type attribute.
4336 (ix86_attribute_table): Change indirect_return to function
4338 * doc/extend.texi: Update indirect_return attribute.
4340 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4342 * wide-int.h (widest2_int): New.
4343 * gimple-fold.c (arith_overflowed_p): Use it.
4344 * tree.h (widest2_int_cst): New.
4345 * tree-vrp.c (wide_int_binop_overflow): Rename from
4346 vrp_int_const_binop.
4347 Rewrite to work on trees.
4348 (extract_range_from_multiplicative_op_1): Abstract code to...
4349 (wide_int_range_min_max): ...here.
4350 (wide_int_range_cross_product): ...and here.
4351 (extract_range_from_binary_expr_1): Abstract overflow code to...
4352 (wide_int_range_mult_wrapping): ...here.
4353 * tree-vrp.h (wide_int_range_cross_product): New.
4354 (wide_int_range_mult_wrapping): New.
4356 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4357 Julia Koval <julia.koval@intel.com>
4359 * config/i386/x86-tune-costs.h (skylake_memcpy,
4360 skylake_memset): Replace rep_prefix with unrolling for size 512.
4362 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4365 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4366 comparision with EQ_EXPR in last stmt.
4368 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4370 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4371 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4372 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4373 previously known as "PowerPC AltiVec Built-in Functions". Move
4374 some material to new subsubsections "PowerPC AltiVec Built-in
4375 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
4377 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
4378 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
4379 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
4380 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
4382 2018-07-18 Richard Biener <rguenther@suse.de>
4384 PR tree-optimization/86557
4385 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
4388 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
4390 * config/s390/s390.c (s390_function_profiler): Generate CFI.
4392 2018-07-17 Jeff Law <law@redhat.com>
4394 * config/arm/arm.c (get_label_padding): Update for recent
4395 changes to label_to_alignment.
4397 PR tree-optimization/86010
4398 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
4400 * config/mips/mips.c (vr4130_align_insns): Update for recent
4401 changes to label_to_alignment.
4403 * config/frv/frv.c (frv_label_align): Update for recent changes
4404 to label_to_alignment.
4406 * config/nios2/nios2.c (nios2_label_align): Update for recent
4407 changes which dropped ALIGN_LABELS_LOG.
4409 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
4411 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
4412 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
4414 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
4416 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4417 deprecated align_labels_log.
4419 2018-07-17 Richard Biener <rguenther@suse.de>
4422 * dwarf2out.c (init_sections_and_labels): Always generate
4423 a debug_line_str_section for early LTO debug.
4424 (dwarf2out_finish): Reset debug_line_str_hash output early.
4425 Bump counter for extra dwarf5 .debug_loc labels to not conflict
4427 (dwarf2out_early_finish): Output debug_line_str.
4429 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4431 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
4432 index register on z196 or later.
4434 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4436 * config/s390/s390.c (s390_default_align): Set default function
4438 (s390_override_options_after_change): Call s390_default align.
4439 (s390_option_override_internal): Call s390_default_align.
4440 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
4442 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4445 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
4446 to _looptemp_ clauses, other than the first two.
4448 2018-07-17 Martin Liska <mliska@suse.cz>
4450 * opts.c: Do not enable OPT_falign_* for -Os.
4452 2018-07-17 Martin Liska <mliska@suse.cz>
4454 * align.h (MAX_CODE_ALIGN): New.
4455 (MAX_CODE_ALIGN_VALUE): New.
4456 * common/config/i386/i386-common.c (ix86_handle_option):
4457 (MAX_CODE_ALIGN): Moved to align.h.
4458 * final.c (MAX_CODE_ALIGN): Likewise.
4459 * opts.c (parse_and_check_align_values):
4460 (MAX_CODE_ALIGN): Likewise.
4461 (MAX_CODE_ALIGN_VALUE): Likewise.
4463 2018-07-17 Martin Liska <mliska@suse.cz>
4465 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
4466 in order to fulfil coding style.
4467 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
4468 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4469 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4470 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
4471 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
4472 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
4473 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
4474 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
4475 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4477 2018-07-17 Martin Liska <mliska@suse.cz>
4479 * align.h: New file.
4480 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
4482 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
4483 align_flags of label_to_alignment.
4484 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
4486 * config/m68k/m68k.c: Do not use removed align_labels_value and
4488 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
4489 (LOOP_ALIGN): Likewise.
4490 (LABEL_ALIGN): Likewise.
4491 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4492 Remove not used macro.
4493 (rs6000_loop_align): Change return type to align_flags.
4494 (rs6000_loop_align_max_skip): Remove.
4495 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
4496 Change return type to align_flags.
4497 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4498 Remove not used macro.
4499 (rs6000_loop_align): Change return type to align_flags.
4500 (rs6000_loop_align_max_skip): Remove.
4501 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
4502 * config/rx/rx-protos.h (rx_align_for_label): Make it
4504 * config/rx/rx.c (rx_align_for_label): Change return type
4506 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
4508 into align_flags class.
4509 (LABEL_ALIGN): Likewise.
4510 (LOOP_ALIGN): Likewise.
4511 * config/s390/s390.c (s390_label_align): Use align_flags
4513 (s390_asm_output_function_label): Likewise.
4514 * config/sh/sh.c (sh_override_options_after_change):
4515 Use align_flags class directly without macros.
4516 (find_barrier): Likewise.
4517 (barrier_align): Likewise.
4518 (sh_loop_align): Likewise.
4519 * config/spu/spu.c (spu_option_override):
4520 Use align_flags_tuple::get_value instead of removed macros.
4521 (spu_sched_init): Likewise.
4522 * config/spu/spu.h (GTY): Likewise.
4523 * config/visium/visium.c (visium_option_override):
4524 Set "8" as default secondary alignment.
4525 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
4526 in order to guarantee secondary alignment of 8.
4527 * coretypes.h: Include align.h header file.
4528 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
4529 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
4530 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
4531 * doc/tm.texi.in: Likewise.
4532 * final.c (struct label_alignment): Remove not used structure.
4533 (LABEL_ALIGN): Change type to align_flags.
4534 (LOOP_ALIGN): Likewise.
4535 (JUMP_ALIGN): Likewise.
4536 (default_loop_align_max_skip): Remove.
4537 (default_label_align_max_skip): Likewise.
4538 (default_jump_align_max_skip): Likewise.
4539 (default_label_align_after_barrier_max_skip):
4540 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
4541 (LABEL_TO_MAX_SKIP): Remove.
4542 (label_to_alignment): Return align_flags type instead of integer.
4543 (label_to_max_skip): Remove.
4544 (align_fuzz): Use align_flags type.
4545 (compute_alignments): Use align_flags type and use align_flags::max
4546 to combine multiple alignments.
4547 (grow_label_align): Grow vec instead of C array.
4548 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
4549 (shorten_branches): Use align_flags type and use align_flags::max
4550 to combine multiple alignments.
4551 (final_scan_insn_1): Remove usage of secondary alignment that comes
4552 from label alignment, but instead use proper secondary alignment
4553 which is computed in grow_label_align.
4554 * flags.h (struct align_flags_tuple): Move to align.h.
4555 (struct align_flags): Likewise.
4556 (state_align_loops): Rename to align_loops.
4557 (state_align_jumps): Rename to align_jumps.
4558 (state_align_labels): Rename to align_labels.
4559 (state_align_functions): Rename to align_functions.
4560 (align_loops_log): Remove.
4561 (align_jumps_log): Remove.
4562 (align_labels_log): Remove.
4563 (align_functions_log): Remove.
4564 (align_loops_max_skip): Remove.
4565 (align_jumps_max_skip): Remove.
4566 (align_labels_max_skip): Remove.
4567 (align_functions_max_skip): Remove.
4568 (align_loops_value): Remove.
4569 (align_jumps_value): Remove.
4570 (align_labels_value): Remove.
4571 (align_functions_value): Remove.
4572 * output.h (label_to_alignment): Change return type to align_flags.
4573 (label_to_max_skip): Remove.
4574 * target.def: Remove loop_align_max_skip, label_align_max_skip,
4575 jump_align_max_skip macros.
4576 * targhooks.h (default_loop_align_max_skip): Remove.
4577 (default_label_align_max_skip): Likewise.
4578 (default_jump_align_max_skip): Likewise.
4579 (default_label_align_after_barrier_max_skip): Remove.
4580 * toplev.c (read_log_maxskip): Use ::normalize function.
4581 (parse_N_M): Remove not used argument and also call ::normalize.
4582 (parse_alignment_opts): Do not pass unused arguments.
4583 * varasm.c (assemble_start_function): Use directly align_functions
4584 instead of removed macros.
4585 * system.h: Do not poison removed macros.
4587 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4590 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
4591 and cond temporaries don't have reference type if iterator has
4592 pointer type. For init use &for_pre_body instead of pre_p if
4593 for_pre_body is non-empty.
4595 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4597 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
4598 double-double modes to SFmode directly directly.
4599 (trunc<mode>sf2_fprs): Delete.
4601 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4603 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
4604 for conversions between IFmode and the decimal floating point modes.
4605 (init_float128_ieee): Use the correct names for conversions between
4606 KFmode and the decimal floating point modes.
4608 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4610 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
4611 for the conversions between TDmode and IFmode.
4612 (init_float128_ieee): Use more correct names for the conversions
4613 between TDmode and KFmode.
4615 2018-07-16 Jakub Jelinek <jakub@redhat.com>
4617 PR tree-optimization/86526
4618 * builtins.c (expand_builtin_memcmp): Formatting fixes.
4619 (inline_expand_builtin_string_cmp): Likewise.
4620 (inline_string_cmp): Likewise. Use c_readstr instead of
4621 builtin_memcpy_read_str. Add unit_mode temporary.
4623 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4626 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
4627 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
4629 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4631 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
4632 Alphabetize prototypes of built-in functions, separating out
4633 built-in functions that are listed in this section but should be
4634 described elsewhere.
4636 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
4639 * expmed.c (emit_store_flag): Do not emit setcc followed by a
4640 conditional move when trapping comparison was split to a
4641 non-trapping one (and vice versa).
4643 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4645 * config/s390/s390.c (s390_function_profiler): Generate nops
4646 instead of profiler call sequences.
4647 * config/s390/s390.opt: Add the new option.
4649 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4651 * config/s390/s390.c (s390_function_profiler): Generate
4652 __mcount_loc section.
4653 * config/s390/s390.opt: Add the new option.
4655 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4657 * common.opt: Add the new warning.
4658 * config/s390/s390.c (s390_function_profiler): Emit "brasl
4659 %r0,__fentry__" when -mfentry is specified.
4660 (s390_option_override_internal): Disallow -mfentry for 31-bit
4662 * config/s390/s390.opt: Add the new option.
4664 2018-07-16 Richard Biener <rguenther@suse.de>
4667 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
4668 for function-local FUNCTION_DECL and RESULT_DECL immediately.
4670 2018-07-16 Martin Liska <mliska@suse.cz>
4673 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
4676 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4678 * config/arc/arcHS.md: Update ARCHS scheduling rules.
4680 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4682 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
4684 * config/arc/arc-cpus.def (hs4x): New cpu.
4686 * config/arc/arc-tables.opt: Regenerate.
4687 * config/arc/arc.c (arc_sched_issue_rate): New function.
4688 (TARGET_SCHED_ISSUE_RATE): Define.
4689 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
4690 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
4692 (attr tune): Add ARCHS4x tune values.
4693 (attr tune_dspmpy): Define.
4694 (*tst): Correct instruction type.
4695 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
4696 * config/arc/arcHS4x.md: New file.
4697 * config/arc/fpu.md: Update instruction type attributes.
4698 * config/arc/t-multilib: Regenerate.
4700 2018-07-16 Tom de Vries <tdevries@suse.de>
4703 * var-tracking.c (vt_initialize): Fix pre_dec handling.
4705 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4707 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
4710 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
4712 PR tree-optimization/86514
4713 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
4714 conversion to a boolean type from a type with greater precision.
4716 2018-07-16 Tom de Vries <tdevries@suse.de>
4718 * var-tracking.c (vt_initialize): Print adjusted insn slim if
4719 dump_flags request TDF_SLIM.
4721 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
4723 * fold-const.c (int_const_binop_1): Abstract...
4724 (wide_int_binop): ...wide int code here.
4725 (poly_int_binop): ...poly int code here.
4726 Abstract the rest of int_const_binop_1 into int_const_binop.
4727 * fold-const.h (wide_int_binop): New.
4728 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
4729 Remove useless PLUS/MINUS_EXPR case.
4730 (zero_nonzero_bits_from_vr): Move wide int code...
4731 (zero_nonzero_bits_from_bounds): ...here.
4732 (extract_range_from_binary_expr_1): Move mask optimization code...
4733 (range_easy_mask_min_max): ...here.
4734 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
4735 (range_easy_mask_min_max): New.
4737 2018-07-15 Jeff Law <law@redhat.com>
4740 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
4743 2018-07-14 Jim Wilson <jimw@sifive.com>
4745 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
4747 2018-07-14 Paul Koning <ni1d@arrl.net>
4749 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
4751 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
4753 * lto-streamer-out.c (copy_function_or_variable): Dump info about
4756 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
4757 Steve Munroe <munroesj52@gmail.com>
4759 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
4760 (_mm_andnot_si128): Likewise.
4761 (_mm_or_si128): Likewise.
4762 (_mm_xor_si128): Likewise.
4764 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
4767 * builtins.c (expand_builtin_memcmp): Inline the calls first
4768 when result_eq is false.
4769 (expand_builtin_strcmp): Inline the calls first.
4770 (expand_builtin_strncmp): Likewise.
4771 (inline_string_cmp): New routine. Expand a string compare
4772 call by using a sequence of char comparison.
4773 (inline_expand_builtin_string_cmp): New routine. Inline expansion
4774 a call to str(n)cmp/memcmp.
4775 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
4777 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
4779 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
4781 * config/arm/driver-arm.c: Include arm-native.h.
4782 (host_detect_local_cpu): Use auto-generated data tables.
4783 (vendors, arm_cpu_table): Delete. Move part information to ...
4784 * config/arm/arm-cpus.in: ... here.
4785 * config/arm/parsecpu.awk (gen_native): New function.
4786 (vendor, part): New CPU fields.
4787 (END): Add support for building the native CPU detection tables.
4788 * config/arm/t-arm (arm-native.h): Add build rule.
4789 (driver-arm.o): Add dependency on arm-native.h.
4791 2018-07-13 Richard Biener <rguenther@suse.de>
4794 * match.pd (addr1 - addr2): Allow either of the operand to
4797 2018-07-13 Tom de Vries <tdevries@suse.de>
4799 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
4802 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
4804 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
4805 arrays instead of numbered variables.
4807 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
4809 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
4810 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
4812 2018-07-13 Richard Biener <rguenther@suse.de>
4815 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
4816 instead of get_context_die.
4818 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
4819 Richard Biener <rguenther@suse.de>
4822 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
4823 that the loop latch destination where phi is defined.
4825 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
4827 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
4828 (riscv_expand_epilogue): Add assertion to check interrupt mode.
4829 (riscv_set_current_function): Extract getting interrupt type to new
4831 (riscv_get_interrupt_type): New function.
4832 (riscv_merge_decl_attributes): New function, checking interrupt type is
4834 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
4836 2018-07-12 Paul Koning <ni1d@arrl.net>
4838 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
4841 2018-07-12 Paul Koning <ni1d@arrl.net>
4843 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
4844 zero reference, add doloop_end instead.
4845 * doc/md.texi (decrement_and_branch_until_zero): Remove.
4846 (Looping patterns): Remove decrement_and_branch_until_zero. Add
4847 detail for doloop_end.
4849 2018-07-12 Martin Sebor <msebor@redhat.com>
4852 * attribs.c (decl_attributes): Reject conflicting attributes before
4853 calling attribute handlers.
4855 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4857 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
4859 (gcc::dump_manager::get_dump_file_name): likewise.
4860 (dump_begin): Likewise.
4861 * dumpfile.h (dump_begin): Update prototype.
4862 (gcc::dump_manager::get_dump_file_name,
4863 gcc::dump_manager::get_dump_file_name): Update prototype.
4865 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4867 * internal-fn.h (vectorizable_internal_fn_p): New function.
4868 * tree-vect-slp.c (compatible_calls_p): Likewise.
4869 (vect_build_slp_tree_1): Remove nops argument. Handle calls
4870 to internal functions.
4871 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
4873 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4875 * fold-const.h (inverse_conditions_p): Declare.
4876 * fold-const.c (inverse_conditions_p): New function.
4877 * match.pd: Use inverse_conditions_p. Add folds of view_converts
4878 that test the inverse condition of a conditional internal function.
4879 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
4880 * internal-fn.c (internal_fn_mask_index): Handle conditional
4882 (vectorized_internal_fn_supported_p): New function.
4883 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
4884 (any_pred_load_store): Replace with...
4885 (need_to_predicate): ...this new variable.
4886 (redundant_ssa_names): New variable.
4887 (ifcvt_can_use_mask_load_store): Move initial checks to...
4888 (ifcvt_can_predicate): ...this new function. Handle tree codes
4889 for which a conditional internal function exists.
4890 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
4891 instead of ifcvt_can_use_mask_load_store. Update after variable
4893 (predicate_load_or_store): New function, split out from
4894 predicate_mem_writes.
4895 (check_redundant_cond_expr): New function.
4896 (value_available_p): Likewise.
4897 (predicate_rhs_code): Likewise.
4898 (predicate_mem_writes): Rename to...
4899 (predicate_statements): ...this. Use predicate_load_or_store
4900 and predicate_rhs_code.
4901 (combine_blocks, tree_if_conversion): Update after above name changes.
4902 (ifcvt_local_dce): Handle redundant_ssa_names.
4903 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
4904 general conditional functions.
4905 * tree-vect-stmts.c (vectorizable_call): Likewise.
4907 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4908 Alan Hayward <alan.hayward@arm.com>
4909 David Sherwood <david.sherwood@arm.com>
4911 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
4912 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
4913 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
4914 plus and minus and convert them into IFN_COND_FMA-based sequences.
4915 (convert_mult_to_fma): Handle conditional plus and minus.
4917 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4919 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
4920 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
4921 (cond_fnms_optab): New optabs.
4922 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
4924 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
4925 * internal-fn.h (get_conditional_internal_fn): Declare.
4926 (get_unconditional_internal_fn): Likewise.
4927 * internal-fn.c (cond_ternary_direct): New macro.
4928 (expand_cond_ternary_optab_fn): Likewise.
4929 (direct_cond_ternary_optab_supported_p): Likewise.
4930 (FOR_EACH_COND_FN_PAIR): Likewise.
4931 (get_conditional_internal_fn): New function.
4932 (get_unconditional_internal_fn): Likewise.
4933 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
4934 (gimple_match_op::gimple_match_op): Add a new overload for 5
4936 (gimple_match_op::set_op): Likewise.
4937 (gimple_resimplify5): Declare.
4938 * genmatch.c (decision_tree::gen): Generate simplifications for
4940 * gimple-match-head.c (gimple_simplify): Define an overload for
4941 5 operands. Handle calls with 5 arguments in the top-level overload.
4942 (convert_conditional_op): Handle conversions from unconditional
4943 internal functions to conditional ones.
4944 (gimple_resimplify5): New function.
4945 (build_call_internal): Pass a fifth operand.
4946 (maybe_push_res_to_seq): Likewise.
4947 (try_conditional_simplification): Try converting conditional
4948 internal functions to unconditional internal functions.
4949 Handle 3-operand unconditional forms.
4950 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
4951 Define ternary equivalents of the current rules for binary conditional
4953 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
4955 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
4956 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
4957 (optab): Handle them.
4958 (SVE_COND_FP_TERNARY): New int iterator.
4959 (sve_fmla_op, sve_fmad_op): New int attributes.
4960 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
4961 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
4962 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
4964 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4966 * target.def (preferred_else_value): New target hook.
4967 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
4968 * doc/tm.texi: Regenerate.
4969 * targhooks.h (default_preferred_else_value): Declare.
4970 * targhooks.c (default_preferred_else_value): New function.
4971 * internal-fn.h (conditional_internal_fn_code): Declare.
4972 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
4973 (get_conditional_internal_fn): Use it.
4974 (conditional_internal_fn_code): New function.
4975 * gimple-match.h (gimple_match_cond): New struct.
4976 (gimple_match_op): Add a cond member function.
4977 (gimple_match_op::gimple_match_op): Update all forms to take a
4979 * genmatch.c (expr::gen_transform): Use the same condition as res_op
4980 for the suboperation, but don't specify a particular else_value.
4981 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
4982 (visit_nary_op, visit_reference_op_load): Pass
4983 gimple_match_cond::UNCOND to the gimple_match_op constructor.
4984 * gimple-match-head.c: Include tree-eh.h
4985 (convert_conditional_op): New function.
4986 (maybe_resimplify_conditional_op): Likewise.
4987 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
4988 (gimple_resimplify2): Likewise.
4989 (gimple_resimplify3): Likewise.
4990 (gimple_resimplify4): Likewise.
4991 (maybe_push_res_to_seq): Return null for conditional operations.
4992 (try_conditional_simplification): New function.
4993 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
4995 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
4997 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
4999 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
5001 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5003 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5005 (hash_tree): Do not hash DECL_FCONTEXT
5006 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
5007 Do not stream DECL_FCONTEXT.
5008 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
5009 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
5011 2018-07-12 Richard Biener <rguenther@suse.de>
5014 * dwarf2out.c (gen_block_die): Only output blocks when they have
5015 at least one !DECL_IGNORED_P variable.
5017 2018-07-12 Richard Biener <rguenther@suse.de>
5020 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5021 Remove -mieee-fp handling.
5023 2018-07-12 Richard Biener <rguenther@suse.de>
5025 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
5026 left-over from last patch.
5028 2018-07-12 Jakub Jelinek <jakub@redhat.com>
5030 PR tree-optimization/86492
5031 * gimple-ssa-store-merging.c
5032 (imm_store_chain_info::coalesce_immediate_stores): Call
5033 check_no_overlap even for the merge_overlapping case. Formatting fix.
5035 2018-07-12 Richard Biener <rguenther@suse.de>
5038 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
5039 move possibly trapping operations into the conditional.
5041 2018-07-12 Richard Biener <rguenther@suse.de>
5043 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
5044 (vn_lookup_simplify_result): Remove recursion limit applied
5046 (vn_nary_build_or_lookup_1): Adjust.
5047 (try_to_simplify): Likewise.
5048 * gimple-match-head.c (gimple_resimplify1): Instead apply one
5050 (gimple_resimplify2): Likewise.
5051 (gimple_resimplify3): Likewise.
5052 (gimple_resimplify4): Likewise.
5054 2018-07-11 Jakub Jelinek <jakub@redhat.com>
5056 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
5057 Use __mmask64 type instead of __mmask8 for __M argument.
5058 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
5059 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
5061 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
5062 __mmask16 for __M argument.
5063 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
5064 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
5065 to __mmask16 instead of __mmask8.
5066 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
5067 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
5068 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
5069 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
5070 instead of __mmask16 for __U argument.
5071 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
5072 __mmask16 instead of __mmask8 for __U argument.
5073 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
5075 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
5077 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
5079 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
5081 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
5083 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
5085 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
5087 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
5089 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
5091 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
5093 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
5094 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
5095 return type as well as __M argument type and all casts from __mmask8
5097 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
5098 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
5099 return type as well as __M argument type and all casts from __mmask8
5101 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
5102 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
5103 return type as well as __M argument type and all casts from __mmask8
5105 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
5106 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
5107 return type as well as __M argument type and all casts from __mmask8
5109 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
5110 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
5113 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
5115 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
5116 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
5119 2018-07-11 Paul Koning <ni1d@arrl.net>
5121 * doc/md.texi (define_subst): Document how multiple occurrences of
5122 the same argument in the replacement pattern are handled.
5124 2018-07-11 Paul Koning <ni1d@arrl.net>
5126 * doc/extend.texi (Common Variable Attributes): Move "mode" into
5128 (Common Type Attributes): Add "mode" attribute.
5130 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
5132 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
5133 stream DECL_ORIGINAL_TYPE.
5134 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
5135 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
5136 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
5137 Do not walk original type.
5138 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
5140 (write_ts_decl_non_common_tree_pointers): Do not stream
5142 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
5143 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
5145 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5147 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
5148 thread twice from the same starting edge.
5150 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5152 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
5153 * gimple.c (gimple_call_nonnull_result_p): ...here...
5154 (gimple_call_nonnull_arg): ...and here.
5155 * gimple.h (gimple_call_nonnull_result_p): New.
5156 (gimple_call_nonnull_arg): New.
5158 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
5160 * config/arm/arm-cpus.in: Move information from fpu field of each
5161 cpu definition to the isa field.
5162 * config/arm/parsecpu.awk (fpu): Delete match rule.
5163 (gen_comm_data): Don't add bits from the CPU's FPU entry.
5165 2018-07-11 Richard Biener <rguenther@suse.de>
5168 * dwarf2out.c (init_sections_and_labels): Use
5169 output_asm_line_debug_info consistently.
5170 (dwarf2out_early_finish): Likewise.
5171 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
5174 2018-07-11 Richard Biener <rguenther@suse.de>
5176 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
5177 Rework father_bb setting in a way to avoid propagating constants
5178 multiple times on a loop body.
5180 2018-07-10 Mark Wielaard <mark@klomp.org>
5183 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
5185 2018-07-10 Richard Biener <rguenther@suse.de>
5187 * hash-map.h (hash_map::iterator::operator*): Return
5188 references to key and value.
5190 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5193 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5194 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5195 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5196 (gimplify_omp_for): For composite loops, move outer
5197 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5198 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5199 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5200 TREE_LIST for both the original class iterator and the "last" helper
5201 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5202 loop, remember has_decl_expr from outer composite loops for the
5203 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5205 2018-07-09 Martin Sebor <msebor@redhat.com>
5209 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5210 accessing implicitly initialized array elements.
5211 * expr.c (string_constant): Handle string initializers of
5212 character arrays within aggregates.
5213 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5214 Store element offset. As a special case, handle zero size.
5215 (fold_nonarray_ctor_reference): Same.
5216 (fold_ctor_reference): Add argument. Store subobject offset.
5217 * gimple-fold.h (fold_ctor_reference): Add argument.
5219 2018-07-09 Paul Koning <ni1d@arrl.net>
5221 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5222 (pdp11_insn_cost): New function.
5223 (pdp11_md_asm_adjust): New function.
5224 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5225 (pdp11_rtx_costs): Update to match machine better.
5226 (output_addr_const_pdp11): Correct format mismatch warnings.
5227 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5228 * config/pdp11/pdp11.md: General change to add base_cost and/or
5229 length attributes for use by new pdp11_insn_cost function.
5230 (MIN_BRANCH): Correct definition.
5232 (doloop_end): Use standard pattern name for looping pattern.
5233 (doloop_end_nocc): New.
5234 (movsf): Add another constraint alternative.
5235 (zero_extendqihi2): Add constraint alternatives for not in place
5237 (zero_extendhisi2): Remove.
5238 (shift patterns): Add CC handling variants.
5242 (define_peephole2): New peephole to recognize mov that sets CC for
5245 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5247 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5248 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5251 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5254 * real.c (real_nextafter): Return true if result is denormal.
5256 2018-07-09 Martin Liska <mliska@suse.cz>
5258 * common.opt: Add back wrongly removed attribute.
5260 2018-07-09 Richard Biener <rguenther@suse.de>
5263 * dwarf2out.c (gen_block_die): For an early generated DIE
5264 always output high/low PC attributes.
5266 2018-07-09 Tom de Vries <tdevries@suse.de>
5268 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5269 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5270 onto VAR_DECL with abstract origin.
5272 2018-07-07 Jim Wilson <jimw@sifive.com>
5274 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5276 2018-07-07 Tom de Vries <tdevries@suse.de>
5278 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5280 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5282 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5284 (combine_bound): Use wide-int overflow calculation instead of
5286 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5288 * fold-const.c (int_const_binop_2): Same.
5289 (extract_muldiv_1): Same.
5290 (fold_div_compare): Same.
5291 (fold_abs_const): Same.
5293 * poly-int.h (add): Same.
5297 * predict.c (predict_iv_comparison): Same.
5298 * profile-count.c (slow_safe_scale_64bit): Same.
5299 * simplify-rtx.c (simplify_const_binary_operation): Same.
5300 * tree-chrec.c (tree_fold_binomial): Same.
5301 * tree-data-ref.c (split_constant_offset_1): Same.
5302 * tree-if-conv.c (idx_within_array_bound): Same.
5303 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5304 * tree-ssa-phiopt.c (minmax_replacement): Same.
5305 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5306 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5307 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5308 * wide-int.cc (wi::add_large): Same.
5309 (wi::mul_internal): Same.
5310 (wi::sub_large): Same.
5311 (wi::divmod_internal): Same.
5312 * wide-int.h: Change overflow type to overflow_type for neg, add,
5313 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5314 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5315 mul_internal, divmod_internal.
5316 (overflow_type): New enum.
5317 (accumulate_overflow): New.
5319 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5321 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5322 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5324 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5326 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5327 argument is checked for zero before entering loop, avoid checking again.
5329 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5331 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5332 * gimplify.h (generic_expr_could_trap_p): Likwise.
5333 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5335 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5337 PR tree-optimization/86401
5338 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5339 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5340 (fold_bit_and_mask): ... here. New helper function for match.pd.
5341 * fold-const.h (fold_bit_and_mask): Declare.
5342 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5344 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5347 * target.def (translate_mode_attribute): New hook.
5348 * targhooks.h (default_translate_mode_attribute): Declare.
5349 * targhooks.c (default_translate_mode_attribute): New function.
5350 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5351 * doc/tm.texi: Regenerate.
5352 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5353 (rs6000_translate_mode_attribute): New function.
5355 2018-07-06 Paul Koning <ni1d@arrl.net>
5357 * doc/md.texi (define_split): Document DONE and FAIL.
5358 (define_peephole2): Ditto.
5360 2018-07-05 Jeff Law <law@redhat.com>
5362 PR tree-optimization/86010
5363 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5364 both the head and tail of mem* and str* calls.
5366 2018-07-05 Jim Wilson <jimw@sifive.com>
5368 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5370 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5372 * config/aarch64/aarch64-simd.md: correct flags text for
5373 MIN_EXPR replacement.
5375 2018-07-05 James Clarke <jrtc27@jrtc27.com>
5377 * configure: Regenerated.
5379 2018-07-05 Carl Love <cel@us.ibm.com>
5381 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
5382 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
5383 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
5384 VSX_BUILTIN_DOUBLEL_V4SF.
5386 2018-07-05 Martin Sebor <msebor@redhat.com>
5389 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5390 than its domain to compute its the upper bound of a char array.
5392 2018-07-05 Nathan Sidwell <nathan@acm.org>
5394 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
5395 * doc/cpp.texi: Update comment.
5396 * doc/tm.texi: Rebuilt.
5397 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
5398 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
5399 * doc/extend.texi (Backwards Compatibility): Clarify it is system
5400 headers affected by extern "C".
5401 * system.h: Poison NO_IMPLICIT_EXTERN_C.
5402 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
5403 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
5404 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
5405 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
5406 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
5407 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
5408 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
5409 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
5410 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
5411 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
5412 config/sparc/sp64-elf.h, config/spu/spu.h,
5413 config/stormy16/stormy16.h, config/v850/v850.h,
5414 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
5415 define NO_IMPLICIT_EXTERN_C.
5416 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
5418 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5421 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5422 instead of GET_MODE_SIZE when comparing Units.
5424 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5427 * rtlanal.c (set_noop_p): Constrain on mode change,
5428 include hard-reg-set.h
5430 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5432 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
5434 2018-07-05 Jakub Jelinek <jakub@redhat.com>
5437 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5440 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5442 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5444 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5447 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5449 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5451 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
5454 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
5456 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
5458 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
5459 neg_*_op* variables.
5461 2018-07-04 Martin Liska <mliska@suse.cz>
5463 * tree-switch-conversion.c: Define
5464 max_ratio_for_speed and max_ratio_for_size constants.
5466 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5467 Martin Liska <mliska@suse.cz>
5472 * common.opt: Rename align options with 'str_' prefix.
5473 * common/config/i386/i386-common.c (set_malign_value): New
5475 (ix86_handle_option): Use it to set -falign-* options/
5476 * config/aarch64/aarch64-protos.h (struct tune_params): Change
5477 type from int to string.
5478 * config/aarch64/aarch64.c: Update default values from int
5480 * config/alpha/alpha.c (alpha_override_options_after_change):
5482 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
5483 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5484 max skip conditionally.
5485 * config/i386/freebsd.h (SUBALIGN_LOG): New.
5486 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5487 max skip conditionally.
5488 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5489 max skip conditionally.
5490 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
5491 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5492 max skip conditionally.
5493 * config/i386/i386.c (struct ptt): Change type from int to
5495 (ix86_default_align): Set default values.
5496 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
5497 max skip conditionally.
5498 * config/i386/iamcu.h (SUBALIGN_LOG): New.
5499 (ASM_OUTPUT_MAX_SKIP_ALIGN):
5500 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
5501 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5502 max skip conditionally.
5503 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
5504 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
5505 * config/i386/x86-64.h (SUBALIGN_LOG): New.
5506 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5507 max skip conditionally.
5508 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5509 * config/ia64/ia64.c (ia64_option_override): Set default values
5510 for alignment options.
5511 * config/m68k/m68k.c: Handle new str_align_* options.
5512 * config/mips/mips.c (mips_set_compression_mode): Change
5514 (mips_option_override): Set default values for options.
5515 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5517 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5519 * config/rx/rx.c (rx_option_override): Likewise.
5520 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
5521 (LABEL_ALIGN): Use align_labels_log.
5522 (LOOP_ALIGN): Use align_loops_align.
5523 * config/s390/s390.c (s390_asm_output_function_label): Use new
5525 * config/sh/sh.c (sh_override_options_after_change):
5526 Change type of constants.
5527 * config/spu/spu.c (spu_sched_init): Likewise.
5528 * config/sparc/sparc.c (sparc_option_override): Set default
5530 * config/visium/visium.c (visium_option_override): Likewise.
5531 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
5532 emit p2align format with last argument if it's not needed.
5533 * doc/invoke.texi: Document extended format of -falign-*.
5534 * final.c: Use align_labels alignment.
5535 * flags.h (struct target_flag_state): Change type to use
5537 (struct align_flags_tuple): New.
5538 (struct align_flags): Likewise.
5539 (align_loops_log): Redefine macro to use new types.
5540 (align_loops_max_skip): Redefine macro to use new types.
5541 (align_jumps_log): Redefine macro to use new types.
5542 (align_jumps_max_skip): Redefine macro to use new types.
5543 (align_labels_log): Redefine macro to use new types.
5544 (align_labels_max_skip): Redefine macro to use new types.
5545 (align_functions_log): Redefine macro to use new types.
5546 (align_loops): Redefine macro to use new types.
5547 (align_jumps): Redefine macro to use new types.
5548 (align_labels): Redefine macro to use new types.
5549 (align_functions): Redefine macro to use new types.
5550 (align_functions_max_skip): Redefine macro to use new types.
5551 (align_loops_value): New macro.
5552 (align_jumps_value): New macro.
5553 (align_labels_value): New macro.
5554 (align_functions_value): New macro.
5555 * function.c (invoke_set_current_function_hook): Propagate
5556 alignment values from flags to global variables default in
5558 * ipa-icf.c (sem_function::equals_wpa): Use
5559 cl_optimization_option_eq instead of memcmp.
5560 * lto-streamer.h (cl_optimization_stream_out): Support streaming
5562 (cl_optimization_stream_in): Likewise.
5563 * optc-save-gen.awk: Support strings in cl_optimization.
5564 * opth-gen.awk: Likewise.
5565 * opts.c (finish_options): Remove error checking of invalid
5567 (MAX_CODE_ALIGN): Remove.
5568 (MAX_CODE_ALIGN_VALUE): Likewise.
5569 (parse_and_check_align_values): New function.
5570 (check_alignment_argument): Likewise.
5571 (common_handle_option): Use check_alignment_argument.
5572 * opts.h (parse_and_check_align_values): Declare.
5573 * toplev.c (init_alignments): Remove.
5574 (read_log_maxskip): New.
5575 (parse_N_M): Likewise.
5576 (parse_alignment_opts): Likewise.
5577 (backend_init_target): Remove usage of init_alignments.
5578 * toplev.h (parse_alignment_opts): Declare.
5579 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
5581 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
5582 * tree.c (cl_option_hasher::equal): New.
5583 * varasm.c: Use new global macros.
5585 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5587 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
5588 Use a simpler align directive also if MAXSKIP = ALIGN-1.
5589 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5590 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5591 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5592 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5593 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
5594 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
5595 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
5596 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5597 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5598 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5599 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5601 2018-07-04 Martin Liska <mliska@suse.cz>
5602 Jonathan Wakely <jwakely@redhat.com>
5604 * coverage.c: Use correct type.
5605 * doc/invoke.texi: Language correction.
5607 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
5610 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
5611 ENDBRANCH for non-tail call which may return via indirect branch.
5612 * doc/extend.texi: Document indirect_return attribute.
5614 2018-07-03 Martin Sebor <msebor@redhat.com>
5616 PR tree-optimization/86274
5617 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
5619 (format_floating): Correct handling of infinities and NaNs.
5621 2018-07-03 Martin Sebor <msebor@redhat.com>
5623 * print-tree.c (print_real_cst): New function.
5624 (print_node_brief): Call it.
5625 (print_node): Ditto.
5627 2018-07-03 Jeff Law <law@redhat.com>
5629 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
5630 into a single pattern.
5632 * config/h8300/h8300.md (ors code_iterator): New.
5633 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
5634 a single pattern and single splitter.
5635 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
5636 (iorqi3_1, xorqi3_1): Likewise.
5637 (iorqi3, xorqi3 expanders): Similarly.
5639 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
5640 (movmd_internal) into a single pattern using the P mode iterator.
5641 (movmd splitters): Similarly.
5642 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
5643 (movsd splitters): Similarly.
5645 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
5646 ADDB, ADDW and ADDL into a single ADD attribute which selects the
5647 right table based on the size of the operand.
5648 * config/h8300/h8300.md (length_table): Corresponding changes. All
5649 references to "addb", "addw" and "addl" changed to "add".
5650 (btst patterns): Merge two variants into a single pattern.
5651 (tstqi, tsthi): Likewise.
5652 (addhi3_incdec, addsi3_incdec): Likewise.
5653 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
5654 (mulhi3, mulsi3): Likewise.
5655 (udivhi3, udivsi3): Likewise.
5656 (divhi3, divsi3): Likewise.
5657 (andorqi3, andorhi3, andorsi3): Likewise.
5659 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
5662 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
5663 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
5665 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5667 PR tree-optimization/85694
5668 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
5669 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
5671 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
5672 (<u>avg<mode>3_ceil): New patterns.
5674 2018-07-03 David Malcolm <dmalcolm@redhat.com>
5676 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
5677 scan-tree-dump directive.
5678 * gcc.dg/vect/slp-perm-2.c: Likewise.
5679 * gcc.dg/vect/slp-perm-3.c: Likewise.
5680 * gcc.dg/vect/slp-perm-5.c: Likewise.
5681 * gcc.dg/vect/slp-perm-6.c: Likewise.
5682 * gcc.dg/vect/slp-perm-7.c: Likewise.
5683 * gcc.dg/vect/slp-perm-8.c: Likewise.
5685 2018-07-03 Marek Polacek <polacek@redhat.com>
5688 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
5690 2018-07-03 Richard Biener <rguenther@suse.de>
5693 * tree-ssa-structalias.c (find_func_clobbers): Properly
5694 handle indirect calls.
5696 2018-07-03 Jeff Law <law@redhat.com>
5698 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
5699 (shifts): New code iterator.
5700 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
5701 expander. Fix HImode handling on H8/SX.
5702 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
5703 (subqi3, subhi3, subsi3 expanders): Likewise.
5704 (andqi3, andhi3, andsi3 expanders): Likewise.
5705 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
5706 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
5707 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
5708 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
5709 (zero_extendqihi2, zero_extendqisi2): Likewise.
5710 (extendqihi2, extendqisi2): Likewise.
5711 (rotlqi3, rotlhi3, rotlsi3): Likewise.
5712 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
5713 (rotlqi3_1, rotlhi3_1): Likewise.
5714 (logicalhi3_sn, logicalsi3_sn): Likewise.
5715 (logicalhi3, logicalsi3): Likewise.
5717 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5719 * tree-vect-patterns.c (vect_recog_rotate_pattern)
5720 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5721 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
5722 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
5723 type to append_pattern_def_seq instead of creating a stmt_vec_info
5725 (build_mask_conversion): Likewise. Remove vinfo argument.
5726 (vect_add_conversion_to_patterm): Likewise, renaming to...
5727 (vect_add_conversion_to_pattern): ...this.
5728 (vect_recog_mask_conversion_pattern): Update call to
5729 build_mask_conversion. Pass the vector type to
5730 append_pattern_def_seq here too.
5731 (vect_recog_gather_scatter_pattern): Update call to
5732 vect_add_conversion_to_pattern.
5734 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5736 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
5737 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5738 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
5739 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
5740 STMT_VINFO_PATTERN_DEF_SEQ to null here.
5741 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
5742 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
5743 append_pattern_def_seq instead of new_pattern_def_seq.
5744 (vect_recog_divmod_pattern): Do both of the above.
5745 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
5748 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5750 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
5751 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
5752 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
5753 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
5754 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
5755 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5756 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
5757 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5758 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5759 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
5760 parameter with a single stmt_vec_info.
5761 (vect_recog_func_ptr): Likewise.
5762 (vect_recog_gather_scatter_pattern): Likewise, folding in...
5763 (vect_try_gather_scatter_pattern): ...this.
5764 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
5765 the stmt_vec_info of the statement to be matched. Don't clear
5766 STMT_VINFO_RELATED_STMT.
5767 (vect_pattern_recog): Update call accordingly.
5769 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5771 PR tree-optimization/85694
5772 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
5773 (uavgM3_ceil): Document new optabs.
5774 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
5775 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
5777 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
5778 (savg_ceil_optab): New optabs.
5779 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
5780 (vect_vect_recog_func_ptrs): Add it.
5781 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
5782 constant directly from the associated lhs.
5784 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5786 * tree-vect-patterns.c (vect_split_statement): New function.
5787 (vect_convert_input): Use it to try to split an existing cast.
5789 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5791 * poly-int.h (print_hex): New function.
5792 * dumpfile.h (dump_dec, dump_hex): Declare.
5793 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
5794 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
5795 min_input_precision, operation_precision and operation_sign.
5796 * tree-vect-patterns.c (vect_get_range_info): New function.
5797 (vect_same_loop_or_bb_p, vect_single_imm_use)
5798 (vect_operation_fits_smaller_type): Delete.
5799 (vect_look_through_possible_promotion): Add an optional
5800 single_use_p parameter.
5801 (vect_recog_over_widening_pattern): Rewrite to use new
5802 stmt_vec_info infomration. Handle one operation at a time.
5803 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
5804 (vect_truncatable_operation_p, vect_set_operation_type)
5805 (vect_set_min_input_precision): New functions.
5806 (vect_determine_min_output_precision_1): Likewise.
5807 (vect_determine_min_output_precision): Likewise.
5808 (vect_determine_precisions_from_range): Likewise.
5809 (vect_determine_precisions_from_users): Likewise.
5810 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
5811 (vect_vect_recog_func_ptrs): Put over_widening first.
5813 (vect_pattern_recog): Call vect_determine_precisions.
5815 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5817 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
5818 statements that have been replaced by further pattern statements.
5819 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
5821 2018-07-03 Richard Biener <rguenther@suse.de>
5823 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
5824 always set *dt. Dump vectype in vectype overload.
5825 * dumpfile.h (dump_gimple_expr): New function.
5826 (dump_gimple_expr_loc): Likewise.
5827 * dumpfile.c (dump_gimple_expr): New function.
5828 (dump_gimple_expr_loc): Likewise.
5830 2018-07-02 Jeff Law <law@redhat.com>
5832 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
5833 the H8/300, H8/300H and H8/S variants into a single pattern.
5834 (movhi_h8300, movqi_h8300hs): Similarly.
5835 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
5836 (QHI mode iterator): New.
5838 * config/h8300/h8300.md: Remove trailing whitespace.
5840 2018-07-02 Jim Wilson <jimw@sifive.com>
5842 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
5843 instead of emit_insn for interrupt returns.
5844 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
5845 (riscv_sret, riscv_uret): Likewise.
5847 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5849 * pretty-print.c (selftest::test_pp_format): Move save and restore
5850 of quotes to class auto_fix_quotes, and add an instance.
5851 * selftest.c: Include "intl.h".
5852 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
5853 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
5854 * selftest.h (selftest::auto_fix_quotes): New class.
5856 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
5858 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
5859 (aarch64_sve_prepare_conditional_op): Remove.
5860 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
5861 Allow aarch64_simd_reg_or_zero as select operand; remove
5862 the aarch64_sve_prepare_conditional_op call.
5863 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
5864 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5865 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
5866 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
5867 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
5868 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
5869 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
5870 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
5871 and a splitters to match all of the *_any patterns.
5872 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
5874 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
5875 (SVE_COND_FP_BINARY_REV): Remove.
5876 (sve_int_op_rev, sve_fp_op_rev): New.
5877 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
5878 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
5879 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
5880 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
5881 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
5882 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
5883 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
5884 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
5885 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
5887 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
5888 Remove match_dup 1 from the inner unspec.
5889 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5891 * config/aarch64/aarch64.md (movprfx): New attr.
5892 (length): Default movprfx to 8.
5893 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
5894 (*madd<SVE_I>, *msub<SVE_I): Likewise.
5895 (*<su>mul<SVE_I>3_highpart): Likewise.
5896 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
5897 (*v<ASHIFT><SVE_I>3): Likewise.
5898 (*<su><MAXMIN><SVE_I>3): Likewise.
5899 (*<su><MAXMIN><SVE_F>3): Likewise.
5900 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
5901 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
5902 (*div<SVE_F>4): Likewise.
5904 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
5906 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
5909 2018-07-02 Richard Biener <rguenther@suse.de>
5911 PR tree-optimization/86363
5912 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
5913 memset argument refers to a non-variable address.
5915 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5917 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
5918 {PLUS,MINUS}_EXPR code to...
5919 (adjust_symbolic_bound): ...here,
5920 (combine_bound): ...here,
5921 (set_value_range_with_overflow): ...and here.
5923 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5925 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
5927 (extract_range_from_abs_expr): ...here.
5929 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
5931 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
5932 -fno-omit-frame-pointer when not optimizing.
5934 2018-07-02 Martin Liska <mliska@suse.cz>
5937 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5938 (propagate_nothrow): Likewise.
5940 2018-07-02 Martin Liska <mliska@suse.cz>
5943 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
5945 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5947 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
5948 function in r262149, changing "loc" param from source_location to
5949 const dump_location_t &.
5950 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
5951 declaration, as above.
5953 2018-07-01 Paul Koning <ni1d@arrl.net>
5955 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5956 -munit-asm, -mgnu-asm, -mdec-asm.
5957 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
5958 (pdp11_output_labelref): New.
5959 (pdp11_output_def): New.
5960 (pdp11_output_addr_vec_elt): New.
5961 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
5962 %# and %@ format codes.
5963 (pdp11_option_override): New.
5964 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
5965 (pdp11_output_ident): New.
5966 (pdp11_asm_named_section): New.
5967 (pdp11_asm_init_sections): New.
5968 (pdp11_file_start): New.
5969 (pdp11_file_end): New.
5970 (output_ascii): Use .ascii/.asciz for -mdec-asm.
5971 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
5972 %o, like %c but octal.
5973 (pdp11_option_override): New.
5974 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
5976 (DATA_SECTION_ASM_OP): Ditto.
5977 (READONLY_DATA_SECTION_ASM_OP): New.
5978 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
5979 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
5980 (ASM_OUTPUT_LABELREF): Ditto.
5981 (ASM_OUTPUT_DEF): Ditto.
5982 (ASM_OUTPUT_EXTERNAL): New.
5983 (ASM_OUTPUT_SOURCE_FILENAME): New.
5984 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
5985 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
5986 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
5987 %# and %@ format codes.
5988 * config/pdp11/pdp11.opt (mgnu-asm): New.
5989 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
5990 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
5991 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
5993 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
5995 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
5996 dereferencing path[] beyond its length.
5998 (debug_all_paths): New.
5999 (rewire_first_differing_edge): New.
6000 (adjust_paths_after_duplication): New.
6001 (duplicate_thread_path): Call adjust_paths_after_duplication.
6003 (thread_through_all_blocks): Add new argument to
6004 duplicate_thread_path.
6006 2018-06-30 Jim Wilson <jimw@sifive.com>
6008 * config/riscv/predicates.md (p2m1_shift_operand): New.
6009 (high_mask_shift_operand): New.
6010 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
6011 pattern using p2m1_shift_operand.
6012 (lshsi3_zero_extend_3+2): New combiner pattern using
6013 high_mask_shift_operand.
6015 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6017 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
6019 (vect_recog_rotate_pattern): ...here.
6020 (vect_convert_input): Try to insert casts of invariants in the
6022 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
6023 preheader to be empty.
6025 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6027 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
6028 vector type. If given, install it in the new statement's
6030 (vect_element_precision): New function.
6031 (vect_unpromoted_value): New struct.
6032 (vect_unpromoted_value::vect_unpromoted_value): New function.
6033 (vect_unpromoted_value::set_op): Likewise.
6034 (vect_look_through_possible_promotion): Likewise.
6035 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
6036 (vect_widened_op_tree, vect_convert_input): Likewise.
6037 (vect_convert_inputs, vect_convert_output): Likewise.
6038 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
6039 to handle the optional cast of the multiplication result and
6040 vect_widened_op_tree to detect the widened multiplication itself.
6041 Do not require the input and output of promotion casts to have
6042 the same sign, but base the signedness of the operation on the
6043 input rather than the result. If the pattern includes two
6044 promotions, check that those promotions have the same sign.
6045 Do not restrict the MULT_EXPR handling to a double-width result;
6046 handle quadruple-width results and wider. Use vect_convert_inputs
6047 to convert the inputs to the common type.
6048 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
6049 to handle the optional cast of the ABS result. Also allow a sign
6050 change or a sign extension between the ABS and MINUS.
6051 Use vect_widened_op_tree to detect the widened subtraction and use
6052 vect_convert_inputs to convert the inputs to the common type.
6053 (vect_handle_widen_op_by_const): Delete.
6054 (vect_recog_widen_op_pattern): New function.
6055 (vect_recog_widen_mult_pattern): Use it.
6056 (vect_recog_widen_shift_pattern): Likewise.
6057 (vect_recog_widen_sum_pattern): Use
6058 vect_look_through_possible_promotion to handle the promoted
6061 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6063 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
6064 the containing gimple_seq *.
6065 * gimple-iterator.h (gsi_for_stmt): Declare it.
6066 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
6067 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
6068 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6069 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6070 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
6072 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
6074 (vect_mark_pattern_stmts): ...here. Handle cases in which the
6075 statement being replaced is part of an existing pattern
6076 definition sequence, inserting the new pattern statements before
6078 (vect_pattern_recog_1): Don't return a bool. If the statement
6079 is already part of a pattern, instead apply pattern matching
6080 to the pattern definition statements. Don't clear the
6081 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
6082 (vect_pattern_recog): Don't break after the first match;
6083 continue processing the pattern definition statements instead.
6084 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
6086 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6088 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6089 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6090 (vect_recog_widen_sum_pattern): Use it.
6092 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6094 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6095 phi is not a pattern statement and has not been replaced by
6096 a pattern statement.
6097 * tree-vect-patterns.c (type_conversion_p): Don't check
6098 STMT_VINFO_IN_PATTERN_P.
6099 (vect_recog_vector_vector_shift_pattern): Likewise.
6100 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
6101 the pattern statement rather than the original statement; check
6102 directly for a WIDEN_MULT_EXPR here.
6103 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
6104 vect_is_simple_use to return the pattern statement rather
6105 than the original statement; use is_pattern_stmt_p to check
6106 for such a pattern statement.
6107 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
6108 to return the pattern statement rather than the original statement;
6109 don't do the same transformation here.
6110 (vect_is_simple_use): If the defining statement has been replaced
6111 by a pattern statement, return the pattern statement instead.
6112 Remove the corresponding (local) transformation from the vectype
6115 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6117 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
6118 end and default to null.
6119 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6120 (vectorizable_reduction): Update calls accordingly, dropping the
6121 gimple ** argument if the passed-back statement isn't needed.
6122 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
6123 (vect_recog_rotate_pattern): Likewise.
6124 (vect_recog_mask_conversion_pattern): Likewise.
6125 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6126 (vect_mask_constant_operand_p): Likewise.
6127 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
6128 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
6129 (get_group_load_store_type, get_load_store_type): Likewise.
6130 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
6131 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
6132 (vectorizable_conversion, vectorizable_assignment): Likewise.
6133 (vectorizable_shift, vectorizable_operation): Likewise.
6134 (vectorizable_store, vect_is_simple_cond): Likewise.
6135 (vectorizable_condition, vectorizable_comparison): Likewise.
6136 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
6137 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
6138 and move it to the end. Cope with null def_stmt_outs.
6140 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6142 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
6144 2018-06-29 Jeff Law <law@redhat.com>
6146 * config/v850/v850.c (v850_legitimate_address_p): Handle large
6147 displacements for TARGET_V850E2V3 and newer.
6148 (TARGET_LRA_P): Remove. Defaults to LRA now.
6149 * config/v850/v850.md (sign23byte_load): Remove.
6150 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
6151 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
6153 2018-06-29 Martin Liska <mliska@suse.cz>
6156 * coverage.c (coverage_init): Mangle full path name.
6157 * doc/invoke.texi: Document the change.
6158 * gcov-io.c (mangle_path): New.
6159 * gcov-io.h (mangle_path): Likewise.
6160 * gcov.c (mangle_name): Use mangle_path for path mangling.
6162 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6164 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
6165 if starting source register is not even.
6167 2018-06-29 Martin Liska <mliska@suse.cz>
6169 PR tree-optimization/86263
6170 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
6171 Make edge redirection.
6173 2018-06-29 David Malcolm <dmalcolm@redhat.com>
6175 * dumpfile.c (dump_loc): Add indentation based on scope depth.
6176 (dump_scope_depth): New variable.
6177 (get_dump_scope_depth): New function.
6178 (dump_begin_scope): New function.
6179 (dump_end_scope): New function.
6180 * dumpfile.h (get_dump_scope_depth): New declaration.
6181 (dump_begin_scope): New declaration.
6182 (dump_end_scope): New declaration.
6183 (class auto_dump_scope): New class.
6184 (AUTO_DUMP_SCOPE): New macro.
6185 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6188 2018-06-29 Richard Biener <rguenther@suse.de>
6190 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6191 compute_all_dependences succeeds.
6192 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6193 exceed --param loop-max-datarefs-for-datadeps.
6195 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6197 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6199 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6202 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6203 alternative 0 in preferred_for_speed attribute.
6205 2018-06-28 Paul Koning <ni1d@arrl.net>
6207 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6208 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6209 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6210 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6211 for shift insn patterns.
6213 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6215 * cgraph.c (cgraph_node::get_body): Replace assignments to
6216 "dump_file" with calls to set_dump_file.
6217 * dumpfile.c (alt_dump_file): Make static, and group with...
6218 (alt_flags): ...this definition.
6219 (dumps_are_enabled): New variable.
6220 (refresh_dumps_are_enabled): New function.
6221 (set_dump_file): New function.
6222 (set_alt_dump_file): New function.
6223 (gcc::dump_manager::dump_start): Replace assignments to
6224 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6226 (gcc::dump_manager::dump_finish): Likewise.
6227 * dumpfile.h (alt_dump_file): Delete decl.
6228 (dumps_are_enabled): New variable decl.
6229 (set_dump_file): New function decl.
6230 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6232 * tree-nested.c (lower_nested_functions): Replace assignments to
6233 "dump_file" with calls to set_dump_file.
6235 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6237 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6238 goto_locus of each outgoing edge of each basic block.
6240 2018-06-28 Richard Biener <rguenther@suse.de>
6242 * dwarf2out.c (decl_scope_table): Remove.
6243 (push_decl_scope): Likewise.
6244 (pop_decl_scope): Likewise.
6245 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6246 (gen_struct_or_union_type_die): Likewise.
6247 (gen_tagged_type_die): Likewise.
6248 (dwarf2out_init): Do not initialize decl_scope_table.
6249 (dwarf2out_c_finalize): Do not free it.
6251 2018-06-28 Richard Biener <rguenther@suse.de>
6253 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6254 deciding whether to not re-use a DIE.
6256 2018-06-28 Richard Biener <rguenther@suse.de>
6258 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6259 DW_AT_abstract_origin attribute.
6261 2018-06-28 Martin Liska <mliska@suse.cz>
6263 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6264 Use newly introduced constants.
6265 * tree-switch-conversion.h (struct jump_table_cluster):
6266 Define max_ratio_for_size and max_ratio_for_speed.
6268 2018-06-28 Martin Liska <mliska@suse.cz>
6270 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6271 Add new checking assert to catch invalid state.
6272 (jump_table_cluster::can_be_handled): Handle single case
6274 (jump_table_cluster::is_beneficial): Bail out for such case.
6275 (bit_test_cluster::find_bit_tests):
6276 Add new checking assert to catch invalid state.
6277 (bit_test_cluster::can_be_handled): Handle single case
6279 (bit_test_cluster::is_beneficial): Bail out for such case.
6280 (switch_decision_tree::analyze_switch_statement):
6283 2018-06-28 Martin Liska <mliska@suse.cz>
6285 * common.opt: Introduce -completion option.
6286 * gcc.c (driver_handle_option): Handle it.
6287 (driver::main): Print completions if completion
6289 * opt-suggestions.c (option_proposer::get_completions):
6291 (option_proposer::suggest_completion): Likewise.
6292 (option_proposer::find_param_completions): Likewise.
6293 (verify_autocompletions): Likewise.
6294 (test_completion_valid_options): Likewise.
6295 (test_completion_valid_params): Likewise.
6296 (in_completion_p): Likewise.
6297 (empty_completion_p): Likewise.
6298 (test_completion_partial_match): Likewise.
6299 (test_completion_garbage): Likewise.
6300 (opt_proposer_c_tests): Likewise.
6301 * opt-suggestions.h: Declare new functions.
6302 * opts.c (common_handle_option): Handle OPT__completion_.
6303 * selftest-run-tests.c (selftest::run_tests): Add
6304 opt_proposer_c_tests.
6305 * selftest.c (assert_str_startswith): New.
6306 * selftest.h (assert_str_startswith): Likewise.
6307 (opt_proposer_c_tests): New.
6308 (ASSERT_STR_STARTSWITH): Likewise.
6310 2018-06-28 Martin Liska <mliska@suse.cz>
6312 * Makefile.in: Add opt-suggestions.o.
6313 * gcc-main.c: Include opt-suggestions.h.
6314 * gcc.c (driver::driver): Likewise.
6315 (driver::~driver): Remove m_option_suggestions.
6316 (driver::build_option_suggestions): Moved to option_proposer.
6317 (driver::suggest_option): Likewise.
6318 (driver::handle_unrecognized_options): Use option_proposer.
6319 * gcc.h (class driver): Add new memver m_option_proposer.
6320 * opt-suggestions.c: New file.
6321 * opt-suggestions.h: New file.
6323 2018-06-28 Martin Liska <mliska@suse.cz>
6325 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6326 (auto_string_vec::~auto_string_vec): Likewise.
6328 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6330 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6331 prevent_decl_creation_for_types fields up and add reset_location field.
6332 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6333 statement if id->reset_location is true.
6334 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6335 id->reset_location is true.
6336 (copy_phis_for_bb): Force input_location on the arguments if
6337 id->reset_location is true.
6338 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6339 is set on the function to be inlined.
6341 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6343 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6345 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6347 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6348 registers for Pmode.
6349 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6350 hard registers for the clobbered pseudo.
6352 2018-06-27 Paul Koning <ni1d@arrl.net>
6354 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6355 mutually exclusive options.
6356 * config/pdp11/constraints.md (h): New constraint.
6357 (O): Update definition to match shift code generation.
6358 (D): New constraint.
6359 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6361 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6363 (output_jump): Change arguments.
6364 (pdp11_fixed_cc_regs): New function.
6365 (pdp11_cc_mode): Ditto.
6366 (pdp11_expand_shift): Ditto.
6367 (pdp11_assemble_shift): Ditto.
6368 (pdp11_small_shift): Ditto.
6369 (pdp11_branch_cost): Remove.
6370 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6372 (pdp11_register_move_cost): Update for CC registers.
6373 (pdp11_rtx_costs): Add case for LSHIFTRT.
6374 (pdp11_output_jump): Add CCNZ mode conditional branches.
6375 (notice_update_cc_on_set): Remove.
6376 (pdp11_cc_mode): New function.
6377 (simple_memory_operand): Correct pre/post decrement case.
6378 (no_side_effect_operand): New function.
6379 (pdp11_regno_reg_class): Add CC_REGS class.
6380 (pdp11_fixed_cc_regs): New function.
6381 (pdp11_small_shift): New function.
6382 (pdp11_expand_shift): New function to expand shift insns.
6383 (pdp11_assemble_shift): New function to output shifts.
6384 (pdp11_branch_cost): Remove.
6385 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
6386 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
6387 (WCHAR_TYPE): Ditto.
6388 (PTRDIFF_TYPE): Ditto.
6389 (ADJUST_INSN_LENGTH): New macro.
6390 (FIXED_REGISTERS): Add CC registers.
6391 (CALL_USED_REGISTERS): Ditto.
6393 (REG_CLASS_NAMES): Ditto.
6394 (REG_CLASS_CONTENTS): Ditto.
6395 (SELECT_CC_MODE): Use new function.
6396 (TARGET_FLAGS_REGNUM): New macro.
6397 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
6398 (cc0_reg_rtx): Remove.
6399 (CC_STATUS_MDEP): Remove.
6400 (CC_STATUS_MDEFP_INIT): Remove.
6401 (CC_IN_FPU): Remove.
6402 (NOTICE_UPDATE_CC): Remove.
6403 (REGISTER_NAMES): Add CC registers.
6404 (BRANCH_COST): Change to constant 1.
6405 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
6407 * config/pdp11/pdp11.opt (mbcopy): Remove.
6408 (mbcopy-builtin): Remove.
6409 (mbranch-cheap): Remove.
6410 (mbranch-expensive): Remove.
6411 * config/pdp11/predicates.md (expand_shift_operand): Update to
6412 match shift code generation.
6413 (ccnz_operator): New predicate.
6414 * doc/invoke.texi (PDP-11 Options): Remove deleted options
6415 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
6416 Remove non-existent option -mabshi, -mno-abshi. Document mutually
6418 * doc/md.texi (PDP-11): Document new D and h constraints. Update
6419 description of O constraint.
6421 2018-06-27 Jeff Law <law@redhat.com>
6422 Austin Law <austinklaw@gmail.com>
6424 * config/v850/v850.md (addsi3_set_flags): New pattern.
6425 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
6426 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
6427 (zero_extendhisi2_v850_set_flags): Likewise.
6428 (zero_extendqisi2_v850_set_flags): Likewise.
6429 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
6430 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
6431 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
6433 * config/v850/v850-protos.h (notice_update_cc): Remove.
6434 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
6435 (v850_print_operand): Handle 'D' and "d".
6436 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
6437 Add handling of arithmetic/logical operations compared against zero.
6438 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
6439 Do not look at v850_compare_op, instead get mode from last argument.
6440 (v850_gen_compare): Remove
6441 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
6442 after reload for prologue insns.
6443 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
6445 (construct_save_jarl): Likewise.
6446 (TARGET_FLAGS_REGNUM): Define.
6447 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
6448 (NOTICE_UPDATE_CC): Remove.
6449 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
6450 than cc0. Conditionalize on reload_completed.
6451 (cmpsi_insn, setfcc_insn): Likewise.
6452 (tst1 splitter): Turn into define_and_split which sets the flags
6454 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
6455 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
6456 (cstoresf4, cstoredf4): Clobber the flags.
6457 (cmpsi, cmpsf, cmpdf): Remove expanders.
6458 (setf_insn): Remove pattern.
6459 (addsi3): Turn into define_and_split which clobbers the flags after
6460 reload and a suitable pattern (addsi3_clobber_flags) for use after
6462 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
6463 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
6464 (ashrsi3, ashrsi3_v850e2): Likewise.
6465 (bins): Clobber the flags.
6466 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
6467 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
6468 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
6469 (call_value_internal_short, call_value_internal_long): Likewise.
6470 (callt_save_interrupt, callt_return_interrupt): Likewise.
6471 (save_interrupt, return_interrupt): Likewise.
6472 (callt_save_all_interrupt, save_all_interrupt): Likewise.
6473 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
6474 (restore_all_interrupt, _restore_all_interrupt): Likewise.
6475 (All FP comparisons): Only allow after reload has completed.
6477 (divh, divhu): Tweak output template.
6478 (branch_z_normal, branch_z_invert): Remove
6479 (branch_nz_normal, branch_nz_invert): Likewise.
6480 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
6482 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
6483 * config/v850/v850.c (notice_update_cc): Remove.
6484 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
6485 (CC_NO_CARRY): Likewise.
6486 (NOTICE_UPDATE_CC): Define to nothing.
6487 * config/v850/v850.md: Remove block comment on cc0 handling
6488 Remove "cc" attribute from all patterns. Remove cc_status handling
6489 from all patterns. Minor formatting fixes.
6491 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6493 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
6494 (cortex-a76.cortex-a55): Likewise.
6495 * config/aarch64/aarch64-tune.md: Regenerate.
6496 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
6497 cortex-a76.cortex-a55.
6499 2018-06-27 Jeff Law <law@redhat.com>
6501 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
6502 (MULTILIB_DIRNAMES): Similarly.
6504 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
6506 * gimple.h (gimple_return_retbnd): Delete.
6507 (gimple_return_set_retbnd): Likewise.
6508 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
6509 gimple_return_set_retbnd.
6510 * gimple-pretty-print.c (dump_gimple_return): Remove call to
6511 gimple_return_retbnd and adjust.
6512 * tree-inline.h (struct copy_body_data): Remove retbnd field.
6513 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
6514 Explicitly return NULL in a couple more cases. Move assertion
6515 on debug statements and remove unreachable code.
6516 (reset_debug_binding): Do not test id->retbnd.
6517 (expand_call_inline): Do not set it.
6519 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6521 * configure.ac: Add --disable-gcov option.
6522 * configure: Regenerate.
6523 * Makefile.in: Honour @enable_gcov@.
6524 * doc/install.texi: Document --disable-gcov.
6526 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6528 * config/arm/arm-cpus.in (cortex-a76): New entry.
6529 (cortex-a76.cortex-a55): Likewise.
6530 * config/arm/arm-tables.opt: Regenerate.
6531 * config/arm/arm-tune.md: Likewise.
6532 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
6533 * doc/invoke.texi (ARM Options): Document cortex-a76 and
6534 cortex-a76.cortex-a55.
6536 2018-06-27 Tamar Christina <tamar.christina@arm.com>
6539 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
6541 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
6543 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
6545 (EPILOGUE_USES): Likewise.
6547 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
6549 * tree-inline.c (remap_location): New function extracted from...
6550 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
6551 (copy_phis_for_bb): ...here. Call remap_location.
6552 (copy_cfg_body): Adjust call to copy_edges_for_bb.
6554 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
6556 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
6557 unaligned vsx for 16B memset.
6559 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
6562 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
6563 ieee128_float_type_node to long_double_type_node unless
6564 TARGET_LONG_DOUBLE_128 is set.
6566 2018-06-26 David Malcolm <dmalcolm@redhat.com>
6568 * cfgloop.c (get_loop_location): Convert return type from
6569 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
6570 by implicit construction from rtx_insn *, and using
6571 dump_user_location_t::from_function_decl for the fallback case.
6572 * cfgloop.h (get_loop_location): Convert return type from
6573 location_t to dump_user_location_t.
6574 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
6575 dump_printf_loc to pass in a dump_location_t rather than a
6576 location_t, via the gimple stmt.
6577 * coverage.c (get_coverage_counts): Update calls to
6578 dump_printf_loc to pass in dump_location_t rather than a
6580 * doc/optinfo.texi (Dump types): Convert example of
6581 dump_printf_loc from taking "locus" to taking "insn". Update
6582 description of the "_loc" calls to cover dump_location_t.
6583 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
6585 (dump_user_location_t::dump_user_location_t): New constructors,
6586 from gimple *stmt and rtx_insn *.
6587 (dump_user_location_t::from_function_decl): New function.
6588 (dump_loc): Make static.
6589 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
6590 const dump_location_t &.
6591 (dump_generic_expr_loc): Delete.
6592 (dump_printf_loc): Convert param "loc" from location_t to
6593 const dump_location_t &.
6594 (selftest::test_impl_location): New function.
6595 (selftest::dumpfile_c_tests): New function.
6596 * dumpfile.h: Include "profile-count.h".
6597 (class dump_user_location_t): New class.
6598 (struct dump_impl_location_t): New struct.
6599 (class dump_location_t): New class.
6600 (dump_printf_loc): Convert 2nd param from source_location to
6601 const dump_location_t &.
6602 (dump_generic_expr_loc): Delete.
6603 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
6604 const dump_location_t &.
6605 * gimple-fold.c (fold_gimple_assign): Update call to
6606 dump_printf_loc to pass in a dump_location_t rather than a
6607 location_t, via the gimple stmt.
6608 (gimple_fold_call): Likewise.
6609 * gimple-loop-interchange.cc
6610 (loop_cand::analyze_iloop_reduction_var): Update for change to
6611 check_reduction_path.
6612 (tree_loop_interchange::interchange): Update for change to
6614 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
6615 change in return-type of find_loop_location.
6616 (graphite_regenerate_ast_isl): Likewise.
6617 * graphite-optimize-isl.c (optimize_isl): Likewise.
6618 * graphite.c (graphite_transform_loops): Likewise.
6619 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
6620 pass in a dump_location_t rather than a location_t, via the
6622 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6623 * ipa.c (walk_polymorphic_call_targets): Likewise.
6624 * loop-unroll.c (report_unroll): Convert "locus" param from
6625 location_t to dump_location_t.
6626 (decide_unrolling): Update for change to get_loop_location's
6628 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
6629 location_t to dump_user_location_t.
6630 (grid_find_single_omp_among_assignments_1): Updates calls to
6631 dump_printf_loc to pass in a dump_location_t rather than a
6632 location_t, via the gimple stmt.
6633 (grid_parallel_clauses_gridifiable): Convert "tloc" from
6634 location_t to dump_location_t. Updates calls to dump_printf_loc
6635 to pass in a dump_location_t rather than a location_t, via the
6637 (grid_inner_loop_gridifiable_p): Likewise.
6638 (grid_dist_follows_simple_pattern): Likewise.
6639 (grid_gfor_follows_tiling_pattern): Likewise.
6640 (grid_target_follows_gridifiable_pattern): Likewise.
6641 (grid_attempt_target_gridification): Convert initialization
6642 of local "grid" from memset to zero-initialization; FIXME: does
6643 this require C++11? Update call to dump_printf_loc to pass in a
6644 optinfo_location rather than a location_t, via the gimple stmt.
6645 * profile.c (read_profile_edge_counts): Updates call to
6646 dump_printf_loc to pass in a dump_location_t rather than a
6648 (compute_branch_probabilities): Likewise.
6649 * selftest-run-tests.c (selftest::run_tests): Call
6651 * selftest.h (dumpfile_c_tests): New decl.
6652 * tree-loop-distribution.c (pass_loop_distribution::execute):
6653 Update for change in return type of find_loop_location.
6654 * tree-parloops.c (parallelize_loops): Likewise.
6655 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
6656 "locus" from location_t to dump_user_location_t.
6657 (canonicalize_loop_induction_variables): Likewise.
6658 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
6659 for change in return type of find_loop_location.
6660 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
6661 to dump_printf_loc to pass in a dump_location_t rather than a
6662 location_t, via the stmt.
6663 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
6665 * tree-vect-loop-manip.c (find_loop_location): Convert return
6666 type from source_location to dump_user_location_t.
6667 (vect_do_peeling): Update for above change.
6668 (vect_loop_versioning): Update for change in type of
6670 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
6671 from location_t to dump_user_location_t.
6672 (vect_estimate_min_profitable_iters): Update for change in type
6674 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
6675 location_t to dump_location_t.
6676 (vect_slp_bb): Update for change in type of vect_location.
6677 * tree-vectorizer.c (vect_location): Convert from source_location
6678 to dump_user_location_t.
6679 (try_vectorize_loop_1): Update for change in vect_location's type.
6680 (vectorize_loops): Likewise.
6681 (increase_alignment): Likewise.
6682 * tree-vectorizer.h (vect_location): Convert from source_location
6683 to dump_user_location_t.
6684 (find_loop_location): Convert return type from source_location to
6685 dump_user_location_t.
6686 (check_reduction_path): Convert 1st param from location_t to
6687 dump_user_location_t.
6688 * value-prof.c (check_counter): Update call to dump_printf_loc to
6689 pass in a dump_user_location_t rather than a location_t; update
6690 call to error_at for change in type of "locus".
6691 (check_ic_target): Update call to dump_printf_loc to
6692 pass in a dump_user_location_t rather than a location_t, via the
6695 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
6697 * config/s390/s390.h (enum processor_flags): Do not use
6698 default tune parameter when -march was specified.
6700 2018-06-26 Jakub Jelinek <jakub@redhat.com>
6703 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
6704 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
6707 2018-06-26 Richard Biener <rguenther@suse.de>
6709 PR tree-optimization/86287
6711 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
6712 (vect_analyze_loop): Initialize n_stmts.
6714 2018-06-26 Richard Biener <rguenther@suse.de>
6717 * fold-const.c (fold_convertible_p): Pointer extension
6720 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
6723 * dwarf2out.c (loc_list_has_views): Adjust comments.
6724 (dw_loc_list): Split single cross-partition range with
6727 2018-06-25 Jeff Law <law@redhat.com>
6729 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
6730 on -mbig-switch by default.
6732 * config/v850/predicates.md (const_float_1_operand): Fix match_code
6734 (const_float_0_operand): Remove unused predicate.
6735 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
6736 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
6737 (recipsf2): New expander. Original pattern now called
6739 (recipdf2, recipdf2_insn): Similarly.
6740 (rsqrtsf2, rsqrtsf2_insn): Similarly
6741 (rsqrtdf2, rsqrtdf2_insn): Similarly
6743 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
6745 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
6746 Simplify logic for FreeBSD (twice).
6748 2018-06-25 Martin Sebor <msebor@redhat.com>
6750 PR tree-optimization/86204
6751 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
6752 a strnlen result if it's less than the length of the string.
6754 2018-06-25 Martin Sebor <msebor@redhat.com>
6756 PR tree-optimization/85700
6757 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
6758 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
6759 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
6761 2018-06-25 Martin Sebor <msebor@redhat.com>
6763 * doc/extend.texi (Zero-length arrays): Update and clarify.
6765 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
6767 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
6768 added IEEE/IBM long double multilib support on PowerPC little
6769 endian Linux systems.
6770 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
6771 (MULTILIB_DEFAULTS): Likewise.
6772 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6774 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
6775 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
6776 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
6778 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
6781 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
6782 (REORDER_45): Likewise.
6784 2018-06-25 Jeff Law <law@redhat.com>
6786 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
6787 dividend to 32 bits. Adjust length.
6788 (udivmodhi4): Cleanup output template. Fix length.
6790 2018-06-25 Carl Love <cel@us.ibm.com>
6792 * config/rs6000/vsx.md: Change word selector to prefered location.
6794 2018-06-25 Richard Biener <rguenther@suse.de>
6796 PR tree-optimization/86304
6797 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
6798 epilogue-if-converted loops as well.
6800 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
6802 * lto-section-out.c (lto_begin_section): Do not print section
6803 name for noaddr and unnumbered dumps.
6805 2018-06-25 Richard Biener <rguenther@suse.de>
6807 * tree-vectorizer.h (struct vec_info_shared): New structure
6808 with parts split out from struct vec_info and loop_nest from
6809 struct _loop_vec_info.
6810 (struct vec_info): Adjust accordingly.
6811 (struct _loop_vec_info): Likewise.
6812 (LOOP_VINFO_LOOP_NEST): Adjust.
6813 (LOOP_VINFO_DATAREFS): Likewise.
6814 (LOOP_VINFO_DDRS): Likewise.
6815 (struct _bb_vec_info): Likewise.
6816 (BB_VINFO_DATAREFS): Likewise.
6817 (BB_VINFO_DDRS): Likewise.
6818 (struct _stmt_vec_info): Add dr_aux member.
6819 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
6820 (DR_MISALIGNMENT_UNINITIALIZED): New.
6821 (set_dr_misalignment): Adjust.
6822 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
6823 (vect_analyze_loop): Adjust prototype.
6824 (vect_analyze_loop_form): Likewise.
6825 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6826 Compute dependences lazily.
6827 (vect_record_base_alignments): Use shared datarefs/ddrs.
6828 (vect_verify_datarefs_alignment): Likewise.
6829 (vect_analyze_data_refs_alignment): Likewise.
6830 (vect_analyze_data_ref_accesses): Likewise.
6831 (vect_analyze_data_refs): Likewise.
6832 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
6833 constructor parameter for shared part.
6834 (vect_analyze_loop_form): Pass in shared part and adjust.
6835 (vect_analyze_loop_2): Pass in storage for the number of
6836 stmts. Move loop nest finding to the caller. Compute
6838 (vect_analyze_loop): Pass in shared part.
6839 (vect_transform_loop): Verify shared datarefs are unchanged.
6840 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
6841 constructor parameter for shared part.
6842 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
6843 (vect_slp_bb): Verify shared datarefs are unchanged before
6845 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
6847 (new_stmt_vec_info): Initialize DR_AUX misalignment to
6848 DR_MISALIGNMENT_UNINITIALIZED.
6849 * tree-vectorizer.c (vec_info::vec_info): Add constructor
6850 parameter for shared part.
6851 (vec_info::~vec_info): Adjust.
6852 (vec_info_shared::vec_info_shared): New.
6853 (vec_info_shared::~vec_info_shared): Likewise.
6854 (vec_info_shared::save_datarefs): Likewise.
6855 (vec_info_shared::check_datarefs): Likewise.
6856 (try_vectorize_loop_1): Construct shared part live for analyses
6857 of a single loop for multiple vector sizes.
6858 * tree-parloops.c (gather_scalar_reductions): Adjust.
6860 2018-06-25 Richard Biener <rguenther@suse.de>
6862 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
6863 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
6864 (vect_analyze_data_refs): Remove similar code from here and
6865 simplify accordingly.
6867 2018-06-25 Richard Biener <rguenther@suse.de>
6869 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
6870 for reverse storage order accesses rather than asserting
6871 they cannot happen here.
6873 2018-06-25 Tom de Vries <tdevries@suse.de>
6876 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
6877 Use data16 instead of .byte for insn prefix.
6879 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
6882 * parser.c (make_char_string_pack): Pass this literal chars
6883 through cpp_interpret_string.
6884 (cp_parser_userdef_numeric_literal): Check the result of
6885 make_char_string_pack.
6887 2018-06-24 Maya Rashish <coypu@sdf.org>
6889 * ginclude/stddef.h: Simplify conditions around avoiding
6890 re-definition of __size_t.
6892 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6894 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
6895 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6897 2018-06-22 Maya Rashish <coypu@sdf.org>
6899 * doc/invoke.texi (mno-fancy-math-387): Update for changes
6900 made to OpenBSD and NetBSD through the years.
6902 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6904 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6905 behavior of vec_pack (vector double, vector double) to match
6906 behavior of vec_float2 (vector double, vector double).
6908 2018-06-22 Olivier Hainque <hainque@adacore.com>
6910 * gimplify.c (gimplify_function_tree): Prevent creation
6911 of a trampoline for the address of the current function
6912 passed to entry/exit instrumentation hooks.
6914 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
6917 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
6920 2018-06-22 Martin Liska <mliska@suse.cz>
6922 PR tree-optimization/86263
6923 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6924 Bail out if is_enabled is false.
6925 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
6927 (jump_table_cluster::is_enabled): New function.
6929 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6931 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6932 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
6933 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
6934 (lto_input_ts_binfo_tree_pointers): Likewise.
6935 * tree-streamer-out.c (streamer_write_tree_bitfields,
6936 write_ts_binfo_tree_pointers): Likewise.
6937 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
6939 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6941 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
6943 2018-06-22 Martin Liska <mliska@suse.cz>
6945 * symbol-summary.h (get): Make it pure and inline move
6946 functionality from ::get function.
6947 (get): Remove and inline into ::get and ::get_create.
6948 (get_create): Move code from ::get function.
6950 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6953 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
6954 -x assembler-with-cpp.
6956 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6958 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
6959 _FILE_OFFSET_BITS=64 for C++.
6961 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
6963 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
6964 conversion insn that shows up when pr85657-3.c is compiled using
6965 IEEE 128-bit long double.
6966 (neg<mode>2_internal): Use the correct mode to check whether the
6967 mode is IBM extended.
6968 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
6969 multiply and divide external functions from being created more
6972 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6974 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
6976 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
6977 the edge can be forwarded.
6978 (cfg_layout_merge_blocks): Likewise.
6980 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6982 * except.c (finish_eh_generation): Commit edge insertions only after
6983 the EH edges have been redirected from post-landing to landing pads.
6985 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6987 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
6988 create_tmp_var_for to create the FRAME decl.
6989 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
6991 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6993 * tree-inline.c (copy_edges_for_bb): Minor tweak.
6994 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
6995 debug statement when resetting its value.
6996 (expand_call_inline): Copy the locus of the call onto the assignment
6997 of the return value, if any. Use local variable in more cases.
6999 2018-06-21 Martin Liska <mliska@suse.cz>
7001 * ipa-pure-const.c (propagate_nothrow): Use
7002 funct_state_summaries->get.
7003 (dump_malloc_lattice): Likewise.
7004 (propagate_malloc): Likewise.
7006 2018-06-21 Richard Biener <rguenther@suse.de>
7008 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
7009 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
7010 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
7012 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
7013 BLOCK_ABSTRACT_ORIGIN unconditionally.
7015 2018-06-21 David Malcolm <dmalcolm@redhat.com>
7017 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
7019 * ipa-reference.c (ipa_reference_c_finalize): Delete
7020 ipa_ref_opt_sum_summaries and set it to NULL.
7022 2018-06-21 Tom de Vries <tdevries@suse.de>
7024 PR tree-optimization/85859
7025 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
7026 test with comment from bb_no_side_effects_p.
7028 2018-06-21 Richard Biener <rguenther@suse.de>
7030 PR tree-optimization/86232
7031 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
7032 max for constant niter.
7034 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7036 * config/aarch64/aarch64-simd.md
7037 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
7039 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7041 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
7042 Make opernads of the unspec commutative.
7044 2018-06-21 Richard Biener <rguenther@suse.de>
7046 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
7047 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
7048 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
7049 (vect_analyze_data_ref_dependence): Re-order checks to deal with
7051 (vect_record_base_alignments): Do not record base alignment
7052 for gathers or scatters.
7053 (vect_compute_data_ref_alignment): Drop return value that is always
7054 true. Bail out early for gathers or scatters.
7055 (vect_enhance_data_refs_alignment): Bail out early for gathers
7057 (vect_find_same_alignment_drs): Likewise.
7058 (vect_analyze_data_refs_alignment): Remove dead code.
7059 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7060 (vect_analyze_data_refs): For possible gathers or scatters do
7061 not create an alternate DR, just check their possible validity
7062 and mark them. Adjust DECL_NONALIASED handling to not rely
7064 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
7065 update inits of gathers or scatters.
7066 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
7067 Also copy gather/scatter flag to pattern vinfo.
7069 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
7071 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7072 behavior of vec_packsu (vector unsigned long long, vector unsigned
7073 long long) to match behavior of vec_packs with same signature.
7075 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
7076 Thomas Schwinge <thomas@codesourcery.com>
7077 Cesar Philippidis <cesar@codesourcery.com>
7079 * gimplify.c (gimplify_scan_omp_clauses): Add support for
7080 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7081 (gimplify_adjust_omp_clauses): Likewise.
7082 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
7083 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
7084 (gimplify_omp_target_update): Update handling of acc update and
7086 * omp-low.c (install_var_field): Remove unused parameter
7087 base_pointers_restrict.
7088 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
7089 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
7091 (omp_target_base_pointers_restrict_p): Delete.
7092 (scan_omp_target): Update call to scan_sharing_clauses.
7093 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
7095 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
7096 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7097 (convert_local_omp_clauses): Likewise.
7098 * tree-pretty-print.c (dump_omp_clause): Likewise.
7099 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
7101 (omp_clause_code_name): Likewise.
7103 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7106 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
7109 PR tree-optimization/86231
7110 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
7111 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7113 2018-06-20 Tom de Vries <tdevries@suse.de>
7115 PR tree-optimization/86097
7116 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
7117 iv type if signedness of iv type is not the same as that of *nit.
7119 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7121 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
7122 EDGE_EH edges, verify they are all EDGE_EH.
7124 2018-06-20 Maya Rashish <coypu@sdf.org>
7126 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
7128 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7130 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
7131 * config/aarch64/aarch64.c (xgene1_tunings): Add
7132 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
7133 (aarch64_mode_valid_for_sched_fusion_p):
7134 Allow 16-byte modes.
7135 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
7136 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
7138 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
7140 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
7141 * config/aarch64/iterators.md (VQ2): New mode iterator.
7143 2018-06-20 Martin Liska <mliska@suse.cz>
7145 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7146 Change default ratio from 10 to 8.
7148 2018-06-20 Martin Liska <mliska@suse.cz>
7150 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7152 (bit_test_cluster::find_bit_tests): Likewise.
7153 (switch_decision_tree::analyze_switch_statement): Find clusters.
7154 * tree-switch-conversion.h (struct jump_table_cluster): Document
7157 2018-06-20 Martin Liska <mliska@suse.cz>
7159 * tree-switch-conversion.c (switch_conversion::collect):
7160 Record m_uniq property.
7161 (switch_conversion::expand): Bail out for special conditions.
7162 (group_cluster::~group_cluster): New.
7163 (group_cluster::group_cluster): Likewise.
7164 (group_cluster::dump): Likewise.
7165 (jump_table_cluster::emit): New.
7166 (switch_decision_tree::fix_phi_operands_for_edges): New.
7167 (struct case_node): Remove struct.
7168 (jump_table_cluster::can_be_handled): New.
7169 (case_values_threshold): Moved to header.
7170 (reset_out_edges_aux): Likewise.
7171 (jump_table_cluster::is_beneficial): New.
7172 (bit_test_cluster::can_be_handled): Likewise.
7173 (add_case_node): Remove.
7174 (bit_test_cluster::is_beneficial): New.
7175 (case_bit_test::cmp): New.
7176 (bit_test_cluster::emit): New.
7177 (expand_switch_as_decision_tree_p): Remove.
7178 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
7179 (fix_phi_operands_for_edge): Likewise.
7180 (switch_decision_tree::analyze_switch_statement): New.
7181 (compute_cases_per_edge): Move ...
7182 (switch_decision_tree::compute_cases_per_edge): ... here.
7183 (try_switch_expansion): Likewise.
7184 (switch_decision_tree::try_switch_expansion): Likewise.
7185 (record_phi_operand_mapping): Likewise.
7186 (switch_decision_tree::record_phi_operand_mapping): Likewise.
7187 (emit_case_decision_tree): Likewise.
7188 (switch_decision_tree::emit): Likewise.
7189 (balance_case_nodes): Likewise.
7190 (switch_decision_tree::balance_case_nodes): Likewise.
7191 (dump_case_nodes): Likewise.
7192 (switch_decision_tree::dump_case_nodes): Likewise.
7193 (emit_jump): Likewise.
7194 (switch_decision_tree::emit_jump): Likewise.
7195 (emit_cmp_and_jump_insns): Likewise.
7196 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7197 (emit_case_nodes): Likewise.
7198 (switch_decision_tree::emit_case_nodes): Likewise.
7199 (conditional_probability): Remove.
7200 * tree-switch-conversion.h (enum cluster_type): New.
7202 (struct cluster): Likewise.
7203 (cluster::cluster): Likewise.
7204 (struct simple_cluster): Likewise.
7205 (simple_cluster::simple_cluster): Likewise.
7206 (struct group_cluster): Likewise.
7207 (struct jump_table_cluster): Likewise.
7208 (struct bit_test_cluster): Likewise.
7209 (struct min_cluster_item): Likewise.
7210 (struct case_tree_node): Likewise.
7211 (case_tree_node::case_tree_node): Likewise.
7212 (jump_table_cluster::case_values_threshold): Likewise.
7213 (struct case_bit_test): Likewise.
7214 (struct switch_decision_tree): Likewise.
7215 (struct switch_conversion): Likewise.
7216 (switch_decision_tree::reset_out_edges_aux): Likewise.
7218 2018-06-20 Martin Liska <mliska@suse.cz>
7220 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7221 (hoist_edge_and_branch_if_true): Likewise.
7222 (expand_switch_using_bit_tests_p): Likewise.
7223 (struct case_bit_test): Likewise.
7224 (case_bit_test_cmp): Likewise.
7225 (emit_case_bit_tests): Likewise.
7226 (switch_conversion::switch_conversion): New class.
7227 (struct switch_conv_info): Remove old struct.
7228 (collect_switch_conv_info): More to ...
7229 (switch_conversion::collect): ... this.
7230 (check_range): Likewise.
7231 (switch_conversion::check_range): Likewise.
7232 (check_all_empty_except_final): Likewise.
7233 (switch_conversion::check_all_empty_except_final): Likewise.
7234 (check_final_bb): Likewise.
7235 (switch_conversion::check_final_bb): Likewise.
7236 (create_temp_arrays): Likewise.
7237 (switch_conversion::create_temp_arrays): Likewise.
7238 (free_temp_arrays): Likewise.
7239 (gather_default_values): Likewise.
7240 (switch_conversion::gather_default_values): Likewise.
7241 (build_constructors): Likewise.
7242 (switch_conversion::build_constructors): Likewise.
7243 (constructor_contains_same_values_p): Likewise.
7244 (switch_conversion::contains_same_values_p): Likewise.
7245 (array_value_type): Likewise.
7246 (switch_conversion::array_value_type): Likewise.
7247 (build_one_array): Likewise.
7248 (switch_conversion::build_one_array): Likewise.
7249 (build_arrays): Likewise.
7250 (switch_conversion::build_arrays): Likewise.
7251 (gen_def_assigns): Likewise.
7252 (switch_conversion::gen_def_assigns): Likewise.
7253 (prune_bbs): Likewise.
7254 (switch_conversion::prune_bbs): Likewise.
7255 (fix_phi_nodes): Likewise.
7256 (switch_conversion::fix_phi_nodes): Likewise.
7257 (gen_inbound_check): Likewise.
7258 (switch_conversion::gen_inbound_check): Likewise.
7259 (process_switch): Use the newly created class.
7260 (switch_conversion::expand): New.
7261 (switch_conversion::~switch_conversion): New.
7262 * tree-switch-conversion.h: New file.
7264 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7266 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7267 tree-vect-patterns.c.
7268 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7269 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7270 (vect_recog_sad_pattern): Likewise.
7271 (vect_recog_widen_sum_pattern): Likewise.
7272 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7273 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7274 (vect_recog_rotate_pattern): Likewise.
7275 (vect_recog_mult_pattern): Likewise.
7276 (vect_recog_vector_vector_shift_pattern): Likewise.
7277 (vect_recog_divmod_pattern): Likewise.
7278 (vect_recog_mixed_size_cond_pattern): Likewise.
7279 (vect_recog_bool_pattern): Likewise.
7280 (vect_recog_mask_conversion_pattern): Likewise.
7281 (vect_try_gather_scatter_pattern): Likewise.
7282 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7283 (vect_recog_over_widening_pattern): Likewise.
7284 (vect_recog_gather_scatter_pattern): Likewise.
7285 (vect_recog_func_ptr): Move from tree-vectorizer.h
7286 (vect_vect_recog_func_ptrs): Move further down the file.
7287 (vect_recog_func): Likewise. Remove the third argument.
7288 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7289 (vect_pattern_recog_1): Expect the pattern function to do any
7290 necessary target tests. Also expect it to provide a vector type.
7291 Remove the type_in handling.
7293 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7295 * tree-vect-patterns.c (vect_pattern_detected): New function.
7296 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7297 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7298 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7299 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7300 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7301 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7302 (vect_recog_mask_conversion_pattern)
7303 (vect_try_gather_scatter_pattern): Likewise.
7305 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7307 * tree-vect-patterns.c (vect_get_internal_def): New function.
7308 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7309 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7310 (search_type_for_mask_1): Use it.
7312 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7314 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7315 redundant WIDEN_SUM_EXPR handling.
7316 (vect_recog_sad_pattern): Likewise.
7318 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7320 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7321 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7322 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7323 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7324 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7326 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7328 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7329 use the stmt_vec_info of the original bb statement for the
7330 new zero assignment, even if the call is part of a pattern.
7332 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7334 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7335 that the sequence is attached to the original statement rather
7336 than the pattern statement.
7337 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7338 PATTERN_DEF_SEQ from the original statement rather than
7339 the main pattern statement.
7340 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7341 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7342 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7344 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7346 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7347 definition statements before the early exit for statements that aren't
7349 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7351 (vect_transform_loop): ...here. Process pattern definition
7352 statements without first checking whether the main pattern
7353 statement is live or relevant.
7355 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7357 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7358 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7360 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7362 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7363 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7364 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7365 (expand_block_compare): Change select_block_compare_mode call.
7366 (expand_strncmp_align_check): Use new functions, fix comment.
7367 (emit_final_str_compare_gpr): New function.
7368 (expand_strn_compare): Refactor and clean up code.
7369 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7371 2018-06-19 Tony Reix <tony.reix@atos.com>
7372 Damien Bergamini <damien.bergamini@atos.com>
7373 David Edelsohn <dje.gcc@gmail.com>
7375 * collect2.c (static_obj): New variable.
7376 (static_libs): New variable.
7377 (is_in_list): Uncomment declaration.
7378 (main): Track AIX libraries linked statically.
7379 (is_in_list): Uncomment definition.
7380 (scan_prog_file): Don't add AIX shared libraries initializer
7381 to constructor list if linking statically.
7383 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
7385 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
7387 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
7389 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
7391 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
7394 2018-06-19 Martin Liska <mliska@suse.cz>
7396 * config/i386/i386.c (ix86_can_inline_p): Do not use
7397 ipa_fn_summaries::get_create.
7398 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
7400 (devirtualization_time_bonus): Likewise.
7401 (ipcp_propagate_stage): Likewise.
7402 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7403 (edge_set_predicate): Likewise.
7404 (evaluate_conditions_for_known_args): Likewise.
7405 (evaluate_properties_for_edge): Likewise.
7406 (ipa_call_summary::reset): Tranform to ...
7407 (ipa_call_summary::~ipa_call_summary): ... this.
7408 (ipa_fn_summary::reset): Transform to ...
7409 (ipa_fn_summary::~ipa_fn_summary): ... this.
7410 (ipa_fn_summary_t::remove): Rename to ...
7411 (ipa_fn_summary_t::remove_callees): ... this.
7412 (ipa_fn_summary_t::duplicate): Use placement new
7413 instead of memory copy.
7414 (ipa_call_summary_t::duplicate): Likewise.
7415 (ipa_call_summary_t::remove): Remove.
7416 (dump_ipa_call_summary): Change get_create to get.
7417 (ipa_dump_fn_summary): Dump only when summary exists.
7418 (analyze_function_body): Use symbol_summary::get instead
7420 (compute_fn_summary): Likewise.
7421 (estimate_edge_devirt_benefit): Likewise.
7422 (estimate_edge_size_and_time): Likewise.
7423 (inline_update_callee_summaries): Likewise.
7424 (remap_edge_change_prob): Likewise.
7425 (remap_edge_summaries): Likewise.
7426 (ipa_merge_fn_summary_after_inlining): Likewise.
7427 (write_ipa_call_summary): Likewise.
7428 (ipa_fn_summary_write): Likewise.
7429 (ipa_free_fn_summary): Likewise.
7430 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
7431 (struct ipa_call_summary): Likewise.
7432 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
7434 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
7435 (estimate_size_after_inlining): Likewise.
7436 (estimate_growth): Likewise.
7437 (growth_likely_positive): Likewise.
7438 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7439 (inline_call): Likewise.
7440 * ipa-inline.c (caller_growth_limits): Likewise.
7441 (can_inline_edge_p): Likewise.
7442 (can_inline_edge_by_limits_p): Likewise.
7443 (compute_uninlined_call_time): Likewise.
7444 (compute_inlined_call_time): Likewise.
7445 (want_inline_small_function_p): Likewise.
7446 (edge_badness): Likewise.
7447 (update_caller_keys): Likewise.
7448 (update_callee_keys): Likewise.
7449 (inline_small_functions): Likewise.
7450 (inline_to_all_callers_1): Likewise.
7451 (dump_overall_stats): Likewise.
7452 (early_inline_small_functions): Likewise.
7453 (early_inliner): Likewise.
7454 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7455 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7456 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7457 * ipa-split.c (execute_split_functions): Likewise.
7458 * symbol-summary.h: Likewise.
7459 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7461 2018-06-19 Richard Biener <rguenther@suse.de>
7463 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
7464 (vectorize_loops): ... here. Fix dbgcnt handling.
7465 (try_vectorize_loop): Wrap try_vectorize_loop_1.
7467 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
7470 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
7471 ieee128 argument takes up only one (vector) register, not two (floating
7474 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7476 * gimplify.c (gimplify_init_constructor): Really never clear for an
7477 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
7479 2018-06-19 Richard Biener <rguenther@suse.de>
7481 PR tree-optimization/86179
7482 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
7483 after failed recognition.
7485 2018-06-18 Martin Sebor <msebor@redhat.com>
7488 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
7489 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
7490 Handle integer subtraction.
7491 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
7492 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
7494 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7496 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
7497 param from rtx to rtx_insn *.
7498 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
7500 (frv_ifcvt_modify_insn): Likwise.
7501 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
7502 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
7503 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
7504 as_a <rtx_insn *> cast to local "unprotected_region" once
7505 it's been established that it's not NULL or pc_rtx.
7506 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
7507 param "sethi" from rtx to rtx_insn *.
7508 (nds32_group_float_insns): Likewise for param "insn".
7509 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
7511 (vax_output_int_subtract): Likewise.
7512 * config/vax/vax.c (vax_output_int_add): Likewise for param
7514 (vax_output_int_subtract): Likewise.
7515 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
7516 (emit_pattern_after): Likewise for param "after".
7517 (emit_insn_after): Likewise.
7518 (emit_jump_insn_after): Likewise.
7519 (emit_call_insn_after): Likewise.
7520 (emit_debug_insn_after): Likewise.
7521 (emit_pattern_before): Likewise for param "before".
7522 (emit_insn_before): Likewise.
7523 (emit_jump_insn_before): Likewise.
7524 * final.c (get_insn_template): Likewise for param "insn", removing
7526 * output.h (get_insn_template): Likewise for 2nd param.
7527 * rtl.h (emit_insn_before): Likewise.
7528 (emit_jump_insn_before): Likewise.
7529 (emit_debug_insn_before_noloc): Likewise.
7530 (emit_insn_after): Likewise.
7531 (emit_jump_insn_after): Likewise.
7532 (emit_call_insn_after): Likewise.
7533 (emit_debug_insn_after): Likewise.
7534 (set_insn_deleted): Likewise for param.
7536 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
7539 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
7540 floating point modes, so that IFmode is numerically greater than
7541 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
7542 to declare the ordering. This prevents IFmode from being
7543 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
7544 machine. Include rs6000-modes.h to share the fractional values
7545 between genmodes* and the rest of the compiler.
7549 * config/rs6000/rs6000-modes.h: New file.
7550 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
7551 meaning of rs6000_long_double_size so that 126..128 selects an
7552 appropriate 128-bit floating point type.
7553 (rs6000_option_override_internal): Likewise.
7554 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
7555 (TARGET_LONG_DOUBLE_128): Change the meaning of
7556 rs6000_long_double_size so that 126..128 selects an appropriate
7557 128-bit floating point type.
7558 (LONG_DOUBLE_TYPE_SIZE): Update comment.
7559 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
7560 source and destination to match the standard usage.
7561 (truncifkf2): Likewise.
7562 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
7563 ISA 2.07 to use an explicit clobber, instead of passing in a
7565 (copysign<mode>3_soft): Likewise.
7567 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7569 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7570 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
7571 (vect_slp_analyze_instance_dependence): Likewise.
7572 (vect_enhance_data_refs_alignment): Likewise.
7573 (vect_analyze_data_refs_alignment): Likewise.
7574 (vect_slp_analyze_and_verify_instance_alignment
7575 (vect_analyze_data_ref_accesses): Likewise.
7576 (vect_prune_runtime_alias_test_list): Likewise.
7577 (vect_analyze_data_refs): Likewise.
7578 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
7579 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7580 (vect_analyze_scalar_cycles_1): Likewise.
7581 (vect_get_loop_niters): Likewise.
7582 (vect_analyze_loop_form_1): Likewise.
7583 (vect_update_vf_for_slp): Likewise.
7584 (vect_analyze_loop_operations): Likewise.
7585 (vect_analyze_loop): Likewise.
7586 (vectorizable_induction): Likewise.
7587 (vect_transform_loop): Likewise.
7588 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
7589 * tree-vect-slp.c (vect_analyze_slp): Likewise.
7590 (vect_make_slp_decision): Likewise.
7591 (vect_detect_hybrid_slp): Likewise.
7592 (vect_slp_analyze_operations): Likewise.
7593 (vect_slp_bb): Likewise.
7594 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
7595 (vectorizable_bswap): Likewise.
7596 (vectorizable_call): Likewise.
7597 (vectorizable_simd_clone_call): Likewise.
7598 (vectorizable_conversion): Likewise.
7599 (vectorizable_assignment): Likewise.
7600 (vectorizable_shift): Likewise.
7601 (vectorizable_operation): Likewise.
7602 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
7604 2018-06-18 Martin Sebor <msebor@redhat.com>
7606 PR tree-optimization/81384
7607 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
7608 * builtins.c (expand_builtin_strnlen): New function.
7609 (expand_builtin): Call it.
7610 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
7611 * builtins.def (BUILT_IN_STRNLEN): New.
7612 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
7613 Warn for bounds in excess of maximum object size.
7614 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
7615 single-value ranges. Handle strnlen.
7616 (handle_builtin_strlen): Handle strnlen.
7617 (strlen_check_and_optimize_stmt): Same.
7618 * doc/extend.texi (Other Builtins): Document strnlen.
7620 2018-06-18 Maya Rashish <coypu@sdf.org>
7622 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
7623 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
7624 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
7626 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
7628 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
7630 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7632 * tree.c (escaped_string::escape): Replace cast to char * by
7633 const_cast<char *> (unescaped).
7635 2018-06-18 Nick Clifton <nickc@redhat.com>
7638 * tree.c (escaped_string): New class. Converts an unescaped
7639 string into its escaped equivalent.
7640 (warn_deprecated_use): Use the new class to convert the
7641 deprecation message, if present.
7642 (test_escaped_strings): New self test.
7643 (test_c_tests): Add test_escaped_strings.
7644 * doc/extend.texi (deprecated): Add a note that the
7645 deprecation message is affected by the -fmessage-length
7646 option, and that control characters will be escaped.
7647 (#pragma GCC error): Document this pragma.
7648 (#pragma GCC warning): Likewise.
7649 * doc/invoke.texi (-fmessage-length): Document this option's
7650 effect on the #warning and #error preprocessor directives and
7651 the deprecated attribute.
7653 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
7655 * tree.c (decl_value_expr_lookup): Revert latest change.
7656 (decl_value_expr_insert): Likewise.
7658 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
7660 * gimplify.c (nonlocal_vlas): Delete.
7661 (nonlocal_vla_vars): Likewise.
7662 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
7664 (gimplify_body): Do not create and destroy nonlocal_vlas.
7665 * tree-nested.c: Include diagnostic.h.
7666 (use_pointer_in_frame): Tweak.
7667 (lookup_field_for_decl): Add assertion and declare the transformation.
7668 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
7669 internal error when the reference is in a wrong context. Do not
7670 create a debug decl by default.
7671 (note_nonlocal_block_vlas): Delete.
7672 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
7673 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
7674 create a debug decl by default.
7675 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
7676 call is in a wrong context.
7677 (fixup_vla_decls): New function.
7678 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
7679 debug variables were created.
7680 * tree.c (decl_value_expr_lookup): Add checking assertion.
7681 (decl_value_expr_insert): Likewise.
7683 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7686 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
7687 * tree-scalar-evolution.c (interpret_expr): Likewise.
7688 (expression_expensive_p): Likewise.
7689 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
7690 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
7691 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
7692 (ssa_defined_by_minus_one_stmt_p): New.
7694 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7697 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
7698 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
7699 * dojump.c (do_jump): Likewise.
7700 * expr.c (expand_expr_real_2): Check operand type's sign.
7701 * fold-const.c (const_unop): Handle ABSU_EXPR.
7702 (fold_abs_const): Likewise.
7703 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
7704 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
7705 (strip_sign_op_1): Likesise.
7706 * match.pd: Add new pattern to generate ABSU_EXPR.
7707 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
7708 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7709 * tree-eh.c (operation_could_trap_helper_p): Likewise.
7710 * tree-inline.c (estimate_operator_cost): Likewise.
7711 * tree-pretty-print.c (dump_generic_node): Likewise.
7712 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
7713 * tree.def (ABSU_EXPR): New.
7715 2018-06-16 Jakub Jelinek <jakub@redhat.com>
7718 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
7719 documented as preserved for backward compatibility only.
7720 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
7722 PR rtl-optimization/86108
7723 * bb-reorder.c (create_forwarder_block): Renamed to ...
7724 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
7725 jump from new landing pad to the second part.
7726 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
7729 2018-06-15 Jakub Jelinek <jakub@redhat.com>
7732 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
7733 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
7734 Only call store_expr for halves if the mode is the same.
7737 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
7738 Fix up comment formatting.
7740 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
7742 * typed-splay-tree.h (typed_splay_tree::remove): New function.
7743 (typed_splay_tree::closure,
7744 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
7745 (typed_splay_tree::typed_splay_tree,
7746 typed_splay_tree::operator =): Declared private.
7747 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
7748 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
7749 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
7750 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
7751 typed_splay_tree::splay_tree_splay,
7752 typed_splay_tree::splay_tree_foreach_helper,
7753 typed_splay_tree::splay_tree_insert,
7754 typed_splay_tree::splay_tree_remove,
7755 typed_splay_tree::splay_tree_lookup,
7756 typed_splay_tree::splay_tree_predecessor,
7757 typed_splay_tree::splay_tree_successor,
7758 typed_splay_tree::splay_tree_min,
7759 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
7760 (typed_splay_tree::root, typed_splay_tree::comp,
7761 typed_splay_tree::delete_key,
7762 typed_splay_tree::delete_value): New data members.
7763 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
7764 typed_splay_tree::remove.
7766 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
7768 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
7769 -mginv and -mno-ginv to the assembler.
7770 * config/mips/mips.opt (-mcrc): New option.
7772 * doc/invoke.text (-mcrc): Document.
7775 2018-06-15 Nick Clifton <nickc@redhat.com>
7778 * tree.c (escaped_string): New class. Converts an unescaped
7779 string into its escaped equivalent.
7780 (warn_deprecated_use): Use the new class to convert the
7781 deprecation message, if present.
7782 (test_escaped_strings): New self test.
7783 (test_c_tests): Add test_escaped_strings.
7784 * doc/extend.texi (deprecated): Add a note that the
7785 deprecation message is affected by the -fmessage-length
7786 option, and that control characters will be escaped.
7787 (#pragma GCC error): Document this pragma.
7788 (#pragma GCC warning): Likewise.
7789 * doc/invoke.texi (-fmessage-length): Document this option's
7790 effect on the #warning and #error preprocessor directives and
7791 the deprecated attribute.
7793 2018-06-15 Richard Biener <rguenther@suse.de>
7795 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
7796 here, also noting vector size used.
7797 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
7798 size used in MSG_OPTIMIZED_LOCATIONS dump.
7799 (pass_slp_vectorize::execute): Adjust.
7801 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
7804 * config/arc/arc.c (arc_return_address_register): Fix
7807 2018-06-15 Richard Biener <rguenther@suse.de>
7810 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
7811 leave useless conversion stripping to force_gimple_operand_gsi.
7812 (gimplify_build2): Likewise.
7813 (gimplify_build1): Likewise.
7815 2018-06-15 Richard Biener <rguenther@suse.de>
7818 * tree-cfg.c (move_stmt_op): unshare invariant addresses
7819 before adjusting their block.
7821 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
7823 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
7824 multilibs for *-*-rtems*.
7825 * config/riscv/t-rtems: New file.
7827 2018-06-14 Jakub Jelinek <jakub@redhat.com>
7830 * match.pd ((A +- CST1) +- CST2): Punt if last resort
7831 unsigned_type_for returns NULL.
7834 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
7835 subregs of multi-word pseudos unless the float mode has word size.
7837 2018-06-14 Richard Biener <rguenther@suse.de>
7840 * tree-vect-generic.c (build_word_mode_vector_type): Remove
7841 duplicate and harmful type_hash_canon.
7842 * tree.c (type_hash_canon): Assert we didn't find ourselves.
7844 2018-06-14 Richard Biener <rguenther@suse.de>
7847 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
7850 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
7852 * config/rtems.h (STDINT_LONG32): Define.
7854 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
7855 Prachi Godbole <prachi.godbole@imgtec.com>
7857 * config/mips/mips-cpus.def: Define P6600.
7858 * config/mips/mips-tables.opt: Regenerate.
7859 * config/mips/mips.c (mips_ucbranch_type): New enum.
7860 (mips_rtx_cost_data): Add support for P6600.
7861 (mips_issue_rate): Likewise.
7862 (mips_multipass_dfa_lookahead): Likewise.
7863 (mips_avoid_hazard): Likewise.
7864 (mips_reorg_process_insns): Likewise.
7865 (mips_classify_branch_p6600): New function.
7866 * config/mips/mips.h (TUNE_P6600): New define.
7867 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
7868 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
7869 * config/mips/mips.md: Include p6600.md.
7870 (processor): Add p6600.
7871 * config/mips/p6600.md: New file.
7872 * doc/invoke.texi: Add p6600 to supported architectures.
7874 2018-06-13 Martin Sebor <msebor@redhat.com>
7876 PR tree-optimization/86114
7877 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
7879 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
7881 2018-06-13 Richard Biener <rguenther@suse.de>
7883 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
7884 Properly set vector type of the intermediate stmt.
7885 * tree-vect-stmts.c (vectorizable_operation): The destination
7886 var always has vectype_out type.
7888 2018-06-13 Jeff Law <law@redhat.com>
7890 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
7891 integer 0 for argument to print_rtl_with_bb.
7892 (rl78_reorg): Likewise.
7894 2018-06-13 David Malcolm <dmalcolm@redhat.com>
7896 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
7897 from rtx to rtx_insn *.
7898 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
7900 (add_sched_insns_for_speculation): Likewise for local "target",
7901 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7902 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
7903 from rtx_insn ** to rtx_code_label **.
7904 (reorg_emit_nops): Likewise.
7905 (c6x_reorg): Likewise for local "call_labels".
7906 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
7908 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
7909 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
7910 the loops over LABEL_REFS.
7911 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
7913 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7914 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
7915 (split_branches): Strengthen local "olabel" from rtx to
7916 rtx_insn *, adding a safe_as_a cast.
7917 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
7919 (add_insn_after): Likewise for first two params.
7920 (add_insn_before): Likewise.
7921 (remove_insn): Likewise for param.
7922 (emit_pattern_before_noloc): Likewise for second and third params.
7923 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
7924 (emit_call_insn_before_noloc): Likewise.
7925 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
7927 (emit_barrier_before): Likewise.
7928 (emit_label_before): Strengthen "label" param from "rtx" to
7929 "rtx_code_label *". Strengthen "before" param from "rtx" to
7931 (emit_insn_after_1): Strengthen "after" param from "rtx" to
7933 (emit_pattern_after_noloc): Likewise.
7934 (emit_insn_after_noloc): Likewise.
7935 (emit_jump_insn_after_noloc): Likewise.
7936 (emit_call_insn_after_noloc): Likewise.
7937 (emit_debug_insn_after_noloc): Likewise.
7938 (emit_barrier_after): Likewise.
7939 (emit_label_after): Likewise for both params.
7940 (emit_pattern_after_setloc): Likewise for "after" param. Convert
7941 "loc" param from "int" to "location_t".
7942 (emit_insn_after_setloc): Likewise.
7943 (emit_jump_insn_after_setloc): Likewise.
7944 (emit_call_insn_after_setloc): Likewise.
7945 (emit_debug_insn_after_setloc): Likewise.
7946 (emit_pattern_before_setloc): Likewise for "before" param. Convert
7947 "loc" param from "int" to "location_t".
7948 (emit_pattern_before): Convert NULL_RTX to NULL.
7949 (emit_insn_before_setloc): Convert "loc" param from "int" to
7951 (emit_jump_insn_before_setloc): Likewise.
7952 (emit_call_insn_before_setloc): Likewise.
7953 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
7954 rtx_insn *. Convert "loc" param from "int" to "location_t".
7955 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
7956 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
7957 Convert 3rd param from "int" to "location_t".
7958 (emit_barrier_before, emit_barrier_after, next_real_insn):
7959 Strengthen param from rtx to rtx_insn *.
7960 (emit_label_before): Strengthen 1st param from "rtx" to
7961 "rtx_code_label *". Strengthen 2nd param from "rtx" to
7963 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
7964 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
7965 Strengthen 2nd param from "rtx" to "rtx_insn *".
7966 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
7967 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
7968 Likewise. Convert 3rd param from "int" to "location_t".
7969 (emit_label_after): Strengthen 1st param from "rtx" to
7971 (next_real_insn, remove_insn): Strengthen param from "rtx" to
7973 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
7974 from "rtx" to "rtx_insn *".
7976 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
7978 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
7979 bodies streamed in with -Q.
7980 * dumpfile.c (dump_files): Add lto-stream-out dump file.
7981 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7982 * gimple-streamer-out.c: Include gimple-pretty-print.h
7983 (output_bb): Dump stmts streamed.
7984 * lto-section-out.c: Include print-tree.h
7985 (lto_begin_section): Dump sections created.
7986 (lto_output_decl_index): Dump decl encoded.
7987 * lto-streamer-out.c: Include print-tree.h
7988 (create_output_block): Dump output block created.
7989 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
7990 (output_function): Dump function output.
7991 (output_constructor): Dump constructor streamed.
7992 (write_global_stream): Output indexes encoded.
7993 (produce_asm_for_decls): Dump streams encoded.
7994 * lto-streamer.c (streamer_dump_file): New global var.
7995 * lto-streamer.h (streamer_dump_file): Declare.
7996 * passes.c (ipa_write_summaries): Initialize streamer dump.
7997 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
8000 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
8003 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
8004 offsets for register save directives. Emit a second batch of save
8005 directives, if need be, when the function accesses prior frames.
8007 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8009 * config/arc/fpu.md (fmasf4): Force operand to register.
8010 (fnmasf4): Likewise.
8012 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8014 * config/arc/arc-protos.h (arc_pad_return): Remove.
8015 * config/arc/arc.c (machine_function): Remove force_short_suffix
8017 (arc_print_operand): Adjust printing of '&'.
8018 (arc_verify_short): Remove conditional printing of short suffix.
8019 (arc_final_prescan_insn): Remove reference to size_reason.
8020 (pad_return): New function.
8021 (arc_reorg): Call pad_return.
8022 (arc_pad_return): Remove.
8023 (arc_init_machine_status): Remove reference to force_short_suffix.
8024 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
8025 (attr length): When attribute iscompact is true force to 2
8026 regardless; in the case of maybe check if we want to force the
8027 instruction to have 4 bytes length.
8028 (nopv): Change it to generate 4 byte long nop as well.
8029 (blockage): New pattern.
8030 (simple_return): Remove call to arc_pad_return.
8031 (p_return_i): Likewise.
8033 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8035 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
8037 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8039 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
8042 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8044 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
8045 for ARC700 and ARCv2.
8047 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
8050 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
8051 operands[2] instead of operands[1].
8054 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8056 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
8057 case, check whether the outer register overlaps an unallocatable
8058 register, not just whether it fits the required class.
8060 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8062 * poly-int.h (can_div_trunc_p): Add new overload in which all values
8064 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
8065 (memrefs_conflict_p): Likewise.
8066 (init_alias_analysis): Likewise.
8067 * cfgexpand.c (expand_debug_expr): Likewise.
8068 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
8069 * cse.c (fold_rtx): Likewise.
8070 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
8071 * expr.c (emit_block_move_hints): Likewise.
8072 (clear_storage_hints, push_block, emit_push_insn): Likewise.
8073 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
8074 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
8075 (emit_group_store): Likewise.
8076 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
8077 to read the PRE/POST_MODIFY increment.
8078 * calls.c (store_one_arg): Use strip_offset.
8079 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
8081 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
8083 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
8084 (simplify_binary_operation_1): Extend CONST_INT handling to
8086 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
8087 than a HOST_WIDE_INT.
8088 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
8090 (adjust_mems, add_stores): Update accodingly.
8091 (vt_canonicalize_addr): Track polynomial offsets.
8092 (emit_note_insn_var_location): Likewise.
8093 (vt_add_function_parameter): Likewise.
8094 (vt_initialize): Likewise.
8096 2018-06-12 Jeff Law <law@redhat.com>
8098 * config.gcc (alpha*-*-freebsd*): Remove.
8099 * config/alpha/freebsd.h: Remove.
8101 2018-06-12 David Malcolm <dmalcolm@redhat.com>
8104 * spellcheck-tree.c (levenshtein_distance): Rename to...
8105 (get_edit_distance): ...this, and update for underlying renaming.
8106 * spellcheck-tree.h (levenshtein_distance): Rename to...
8107 (get_edit_distance): ...this.
8108 * spellcheck.c (levenshtein_distance): Rename to...
8109 (get_edit_distance): ...this. Convert from Levenshtein distance
8110 to Damerau-Levenshtein distance by supporting transpositions of
8111 adjacent characters. Rename "v1" to "v_next" and "v0" to
8113 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
8114 (selftest::test_edit_distance_unit_test_oneway): ...this, and
8115 update for underlying renaming.
8116 (selftest::levenshtein_distance_unit_test): Rename to...
8117 (selftest::test_get_edit_distance_unit): ...this, and update for
8118 underlying renaming.
8119 (selftest::test_find_closest_string): Add example from PR 69968
8120 where transposition helps
8121 (selftest::test_metric_conditions): Update for renaming.
8122 (selftest::test_metric_conditions): Likewise.
8123 (selftest::spellcheck_c_tests): Likewise.
8124 * spellcheck.h (levenshtein_distance): Rename both overloads to...
8125 (get_edit_distance): ...this.
8126 (best_match::consider): Update for renaming.
8128 2018-06-12 Martin Sebor <msebor@redhat.com>
8130 PR tree-optimization/85259
8131 * builtins.c (compute_objsize): Handle constant offsets.
8132 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
8133 true iff a warning has been issued.
8134 * gimple.h (gimple_nonartificial_location): New function.
8135 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
8136 gimple_nonartificial_location and handle -Wno-system-headers.
8137 (handle_builtin_stxncpy): Same.
8139 2018-06-12 Martin Sebor <msebor@redhat.com>
8142 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
8144 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
8146 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8147 BUILTIN_VEC_XST entries for pointer to double and long long.
8149 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
8152 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8154 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8157 2018-06-12 Martin Liska <mliska@suse.cz>
8159 * doc/options.texi: Document IntegerRange.
8161 2018-06-12 Martin Liska <mliska@suse.cz>
8163 * config/i386/i386.opt: Make MPX-related options as Deprecated.
8164 * opt-functions.awk: Handle Deprecated flag.
8165 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
8167 (read_cmdline_option): Report warning for a deprecated option.
8168 * opts.h (struct cl_option): Add new field cl_deprecated.
8169 (CL_ERR_DEPRECATED): New.
8171 2018-06-12 Martin Liska <mliska@suse.cz>
8173 * doc/options.texi: Document Deprecated option flag.
8175 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8177 * config/arc/arc-arch.h (arc_extras): New enum.
8178 (arc_cpu_t):Add field extra.
8179 (arc_cpu_types): Consider the extras.
8180 * config/arc/arc-cpus.def: Add extras info.
8181 * config/arc/arc-opts.h (processor_type): Consider extra field.
8182 * config/arc/arc.c (arc_override_options): Handle extra field.
8184 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8186 * config/arc/arc-arch.h: Update ARC_OPTX macro.
8187 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8189 * config/arc/arc.c (arc_init): Update pic warning.
8190 (irq_range): Update irq range parsing warnings.
8191 (arc_override_options): Update various warning messages.
8192 (arc_handle_aux_attribute): Likewise.
8194 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8196 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8198 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8200 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8201 compiler messages directives.
8203 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8205 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8206 * config/mips/mips-tables.opt: Regenerate.
8207 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8209 * doc/invoke.texi: Document -march=i6500.
8211 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8213 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8214 (i6400_gpmul): Add cpu_unit.
8215 (i6400_gpdiv): Likewise.
8216 (i6400_msa_add_d): Update reservations.
8217 (i6400_msa_int_add) Likewise.
8218 (i6400_msa_short_logic3) Likewise.
8219 (i6400_msa_short_logic2) Likewise.
8220 (i6400_msa_short_logic) Likewise.
8221 (i6400_msa_move) Likewise.
8222 (i6400_msa_cmp) Likewise.
8223 (i6400_msa_short_float2) Likewise.
8224 (i6400_msa_div_d) Likewise.
8225 (i6400_msa_long_logic1) Likewise.
8226 (i6400_msa_long_logic2) Likewise.
8227 (i6400_msa_mult) Likewise.
8228 (i6400_msa_long_float2) Likewise.
8229 (i6400_msa_long_float4) Likewise.
8230 (i6400_msa_long_float5) Likewise.
8231 (i6400_msa_long_float8) Likewise.
8232 (i6400_fpu_fadd): Include frint type.
8233 (i6400_fpu_store): New define_insn_reservation.
8234 (i6400_fpu_load): Likewise.
8235 (i6400_fpu_move): Likewise.
8236 (i6400_fpu_fcmp): Likewise.
8237 (i6400_fpu_fmadd): Likewise.
8238 (i6400_int_mult): Include imul3nc type and update reservation.
8239 (i6400_int_div): Include idiv3 type and update reservation.
8240 (i6400_int_load): Update to check type not move_type.
8241 (i6400_int_store): Likewise.
8242 (i6400_int_prefetch): Set zero latency.
8244 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8246 * gcc.c: Document new %@{...} sequence.
8247 (LINK_COMMAND_SPEC): Use it for the -L switches.
8248 (cpp_unique_options): Use it for the -I switches.
8249 (at_file_argbuf): New global variable.
8250 (in_at_file): Likewise.
8251 (alloc_args): Create at_file_argbuf.
8252 (clear_args): Truncate at_file_argbuf.
8253 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8254 (open_at_file): New function.
8255 (close_at_file): Likewise.
8256 (create_at_file): Delete.
8257 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8260 (validate_switches_from_spec): Deal with %@{...} sequence.
8261 (validate_switches): Likewise.
8262 (driver::finalize): Call clear_args.
8264 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8266 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8268 2018-06-11 Martin Sebor <msebor@redhat.com>
8270 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8271 (Wno-class-memaccess): Add @opindex.
8272 (Wno-templates, Wno-multiple-inheritance): Same.
8273 (Wno-virtual-inheritance, Wno-namespaces): Same.
8274 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8275 (Wno-format-overflow, Wno-format-truncation): Same.
8276 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8277 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8278 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8279 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8280 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8281 (Wno-misspelled-isr): Same.
8283 2018-06-11 Martin Sebor <msebor@redhat.com>
8285 * PR tree-optimization/86083
8286 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8288 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8290 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8292 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8295 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8296 on the correct operand.
8297 (*movdi_internal64): Ditto.
8299 2018-06-11 Martin Liska <mliska@suse.cz>
8301 PR tree-optimization/86089
8302 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8304 2018-06-11 Julia Koval <julia.koval@intel.com>
8306 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8307 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8308 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8310 2018-06-11 Olivier Hainque <hainque@adacore.com>
8312 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8313 for Ada with strict dwarf2.
8315 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8318 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8321 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8323 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8325 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8327 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8330 2018-06-08 Martin Liska <mliska@suse.cz>
8332 * tree-cfg.h (debug_function): Fix argument type to match
8335 2018-06-08 Martin Liska <mliska@suse.cz>
8337 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8338 Remove usage of MPX-related (and removed) fields.
8339 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8341 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8343 * cfg.c (debug): Use TDF_NONE rather than 0.
8344 * cfghooks.c (debug): Likewise.
8345 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8346 (struct dump_option_value_info): Convert to...
8347 (struct kv_pair): ...this template type.
8348 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8350 (optinfo_verbosity_options): Likewise.
8351 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8353 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8354 than int for "optgroup_flags" param.
8355 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8357 (dump_dec): Likewise.
8358 (dump_finish): Use TDF_NONE rather than 0.
8359 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8360 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8361 than 0. Update for change to option_ptr.
8362 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8363 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8364 0. Update for changes to optinfo_verbosity_options and
8366 (opt_info_switch_p): Convert optgroup_flags from int to
8368 (dump_basic_block): Use dump_flags_t rather than int
8369 for "dump_kind" param.
8370 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8371 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8372 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8373 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8374 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8375 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8376 TDF_NONE): Convert from macros to...
8377 (enum dump_flag): ...this new enum.
8378 (dump_flags_t): Update to use enum.
8379 (operator|, operator&, operator~, operator|=, operator&=):
8380 Implement for dump_flags_t.
8381 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
8382 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
8383 Convert from macros to...
8384 (enum optgroup_flag): ...this new enum.
8385 (optgroup_flags_t): New typedef.
8386 (operator|, operator|=): Implement for optgroup_flags_t.
8387 (struct dump_file_info): Convert field "alt_flags" to
8388 dump_flags_t. Convert field "optgroup_flags" to
8390 (dump_basic_block): Use dump_flags_t rather than int for param.
8391 (dump_generic_expr_loc): Likewise.
8392 (dump_dec): Likewise.
8393 (dump_register): Convert param "optgroup_flags" to
8395 (opt_info_enable_passes): Likewise.
8396 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
8398 * gimple-pretty-print.c (debug): Likewise.
8399 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
8400 (merged_store_group::apply_stores): Likewise.
8401 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
8402 * gimple.c (verify_gimple_pp): Likewise.
8403 * graphite-poly.c (print_pbb_body): Likewise.
8404 * passes.c (pass_manager::register_one_dump_file): Convert
8405 local "optgroup_flags" to optgroup_flags_t.
8406 * print-tree.c (print_node): Use TDF_NONE rather than 0.
8408 (debug_body): Likewise.
8409 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
8410 to optgroup_flags_t.
8411 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
8413 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
8414 (convert_mult_to_fma): Likewise.
8415 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
8416 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
8417 * tree-vect-data-refs.c (dump_lower_bound): Convert param
8418 "dump_kind" to dump_flags_t.
8420 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8422 * config/rs6000/rs6000.c (min, max): Delete.
8424 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8426 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
8427 -mabi=spe and -mabi=no-spe.
8429 2018-06-08 Martin Liska <mliska@suse.cz>
8431 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
8432 where we expect an existing summary.
8434 2018-06-08 Martin Liska <mliska@suse.cz>
8436 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
8437 * ipa-inline.h (estimate_edge_growth): Likewise.
8439 2018-06-08 Martin Liska <mliska@suse.cz>
8441 * cgraph.c (function_version_hasher::hash): Use
8442 cgraph_node::get_uid ().
8443 (function_version_hasher::equal):
8444 * cgraph.h (cgraph_node::get_uid): New method.
8445 * ipa-inline.c (update_caller_keys): Use
8446 cgraph_node::get_uid ().
8447 (update_callee_keys): Likewise.
8448 * ipa-utils.c (searchc): Likewise.
8449 (ipa_reduced_postorder): Likewise.
8450 * lto-cgraph.c (input_node): Likewise.
8451 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
8452 * symbol-summary.h (symtab_insertion): Likewise.
8453 (symtab_removal): Likewise.
8454 (symtab_duplication): Likewise.
8455 * tree-pretty-print.c (dump_function_header): Likewise.
8456 * tree-sra.c (convert_callers_for_node): Likewise.
8458 2018-06-08 Martin Liska <mliska@suse.cz>
8460 * cgraph.c (symbol_table::create_edge): Always assign a new
8462 (symbol_table::free_edge): Do not recycle numbers.
8463 * cgraph.h (cgraph_edge::get): New method.
8464 * symbol-summary.h (symtab_removal): Use it.
8465 (symtab_duplication): Likewise.
8466 (call_summary::hashable_uid): Remove.
8468 2018-06-08 Martin Liska <mliska@suse.cz>
8470 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
8471 (initialize_growth_caches): Remove.
8472 (free_growth_caches): Likewise.
8473 (do_estimate_edge_time): Use edge_growth_cache.
8474 (do_estimate_edge_size): Likewise.
8475 (do_estimate_edge_hints): Likewise.
8476 * ipa-inline.c (reset_edge_caches): Likewise.
8477 (recursive_inlining): Likewise.
8478 (inline_small_functions): Likewise.
8479 * ipa-inline.h (initialize_growth_caches): Remove.
8480 (estimate_edge_size): Likewise.
8481 (estimate_edge_time): Likewise.
8482 (estimate_edge_hints): Likewise.
8483 (reset_edge_growth_cache): Likewise.
8484 * symbol-summary.h (call_summary::remove): New method.
8486 2018-06-08 Martin Liska <mliska@suse.cz>
8488 * ipa-cp.c (class edge_clone_summary): New summary.
8489 (grow_edge_clone_vectors): Remove.
8490 (ipcp_edge_duplication_hook): Remove.
8491 (class edge_clone_summary_t): New call_summary class.
8492 (ipcp_edge_removal_hook): Remove.
8493 (edge_clone_summary_t::duplicate): New function.
8494 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
8495 (create_specialized_node): Likewise.
8496 (ipcp_driver): Initialize edge_clone_summaries and do not
8499 2018-06-08 Martin Liska <mliska@suse.cz>
8501 * symbol-summary.h (get): New function.
8502 (call_summary::m_initialize_when_cloning): New class member.
8504 2018-06-08 Martin Liska <mliska@suse.cz>
8506 * cgraph.c (cgraph_node::remove): Do not recycle uid.
8507 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
8508 (symbol_table::allocate_cgraph_symbol): Do not set uid.
8509 * passes.c (uid_hash_t): Record removed_nodes by their uids.
8510 (remove_cgraph_node_from_order): Use the removed_nodes set.
8511 (do_per_function_toporder): Likwise.
8512 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
8513 instead of summary_uid.
8514 (symtab_removal): Likewise.
8515 (symtab_duplication): Likewise.
8517 2018-06-08 Martin Liska <mliska@suse.cz>
8519 * ipa-cp.c (ipcp_store_bits_results): Use
8520 ipcp_transformation_sum.
8521 (ipcp_store_vr_results): Likewise.
8522 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
8524 (ipcp_transformation_initialize): ... this.
8525 (ipa_set_node_agg_value_chain):
8526 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
8527 (write_ipcp_transformation_info): Likewise.
8528 (read_ipcp_transformation_info): Likewise.
8529 (ipcp_update_bits): Likewise.
8530 (ipcp_update_vr): Likewise.
8531 (ipcp_transform_function): Likewise.
8532 * ipa-prop.h: Rename ipcp_transformation_summary to
8533 ipcp_transformation.
8534 (class ipcp_transformation_t): New function summary.
8535 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
8536 (ipa_get_agg_replacements_for_node): Likewise.
8538 2018-06-08 Martin Liska <mliska@suse.cz>
8540 * ipa-pure-const.c (struct funct_state_d): Do it class instead
8542 (class funct_state_summary_t): New function_summary class.
8543 (has_function_state): Remove.
8544 (get_function_state): Likewise.
8545 (set_function_state): Likewise.
8546 (add_new_function): Likewise.
8547 (funct_state_summary_t::insert): New function.
8548 (duplicate_node_data): Remove.
8549 (remove_node_data): Remove.
8550 (funct_state_summary_t::duplicate): New function.
8551 (register_hooks): Create new funct_state_summaries.
8552 (pure_const_generate_summary): Use it.
8553 (pure_const_write_summary): Likewise.
8554 (pure_const_read_summary): Likewise.
8555 (propagate_pure_const): Likewise.
8556 (propagate_nothrow): Likewise.
8557 (dump_malloc_lattice): Likewise.
8558 (propagate_malloc): Likewise.
8559 (execute): Do not register hooks, just remove summary
8561 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
8564 2018-06-08 Martin Liska <mliska@suse.cz>
8566 * ipa-reference.c (remove_node_data): Remove.
8567 (duplicate_node_data): Likewise.
8568 (class ipa_ref_var_info_summary_t): New class.
8569 (class ipa_ref_opt_summary_t): Likewise.
8570 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
8571 (get_reference_optimization_summary): Use
8572 ipa_ref_opt_sum_summaries.
8573 (set_reference_vars_info): Remove.
8574 (set_reference_optimization_summary): Likewise.
8575 (ipa_init): Create summaries.
8576 (init_function_info): Use function summary.
8577 (ipa_ref_opt_summary_t::duplicate): New function.
8578 (ipa_ref_opt_summary_t::remove): New function.
8579 (get_read_write_all_from_node): Fix GNU coding style.
8580 (propagate): Use function summary.
8581 (write_node_summary_p): Fix GNU coding style.
8582 (stream_out_bitmap): Likewise.
8583 (ipa_reference_read_optimization_summary): Use function summary.
8584 (ipa_reference_c_finalize): Do not release hooks.
8586 2018-06-08 Martin Liska <mliska@suse.cz>
8588 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
8589 (analyze_function_body): Extract multiple calls of get_create.
8590 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8591 * ipa-inline.c (recursive_inlining): Use ::get method.
8592 * ipa-inline.h (estimate_edge_growth): Likewise.
8594 2018-06-08 Martin Liska <mliska@suse.cz>
8596 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
8598 (hsa_function_summary::hsa_function_summary): Use the new enum
8600 (hsa_gpu_implementation_p): Use hsa_summaries::get.
8601 * hsa-gen.c (hsa_get_host_function): Likewise.
8602 (get_brig_function_name): Likewise.
8603 * ipa-hsa.c (process_hsa_functions): Likewise.
8604 (ipa_hsa_write_summary): Likewise.
8605 * symbol-summary.h (symtab_duplication): Use ::get function/
8606 (get): New function.
8608 2018-06-08 Martin Liska <mliska@suse.cz>
8610 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
8612 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
8613 (hsa_register_kernel): Likewise.
8614 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
8615 * hsa-gen.c (hsa_get_host_function): Likewise.
8616 (get_brig_function_name): Likewise.
8617 (generate_hsa): Likewise.
8618 (pass_gen_hsail::execute): Likewise.
8619 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
8620 (devirtualization_time_bonus): Likewise.
8621 (ipcp_propagate_stage): Likewise.
8622 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8623 (edge_set_predicate): Likewise.
8624 (evaluate_conditions_for_known_args): Likewise.
8625 (evaluate_properties_for_edge): Likewise.
8626 (ipa_fn_summary::reset): Likewise.
8627 (ipa_fn_summary_t::duplicate): Likewise.
8628 (dump_ipa_call_summary): Likewise.
8629 (ipa_dump_fn_summary): Likewise.
8630 (analyze_function_body): Likewise.
8631 (compute_fn_summary): Likewise.
8632 (estimate_edge_devirt_benefit): Likewise.
8633 (estimate_edge_size_and_time): Likewise.
8634 (estimate_calls_size_and_time): Likewise.
8635 (estimate_node_size_and_time): Likewise.
8636 (inline_update_callee_summaries): Likewise.
8637 (remap_edge_change_prob): Likewise.
8638 (remap_edge_summaries): Likewise.
8639 (ipa_merge_fn_summary_after_inlining): Likewise.
8640 (ipa_update_overall_fn_summary): Likewise.
8641 (read_ipa_call_summary): Likewise.
8642 (inline_read_section): Likewise.
8643 (write_ipa_call_summary): Likewise.
8644 (ipa_fn_summary_write): Likewise.
8645 (ipa_free_fn_summary): Likewise.
8646 * ipa-hsa.c (process_hsa_functions): Likewise.
8647 (ipa_hsa_write_summary): Likewise.
8648 (ipa_hsa_read_section): Likewise.
8649 * ipa-icf.c (sem_function::merge): Likewise.
8650 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8651 (do_estimate_edge_time): Likewise.
8652 (estimate_size_after_inlining): Likewise.
8653 (estimate_growth): Likewise.
8654 (growth_likely_positive): Likewise.
8655 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8656 (inline_call): Likewise.
8657 * ipa-inline.c (caller_growth_limits): Likewise.
8658 (can_inline_edge_p): Likewise.
8659 (can_inline_edge_by_limits_p): Likewise.
8660 (compute_uninlined_call_time): Likewise.
8661 (compute_inlined_call_time): Likewise.
8662 (want_inline_small_function_p): Likewise.
8663 (edge_badness): Likewise.
8664 (update_caller_keys): Likewise.
8665 (update_callee_keys): Likewise.
8666 (recursive_inlining): Likewise.
8667 (inline_small_functions): Likewise.
8668 (inline_to_all_callers_1): Likewise.
8669 (dump_overall_stats): Likewise.
8670 (early_inline_small_functions): Likewise.
8671 (early_inliner): Likewise.
8672 * ipa-inline.h (estimate_edge_growth): Likewise.
8673 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8674 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8675 * ipa-prop.h (IPA_NODE_REF): Likewise.
8676 (IPA_EDGE_REF): Likewise.
8677 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8678 (propagate_malloc): Likewise.
8679 * ipa-split.c (execute_split_functions): Likewise.
8680 * symbol-summary.h: Rename get to get_create.
8682 (get_create): Likewise.
8683 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8685 2018-06-08 Martin Liska <mliska@suse.cz>
8687 * symbol-summary.h (release): Move definition out of class
8689 (symtab_removal): Likewise.
8690 (symtab_duplication): Likewise.
8692 2018-06-08 Martin Liska <mliska@suse.cz>
8694 * symbol-summary.h (function_summary): Move constructor
8695 implementation out of class declaration.
8696 (release): Likewise.
8697 (symtab_insertion): Likewise.
8698 (symtab_removal): Likewise.
8699 (symtab_duplication): Likewise.
8702 2018-06-08 Martin Liska <mliska@suse.cz>
8704 * Makefile.in: Remove support for MPX (macros, related functions,
8705 fields in cgraph_node, ...).
8706 * builtin-types.def (BT_BND): Likewise.
8707 (BT_FN_BND_CONST_PTR): Likewise.
8708 (BT_FN_CONST_PTR_BND): Likewise.
8709 (BT_FN_VOID_PTR_BND): Likewise.
8710 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
8711 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
8712 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
8713 (expand_builtin_mempcpy_with_bounds): Likewise.
8714 (expand_builtin_memset_with_bounds): Likewise.
8715 (expand_builtin_memset_args): Likewise.
8716 (std_expand_builtin_va_start): Likewise.
8717 (expand_builtin): Likewise.
8718 (expand_builtin_with_bounds): Likewise.
8719 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
8720 (DEF_LIB_BUILTIN_CHKP): Likewise.
8721 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
8722 (DEF_CHKP_BUILTIN): Likewise.
8723 (BUILT_IN_MEMCPY): Likewise.
8724 (BUILT_IN_MEMMOVE): Likewise.
8725 (BUILT_IN_MEMPCPY): Likewise.
8726 (BUILT_IN_MEMSET): Likewise.
8727 (BUILT_IN_STPCPY): Likewise.
8728 (BUILT_IN_STRCAT): Likewise.
8729 (BUILT_IN_STRCHR): Likewise.
8730 (BUILT_IN_STRCPY): Likewise.
8731 (BUILT_IN_STRLEN): Likewise.
8732 (BUILT_IN_MEMCPY_CHK): Likewise.
8733 (BUILT_IN_MEMMOVE_CHK): Likewise.
8734 (BUILT_IN_MEMPCPY_CHK): Likewise.
8735 (BUILT_IN_MEMSET_CHK): Likewise.
8736 (BUILT_IN_STPCPY_CHK): Likewise.
8737 (BUILT_IN_STRCAT_CHK): Likewise.
8738 (BUILT_IN_STRCPY_CHK): Likewise.
8739 * calls.c (store_bounds): Likewise.
8740 (emit_call_1): Likewise.
8741 (special_function_p): Likewise.
8742 (maybe_warn_nonstring_arg): Likewise.
8743 (initialize_argument_information): Likewise.
8744 (finalize_must_preallocate): Likewise.
8745 (compute_argument_addresses): Likewise.
8746 (expand_call): Likewise.
8747 * cfgexpand.c (expand_call_stmt): Likewise.
8748 (expand_return): Likewise.
8749 (expand_gimple_stmt_1): Likewise.
8750 (pass_expand::execute): Likewise.
8751 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
8752 (cgraph_node::remove): Likewise.
8753 (cgraph_node::dump): Likewise.
8754 (cgraph_node::verify_node): Likewise.
8755 * cgraph.h (chkp_function_instrumented_p): Likewise.
8756 (symtab_node::get_alias_target): Likewise.
8757 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
8758 (cgraph_local_p): Likewise.
8759 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
8760 (cgraph_edge::rebuild_references): Likewise.
8761 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
8762 (walk_polymorphic_call_targets): Likewise.
8763 (cgraph_node::expand_thunk): Likewise.
8764 (symbol_table::output_weakrefs): Likewise.
8765 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
8766 (ix86_handle_option): Likewise.
8767 * config/i386/constraints.md: Likewise.
8768 * config/i386/i386-builtin-types.def (BND): Likewise.
8772 * config/i386/i386-builtin.def (BDESC_END): Likewise.
8773 (BDESC_FIRST): Likewise.
8775 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
8776 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
8777 * config/i386/i386.c (enum reg_class): Likewise.
8778 (ix86_target_string): Likewise.
8779 (ix86_option_override_internal): Likewise.
8780 (ix86_conditional_register_usage): Likewise.
8781 (ix86_valid_target_attribute_inner_p): Likewise.
8782 (ix86_set_indirect_branch_type): Likewise.
8783 (ix86_set_current_function): Likewise.
8784 (ix86_function_arg_regno_p): Likewise.
8785 (init_cumulative_args): Likewise.
8786 (ix86_function_arg_advance): Likewise.
8787 (ix86_function_arg): Likewise.
8788 (ix86_pass_by_reference): Likewise.
8789 (ix86_function_value_regno_p): Likewise.
8790 (ix86_function_value_1): Likewise.
8791 (ix86_function_value_bounds): Likewise.
8792 (ix86_return_in_memory): Likewise.
8793 (ix86_setup_incoming_vararg_bounds): Likewise.
8794 (ix86_va_start): Likewise.
8795 (indirect_thunk_need_prefix): Likewise.
8796 (print_reg): Likewise.
8797 (ix86_print_operand): Likewise.
8798 (ix86_expand_call): Likewise.
8799 (ix86_output_function_return): Likewise.
8800 (reg_encoded_number): Likewise.
8801 (BDESC_VERIFYS): Likewise.
8802 (ix86_init_mpx_builtins): Likewise.
8803 (ix86_init_builtins): Likewise.
8804 (ix86_emit_cmove): Likewise.
8805 (ix86_emit_move_max): Likewise.
8806 (ix86_expand_builtin): Likewise.
8807 (ix86_builtin_mpx_function): Likewise.
8808 (ix86_get_arg_address_for_bt): Likewise.
8809 (ix86_load_bounds): Likewise.
8810 (ix86_store_bounds): Likewise.
8811 (ix86_load_returned_bounds): Likewise.
8812 (ix86_store_returned_bounds): Likewise.
8813 (ix86_class_likely_spilled_p): Likewise.
8814 (ix86_hard_regno_mode_ok): Likewise.
8815 (x86_order_regs_for_local_alloc): Likewise.
8816 (ix86_mitigate_rop): Likewise.
8817 (ix86_bnd_prefixed_insn_p): Likewise.
8818 (ix86_mpx_bound_mode): Likewise.
8819 (ix86_make_bounds_constant): Likewise.
8820 (ix86_initialize_bounds): Likewise.
8821 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
8822 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
8823 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
8824 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
8825 (TARGET_CHKP_BOUND_MODE): Likewise.
8826 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
8827 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
8828 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
8829 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
8830 * config/i386/i386.h (TARGET_MPX): Likewise.
8831 (TARGET_MPX_P): Likewise.
8832 (VALID_BND_REG_MODE): Likewise.
8833 (FIRST_BND_REG): Likewise.
8834 (LAST_BND_REG): Likewise.
8835 (enum reg_class): Likewise.
8836 (BND_REG_P): Likewise.
8837 (BND_REGNO_P): Likewise.
8838 (BNDmode): Likewise.
8839 (ADJUST_INSN_LENGTH): Likewise.
8840 * config/i386/i386.md: Likewise.
8841 * config/i386/i386.opt: Likewise.
8842 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
8843 (defined): Likewise.
8844 (LINK_MPX): Likewise.
8845 (MPX_SPEC): Likewise.
8846 (LIBMPX_SPEC): Likewise.
8847 (LIBMPXWRAPPERS_SPEC): Likewise.
8848 (CHKP_SPEC): Likewise.
8849 * config/i386/predicates.md: Likewise.
8850 * dbxout.c (dbxout_type): Likewise.
8851 * doc/extend.texi: Likewise.
8852 * doc/invoke.texi: Likewise.
8853 * doc/md.texi: Likewise.
8854 * doc/tm.texi: Likewise.
8855 * doc/tm.texi.in: Likewise.
8856 * dwarf2out.c (is_base_type): Likewise.
8857 (gen_formal_types_die): Likewise.
8858 (gen_subprogram_die): Likewise.
8859 (gen_type_die_with_usage): Likewise.
8860 (gen_decl_die): Likewise.
8861 (dwarf2out_late_global_decl): Likewise.
8862 * expr.c (expand_assignment): Likewise.
8863 (emit_storent_insn): Likewise.
8864 (store_expr_with_bounds): Likewise.
8865 (store_expr): Likewise.
8866 (expand_expr_real_1): Likewise.
8867 * expr.h (store_expr_with_bounds): Likewise.
8868 * function.c (use_register_for_decl): Likewise.
8869 (struct bounds_parm_data): Likewise.
8870 (assign_parms_augmented_arg_list): Likewise.
8871 (assign_parm_find_entry_rtl): Likewise.
8872 (assign_parm_is_stack_parm): Likewise.
8873 (assign_parm_load_bounds): Likewise.
8874 (assign_bounds): Likewise.
8875 (assign_parms): Likewise.
8876 (expand_function_start): Likewise.
8877 * gcc.c (CHKP_SPEC): Likewise.
8878 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
8879 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
8880 (wrestrict_dom_walker::check_call): Likewise.
8881 * gimple.c (gimple_build_call_from_tree): Likewise.
8882 * gimple.h (enum gf_mask): Likewise.
8883 (gimple_call_with_bounds_p): Likewise.
8884 (gimple_call_set_with_bounds): Likewise.
8885 * gimplify.c (gimplify_init_constructor): Likewise.
8886 * ipa-cp.c (initialize_node_lattices): Likewise.
8887 (propagate_constants_across_call): Likewise.
8888 (find_more_scalar_values_for_callers_subset): Likewise.
8889 * ipa-hsa.c (process_hsa_functions): Likewise.
8890 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
8891 * ipa-icf.c (sem_function::merge): Likewise.
8892 * ipa-inline.c (early_inliner): Likewise.
8893 * ipa-pure-const.c (warn_function_noreturn): Likewise.
8894 (warn_function_cold): Likewise.
8895 (propagate_pure_const): Likewise.
8896 * ipa-ref.h (enum GTY): Likewise.
8897 * ipa-split.c (find_retbnd): Likewise.
8898 (consider_split): Likewise.
8899 (split_function): Likewise.
8900 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8901 * ipa.c (walk_polymorphic_call_targets): Likewise.
8902 (symbol_table::remove_unreachable_nodes): Likewise.
8903 (process_references): Likewise.
8904 (cgraph_build_static_cdtor_1): Likewise.
8905 * lto-cgraph.c (lto_output_node): Likewise.
8906 (output_refs): Likewise.
8907 (compute_ltrans_boundary): Likewise.
8908 (input_overwrite_node): Likewise.
8909 (input_node): Likewise.
8910 (input_cgraph_1): Likewise.
8911 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
8912 * passes.c (pass_manager::execute_early_local_passes): Likewise.
8913 (class pass_chkp_instrumentation_passes): Likewise.
8914 (make_pass_chkp_instrumentation_passes): Likewise.
8915 * passes.def: Likewise.
8916 * rtl.h (struct GTY): Likewise.
8917 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
8918 * stor-layout.c (layout_type): Likewise.
8919 * symtab.c: Likewise.
8920 * target.def: Likewise.
8921 * targhooks.c (default_chkp_bound_type): Likewise.
8922 (default_chkp_bound_mode): Likewise.
8923 (default_builtin_chkp_function): Likewise.
8924 (default_chkp_function_value_bounds): Likewise.
8925 (default_chkp_make_bounds_constant): Likewise.
8926 (default_chkp_initialize_bounds): Likewise.
8927 * targhooks.h (default_chkp_bound_type): Likewise.
8928 (default_chkp_bound_mode): Likewise.
8929 (default_builtin_chkp_function): Likewise.
8930 (default_chkp_function_value_bounds): Likewise.
8931 (default_chkp_make_bounds_constant): Likewise.
8932 (default_chkp_initialize_bounds): Likewise.
8933 * toplev.c (compile_file): Likewise.
8934 (process_options): Likewise.
8935 * tree-core.h (DEF_BUILTIN): Likewise.
8936 (DEF_BUILTIN_CHKP): Likewise.
8937 * tree-inline.c (declare_return_variable): Likewise.
8938 (remap_gimple_stmt): Likewise.
8939 (copy_bb): Likewise.
8940 (initialize_inlined_parameters): Likewise.
8941 (expand_call_inline): Likewise.
8942 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
8943 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
8944 (make_pass_ipa_chkp_produce_thunks): Likewise.
8945 (make_pass_chkp): Likewise.
8946 (make_pass_chkp_opt): Likewise.
8947 (make_pass_chkp_instrumentation_passes): Likewise.
8948 * tree-pretty-print.c (dump_generic_node): Likewise.
8949 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
8950 * tree-ssa-dce.c (propagate_necessity): Likewise.
8951 (eliminate_unnecessary_stmts): Likewise.
8952 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
8953 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
8954 * tree-ssa-sccvn.h: Likewise.
8955 * tree-ssa-strlen.c (get_string_length): Likewise.
8956 (valid_builtin_call): Likewise.
8957 (adjust_last_stmt): Likewise.
8958 (handle_builtin_strchr): Likewise.
8959 (handle_builtin_strcpy): Likewise.
8960 (handle_builtin_stxncpy): Likewise.
8961 (handle_builtin_memcpy): Likewise.
8962 (handle_builtin_strcat): Likewise.
8963 (strlen_check_and_optimize_stmt): Likewise.
8964 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
8965 * tree-streamer-in.c: Likewise.
8966 * tree-streamer.c (record_common_node): Likewise.
8967 * tree.c (tree_code_size): Likewise.
8968 (wide_int_to_tree_1): Likewise.
8969 (type_contains_placeholder_1): Likewise.
8970 (build_common_tree_nodes): Likewise.
8971 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
8972 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
8973 (POINTER_BOUNDS_P): Likewise.
8974 (BOUNDED_TYPE_P): Likewise.
8975 (BOUNDED_P): Likewise.
8976 (CALL_WITH_BOUNDS_P): Likewise.
8977 (pointer_bounds_type_node): Likewise.
8978 * value-prof.c (gimple_ic): Likewise.
8979 * var-tracking.c (vt_add_function_parameters): Likewise.
8980 * varasm.c (make_decl_rtl): Likewise.
8981 (assemble_start_function): Likewise.
8982 (output_constant): Likewise.
8983 (maybe_assemble_visibility): Likewise.
8984 * varpool.c (ctor_for_folding): Likewise.
8985 * chkp-builtins.def: Remove.
8986 * ipa-chkp.c: Remove.
8987 * ipa-chkp.h: Remove.
8988 * rtl-chkp.c: Remove.
8989 * rtl-chkp.h: Remove.
8990 * tree-chkp-opt.c: Remove.
8991 * tree-chkp.c: Remove.
8992 * tree-chkp.h: Remove.
8994 2018-06-07 Carl Love <cel@us.ibm.com>
8996 * config/rs6000/vsx.md (vextract_fp_from_shorth,
8997 vextract_fp_from_shortl): Add BE support.
8999 2018-06-07 Paul Koning <ni1d@arrl.net>
9001 * compare-elim.c (try_merge_compare): Don't merge compare if
9002 address contains a side effect.
9003 (try_eliminate_compare): Likewise.
9005 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
9007 * config.gcc: Support "tremont".
9008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
9009 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9011 * config/i386/i386.c (m_TREMONT): Define.
9012 (processor_target_table): Add "tremont".
9013 (PTA_TREMONT): Define.
9014 (ix86_lea_outperforms): Add TARGET_TREMONT.
9015 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
9016 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
9017 and M_INTEL_GOLDMONT_PLUS.
9018 (fold_builtin_cpu): Add "tremont".
9019 (ix86_add_stmt_cost): Add TARGET_TREMONT.
9020 (ix86_option_override_internal): Add "tremont".
9021 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
9022 (processor_type): Add PROCESSOR_TREMONT.
9023 * config/i386/x86-tune.def: Add m_TREMONT.
9024 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
9026 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9028 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
9029 symbol defined for msp430i* devices to be lower case.
9031 2018-06-07 Richard Biener <rguenther@suse.de>
9033 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
9034 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
9035 Properly wrap signed arithmetic if overflow wraps.
9037 2018-06-07 Jakub Jelinek <jakub@redhat.com>
9039 PR tree-optimization/69615
9040 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
9041 of a cast from a same precision integral SSA_NAME in a bb dominated
9042 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
9043 cast to utype if rhs2 has already a compatible type.
9045 2018-06-07 Richard Biener <rguenther@suse.de>
9047 PR tree-optimization/85935
9048 * graphite-scop-detection.c (find_params_in_bb): Analyze
9049 condition operands with respect to the correct loop. Assert
9050 the analysis doesn't fail.
9052 2018-06-04 Carl Love <cel@us.ibm.com>
9054 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
9055 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
9056 as it is slightly cheaper.
9057 (first_match_or_eos_index_<mode>):
9058 Calculate index using natural element order.
9059 (first_match_index_<mode>):
9060 Calculate index using natural element order.
9061 (first_match_or_eos_index_<mode>):
9062 Calculate index using natural order.
9063 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
9064 for BE and LE modes.
9065 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
9066 P9V_BUILTIN_VCLZLSBB_V16QI.
9067 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
9070 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9072 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
9073 indentation and line wrap for many prototypes. Add missing
9074 @smallexample directives around block of prototypes for vec_xl and
9077 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
9079 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
9080 track if we pass or return IEEE 128-bit floating point.
9081 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
9082 C++ mangling that is compatible with GCC 8.1.
9083 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
9084 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9085 floating point types.
9086 (rs6000_function_arg_advance_1): Likewise.
9087 (rs6000_mangle_type): Optionally generate mangled names that match
9088 what GCC 8.1 generated for IEEE 128-bit floating point types.
9089 (rs6000_globalize_decl_name): If we have an external function that
9090 passes or returns IEEE 128-bit types, generate a weak reference
9091 from the mangled name used in GCC 8.1 to the current mangled
9093 (rs6000_init_builtins): Make __ibm128 use the long double type if
9094 long double is IBM extended double. Make __float128 use the long
9095 double type if long double is IEEE 128-bit.
9098 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
9099 macro for __ibm128 built-in functions.
9100 (PACK_IF): Add __ibm128 pack/unpack functions.
9101 (UNPACK_IF): Likewise.
9102 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
9103 enable long double built-in functions if long double is IEEE
9104 128-bit floating point.
9105 (rs6000_invalid_builtin): Update long double built-in function
9107 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
9108 functions, adjust the built-in function to use the long double
9109 built-in function if __ibm128 and long double are the same type.
9110 * doc/extend.texi (PowerPC builtins): Update documention for
9111 __builtin_{,un}pack_longdouble. Add documentation for
9112 __builtin_{,un}pack_ibm128.
9114 2018-06-06 Jim Wilson <jimw@sifive.com>
9116 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
9117 (struct machine_function): New field interrupt_mode.
9118 (riscv_handle_type_attribute): New function. Add forward declaration.
9119 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
9120 (riscv_expand_epilogue): Check interrupt_mode field.
9121 (riscv_set_current_function): Check interrupt attribute args and
9122 set interrupt_mode field.
9123 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
9124 (riscv_sret, riscv_uret): New.
9125 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
9126 new arguments to interrupt attribute.
9128 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
9131 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
9132 Don't handle -mcpu=power8 if -mpower9-vector is also used.
9134 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9136 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
9137 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
9138 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9139 several redundant entries.
9141 2018-06-06 David Malcolm <dmalcolm@redhat.com>
9143 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
9144 type from "rtx" to "rtx_insn *".
9145 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
9146 for local "call_insn", removing cast.
9147 (ix86_expand_call): Likewise, introducing a "call_insn" local.
9149 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
9151 PR tree-optimization/86066
9152 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
9153 for BIT_INSERT_EXPR stores.
9155 2018-06-06 Richard Biener <rguenther@suse.de>
9157 PR tree-optimization/86062
9158 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9159 component refs ontop
9160 of to be offsetted base.
9162 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9164 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
9165 to be static and remove check on interrupt attribute name.
9167 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
9169 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9170 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
9172 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9175 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
9177 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
9178 Remove second argument, change how error is called.
9179 (aarch64_layout_arg): Remove second argument from
9180 aarch64_err_no_fpadvsimd call.
9181 (aarch64_init_cumulative_args): Ditto.
9182 (aarch64_gimplify_va_arg_expr): Ditto.
9183 * config/aarch64/aarch64.md (mov<mode>): Ditto.
9185 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
9187 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9188 (*simple_return_indirect_internal<mode>): Rename from
9189 simple_return_indirect_internal. Use W mode iterator.
9190 (rstorssp): New expander.
9191 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9192 (clrssbsy): New expander.
9193 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9195 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9197 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9199 (cmse_check_pointed_object): Likewise.
9201 2018-06-05 Martin Liska <mliska@suse.cz>
9203 PR gcov-profile/47618
9204 * doc/invoke.texi: Document how -fprofile-dir format
9207 2018-06-05 Richard Biener <rguenther@suse.de>
9209 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9210 removal pretend DOM info isn't available so we do not update
9211 it and only remove edges, not dominated blocks. Actually free
9212 DOM info in case we removed something. Remove unreachable blocks.
9213 (mfb_keep_latches): Work with either DOM info or marked backedges.
9214 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9215 first. Mark backedges if DOM info isn't available.
9216 (Re-)compute DOM info after cleanup_control_flow_pre.
9218 2018-06-05 Richard Biener <rguenther@suse.de>
9220 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9221 (locus_discrim_hasher::hash): Adjust.
9222 (locus_discrim_hasher::equal): Likewise.
9223 (next_discriminator_for_locus): Work on line directly.
9224 (same_line_p): Pass in expanded locus1 as well.
9225 (assign_discriminators): Avoid redundant location expansions.
9227 2018-06-05 Richard Biener <rguenther@suse.de>
9229 PR tree-optimization/86046
9230 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9231 if required after clearing TREE_ADDRESSABLE.
9233 2018-06-05 Richard Biener <rguenther@suse.de>
9235 PR tree-optimization/86047
9236 * tree-ssa-loop.c (for_each_index): Glob handling of all
9237 decls and constants and really handle all of them.
9239 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9242 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9243 qualifier_void_pointer and qualifier_const_void_pointer.
9244 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9245 (arm_init_builtins): Handle the above.
9246 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9247 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9248 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9251 2018-06-05 Martin Liska <mliska@suse.cz>
9253 * auto-profile.c (read_autofdo_file): Do not use
9254 gcov_ctr_summary struct.
9255 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9256 * coverage.c (struct counts_entry): Likewise.
9257 (read_counts_file): Read just single summary entry.
9258 (get_coverage_counts): Use gcov_summary struct.
9259 * coverage.h (get_coverage_counts): Likewise.
9260 * gcov-dump.c (dump_working_sets): Likewise.
9261 (tag_summary): Dump just single summary.
9262 * gcov-io.c (gcov_write_summary): Write just histogram
9264 (gcov_read_summary): Read just single summary.
9265 (compute_working_sets): Use gcov_summary struct.
9266 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9267 of GCOV_COUNTERS_SUMMABLE.
9268 (GCOV_COUNTERS_SUMMABLE): Remove.
9269 (GCOV_FIRST_VALUE_COUNTER): Replace with
9270 GCOV_COUNTER_V_INTERVAL.
9271 (struct gcov_ctr_summary): Remove.
9272 (struct gcov_summary): Directly use fields of former
9274 (compute_working_sets): Use gcov_summary struct.
9275 * gcov.c (read_count_file): Do not use ctrs fields.
9276 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9278 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9280 * profile.c: Likewise.
9281 * profile.h: Likewise.
9283 2018-06-05 Martin Liska <mliska@suse.cz>
9285 PR gcov-profile/84846
9286 * gcov.c (output_lines): Print working directory only
9287 in intermediate format.
9289 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9291 * config/s390/s390-builtin-types.def: Add void function type.
9292 * config/s390/s390-builtins.def: Use the function type for the
9295 2018-06-04 Jim Wilson <jimw@sifive.com>
9297 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9299 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9300 and maybe_eh_return. Change regno to unsigned int. Use new args to
9301 handle EH_RETURN_DATA_REGNO registers properly.
9302 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9303 (riscv_expand_epilogue): Update comment. Change argument name and
9304 type. Update code to use new name and type. Pass new args to
9305 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9307 * config/riscv/riscv.md (NORMAL_RETURN): New.
9308 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9309 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9310 (eh_return): Call gen_eh_return_internal and emit barrier.
9311 (eh_return_internal): Call riscv_expand_epilogue.
9313 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9315 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9316 bit_insertion field and declare can_be_merged_into method.
9317 (merged_store_group::can_be_merged_into): New method.
9318 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9319 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9320 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9322 2018-06-04 Richard Biener <rguenther@suse.de>
9324 PR tree-optimization/85955
9325 * builtins.c (fold_builtin_sincos): Convert pointers to
9326 destination to appropriate type before dereferencing.
9328 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9330 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9332 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9334 * expr.c (expand_expr_real_1): Force the operand into memory if
9335 its TYPE_MODE is BLKmode and if there is no integer mode for
9336 the number of bits being extracted.
9338 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9342 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9343 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9345 2018-06-04 Richard Biener <rguenther@suse.de>
9347 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9348 (cleanup_tree_cfg_noloop): ... single caller. Do
9349 start_recording_case_labels later.
9351 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9353 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9354 to _IMMINTRIN_H_INCLUDED.
9355 * config/i386/pconfigintrin.h: Ditto.
9356 * config/i386/waitpkgintrin.h: Ditto.
9357 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9358 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9359 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9360 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9361 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9362 waitpkgintrin.h and cldemoteintrin.h.
9364 2018-06-04 Richard Biener <rguenther@suse.de>
9366 PR tree-optimization/86038
9367 * tracer.c (find_best_successor): Check probability for
9368 being initialized, bail out if not.
9370 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9373 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9374 of bits to ignore when comparing architectures.
9376 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9378 PR tree-optimization/69615
9379 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
9380 maximum or minimum of the type, try to merge it also as if
9381 range1 is + [-, x - 1] or + [x + 1, -].
9384 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
9386 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
9388 PR tree-optimization/86034
9389 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
9390 the unsigned bitfield type in a bit insertion sequence if it does not
9391 have a larger precision than the bitfield size.
9392 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
9394 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
9396 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
9398 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9400 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
9401 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9402 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
9403 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
9405 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9407 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
9408 Disable -fdelete-null-pointer-checks for ELF toolchain.
9410 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9411 Kito Cheng <kito.cheng@gmail.com>
9413 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
9414 (nds32le-*-*, nds32be-*-*): Integrate checking process.
9415 (nds32*-*-*): Add glibc and uclibc conditions.
9416 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
9417 (TARGET_EXCEPT_UNWIND_INFO): Define.
9418 * config/nds32/elf.h: New file.
9419 * config/nds32/linux.h: New file.
9420 * config/nds32/nds32-elf.opt: New file.
9421 * config/nds32/nds32-linux.opt: New file.
9422 * config/nds32/nds32-fp-as-gp.c
9423 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
9424 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
9426 (nds32_asm_file_end): Ditto.
9427 (nds32_print_operand): Ditto.
9428 (nds32_insert_attributes): Ditto.
9429 (nds32_init_libfuncs): New function.
9430 (TARGET_HAVE_TLS): Define.
9431 (TARGET_INIT_LIBFUNCS): Define.
9432 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
9434 (TARGET_ELF): Apply different mcmodel setting.
9435 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
9436 been migrated into elf.h and linux.h files.
9437 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
9438 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
9439 (mcmodel): The content has been migrated into nds32-elf.opt and
9440 nds32-linux.opt files.
9441 * config/nds32/t-elf: New file.
9442 * config/nds32/t-linux: New file.
9444 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9445 Shiva Chen <shiva0217@gmail.com>
9447 * config/nds32/constants.md (unspec_volatile_element): Add
9448 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
9449 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
9451 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9452 (make_pass_nds32_fp_as_gp): Declare.
9453 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
9455 (nds32_asm_function_end_prologue): Remove unused asm output.
9456 (nds32_asm_function_begin_epilogue): Remove unused asm output.
9457 (nds32_asm_file_start): Output necessary fp_as_gp information.
9458 (nds32_option_override): Adjust register usage.
9459 (nds32_expand_prologue): Consider fp_as_gp situation.
9460 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
9461 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
9464 (simple_return): Ditto.
9465 (omit_fp_begin): Output special directive for fp_as_gp.
9466 (omit_fp_end): Output special directive for fp_as_gp.
9467 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
9468 mforbid-fp-as-gp): New options.
9470 2018-06-01 Mark Wielaard <mark@klomp.org>
9472 * dwarf2out.c (dwarf2out_finish): Remove generation of
9473 DW_AT_loclists_base.
9475 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
9477 * gimple-ssa-store-merging.c: Include gimple-fold.h.
9478 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
9479 (struct merged_store_group): Add bit_insertion field.
9480 (dump_char_array): Use standard hexadecimal format.
9481 (merged_store_group::merged_store_group): Set bit_insertion to false.
9482 (merged_store_group::apply_stores): Use optimal buffer size. Deal
9483 with BIT_INSERT_EXPR stores. Move up code updating the mask and
9484 also print the mask in the dump file.
9485 (pass_store_merging::gate): Minor tweak.
9486 (imm_store_chain_info::coalesce_immediate): Fix wrong association
9487 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
9488 stores with INTEGER_CST stores.
9489 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
9490 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
9491 and use it throughout. Generate bit insertion sequences if need be.
9492 (pass_store_merging::process_store): Remove redundant condition.
9493 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
9495 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
9497 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
9498 the 128-bit floating point types. Fix function comment.
9500 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9502 * config/aarch64/aarch64-simd.md
9503 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
9505 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
9508 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9510 PR tree-optimization/85989
9511 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
9513 (backprop::intersect_uses): Check it when deciding whether this
9514 is a backedge reference.
9515 (backprop::process_block): Add each phi to m_visited_phis
9516 after visiting it, then clear it at the end.
9518 2018-06-01 Richard Biener <rguenther@suse.de>
9520 * tree-vectorizer.h (vect_dr_stmt): New function.
9521 (vect_get_load_cost): Adjust.
9522 (vect_get_store_cost): Likewise.
9523 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
9524 Use vect_dr_stmt instead of DR_SMTT.
9525 (vect_record_base_alignments): Likewise.
9526 (vect_calculate_target_alignment): Likewise.
9527 (vect_compute_data_ref_alignment): Likewise and make static.
9528 (vect_update_misalignment_for_peel): Likewise.
9529 (vect_verify_datarefs_alignment): Likewise.
9530 (vector_alignment_reachable_p): Likewise.
9531 (vect_get_data_access_cost): Likewise. Pass down
9532 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
9533 (vect_get_peeling_costs_all_drs): Likewise.
9534 (vect_peeling_hash_get_lowest_cost): Likewise.
9535 (vect_enhance_data_refs_alignment): Likewise.
9536 (vect_find_same_alignment_drs): Likewise.
9537 (vect_analyze_data_refs_alignment): Likewise.
9538 (vect_analyze_group_access_1): Likewise.
9539 (vect_analyze_group_access): Likewise.
9540 (vect_analyze_data_ref_access): Likewise.
9541 (vect_analyze_data_ref_accesses): Likewise.
9542 (vect_vfa_segment_size): Likewise.
9543 (vect_small_gap_p): Likewise.
9544 (vectorizable_with_step_bound_p): Likewise.
9545 (vect_prune_runtime_alias_test_list): Likewise.
9546 (vect_analyze_data_refs): Likewise.
9547 (vect_supportable_dr_alignment): Likewise.
9548 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9549 (vect_gen_prolog_loop_niters): Likewise.
9550 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9551 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
9553 (vect_recog_mask_conversion_pattern): Likewise.
9554 (vect_try_gather_scatter_pattern): Likewise.
9555 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
9556 to vect_get_store_cost.
9557 (vect_get_store_cost): Get stmt_info instead of DR.
9558 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
9559 (vect_get_load_cost): Get stmt_info instead of DR.
9561 2018-06-01 Richard Biener <rguenther@suse.de>
9564 * gimple-fold.c (var_decl_component_p): Also allow offsetted
9565 vars wrapped in MEM_REFs.
9567 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9569 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9570 Fix subreg tests so that we only return a choice between
9571 GENERAL_REGS and FP_REGS if the original classes included both.
9573 2018-06-01 Richard Biener <rguenther@suse.de>
9576 * tree-ssa-structalias.c (get_function_part_constraint):
9577 Handle NULL fi->decl.
9578 (find_func_aliases_for_call): Properly handle indirect
9579 fi from direct call.
9580 (find_func_clobbers): Likewise.
9581 (ipa_pta_execute): Likewise.
9582 (create_variable_info_for): For functions that are ifunc_resolver
9583 resolve to a varinfo that contains the result of the resolver call.
9584 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
9587 2018-05-31 Michael Collison <michael.collison@arm.com>
9589 * config/aarch64/aarch64.md:
9590 (*fix_to_zero_extenddfdi2): New pattern.
9591 * gcc.target/aarch64/fix_extend1.c: New testcase.
9593 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
9597 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
9598 and BUILT_IN_STRNCMP_EQ.
9599 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
9600 BUILT_IN_STRNCMP_EQ.
9601 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
9602 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9603 (gimple_fold_builtin): Likewise.
9604 * tree-ssa-strlen.c (compute_string_length): New function.
9605 (determine_min_obsize): New function.
9606 (handle_builtin_string_cmp): New function to handle calls to
9607 string compare functions.
9608 (strlen_optimize_stmt): Add handling to builtin string compare
9610 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
9611 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9612 * tree.c (build_common_builtin_nodes): Add new defines of
9613 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
9615 2018-05-31 Jakub Jelinek <jakub@redhat.com>
9618 * bb-reorder.c (pass_partition_blocks::gate): Return false for
9619 functions with naked attribute.
9621 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
9623 * config/i386/sse.md (avx_vec_concat<mode>):
9624 Substitute concat_tg_mode mode attribute with xtg_mode.
9625 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
9626 (concat_tg_mode): Remove mode attribute.
9628 2018-05-31 Martin Sebor <msebor@redhat.com>
9631 * calls.c (alloc_max_size): Correct a logic error/typo.
9632 Treat excessive arguments as infinite. Warn for invalid arguments.
9633 * doc/invoke.texi (-Walloc-size-larger-than): Update.
9635 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
9638 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
9639 and movx for Haswell.
9641 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
9642 Cesar Philippidis <cesar@codesourcery.com>
9645 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
9646 when emitting error on private/firstprivate reductions.
9647 * omp-low.c (lower_omp_target): Avoid reference-type processing
9648 on pointers for firstprivate clause.
9650 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9652 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
9655 * config/aarch64/aarch64-simd.md
9656 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
9657 (aarch64_ld1_x3_<mode>): Likewise
9658 (aarch64_st1x2<VALLDIF:mode>): Likewise
9659 (aarch64_st1_x2_<mode>): Likewise
9660 (aarch64_st1x3<VALLDIF:mode>): Likewise
9661 (aarch64_st1_x3_<mode>): Likewise
9662 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
9663 (vld1_s8_x3): Likewise.
9664 (vld1_u16_x3): Likewise.
9665 (vld1_s16_x3): Likewise.
9666 (vld1_u32_x3): Likewise.
9667 (vld1_s32_x3): Likewise.
9668 (vld1_u64_x3): Likewise.
9669 (vld1_s64_x3): Likewise.
9670 (vld1_f16_x3): Likewise.
9671 (vld1_f32_x3): Likewise.
9672 (vld1_f64_x3): Likewise.
9673 (vld1_p8_x3): Likewise.
9674 (vld1_p16_x3): Likewise.
9675 (vld1_p64_x3): Likewise.
9676 (vld1q_u8_x3): Likewise.
9677 (vld1q_s8_x3): Likewise.
9678 (vld1q_u16_x3): Likewise.
9679 (vld1q_s16_x3): Likewise.
9680 (vld1q_u32_x3): Likewise.
9681 (vld1q_s32_x3): Likewise.
9682 (vld1q_u64_x3): Likewise.
9683 (vld1q_s64_x3): Likewise.
9684 (vld1q_f16_x3): Likewise.
9685 (vld1q_f32_x3): Likewise.
9686 (vld1q_f64_x3): Likewise.
9687 (vld1q_p8_x3): Likewise.
9688 (vld1q_p16_x3): Likewise.
9689 (vld1q_p64_x3): Likewise.
9690 (vst1_s64_x2): Likewise.
9691 (vst1_u64_x2): Likewise.
9692 (vst1_f64_x2): Likewise.
9693 (vst1_s8_x2): Likewise.
9694 (vst1_p8_x2): Likewise.
9695 (vst1_s16_x2): Likewise.
9696 (vst1_p16_x2): Likewise.
9697 (vst1_s32_x2): Likewise.
9698 (vst1_u8_x2): Likewise.
9699 (vst1_u16_x2): Likewise.
9700 (vst1_u32_x2): Likewise.
9701 (vst1_f16_x2): Likewise.
9702 (vst1_f32_x2): Likewise.
9703 (vst1_p64_x2): Likewise.
9704 (vst1q_s8_x2): Likewise.
9705 (vst1q_p8_x2): Likewise.
9706 (vst1q_s16_x2): Likewise.
9707 (vst1q_p16_x2): Likewise.
9708 (vst1q_s32_x2): Likewise.
9709 (vst1q_s64_x2): Likewise.
9710 (vst1q_u8_x2): Likewise.
9711 (vst1q_u16_x2): Likewise.
9712 (vst1q_u32_x2): Likewise.
9713 (vst1q_u64_x2): Likewise.
9714 (vst1q_f16_x2): Likewise.
9715 (vst1q_f32_x2): Likewise.
9716 (vst1q_f64_x2): Likewise.
9717 (vst1q_p64_x2): Likewise.
9718 (vst1_s64_x3): Likewise.
9719 (vst1_u64_x3): Likewise.
9720 (vst1_f64_x3): Likewise.
9721 (vst1_s8_x3): Likewise.
9722 (vst1_p8_x3): Likewise.
9723 (vst1_s16_x3): Likewise.
9724 (vst1_p16_x3): Likewise.
9725 (vst1_s32_x3): Likewise.
9726 (vst1_u8_x3): Likewise.
9727 (vst1_u16_x3): Likewise.
9728 (vst1_u32_x3): Likewise.
9729 (vst1_f16_x3): Likewise.
9730 (vst1_f32_x3): Likewise.
9731 (vst1_p64_x3): Likewise.
9732 (vst1q_s8_x3): Likewise.
9733 (vst1q_p8_x3): Likewise.
9734 (vst1q_s16_x3): Likewise.
9735 (vst1q_p16_x3): Likewise.
9736 (vst1q_s32_x3): Likewise.
9737 (vst1q_s64_x3): Likewise.
9738 (vst1q_u8_x3): Likewise.
9739 (vst1q_u16_x3): Likewise.
9740 (vst1q_u32_x3): Likewise.
9741 (vst1q_u64_x3): Likewise.
9742 (vst1q_f16_x3): Likewise.
9743 (vst1q_f32_x3): Likewise.
9744 (vst1q_f64_x3): Likewise.
9745 (vst1q_p64_x3): Likewise.
9747 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9749 * config/msp430/msp430.c (msp430_output_labelref): Prepend
9750 user_label_prefix to name.
9752 * tree-core.h: Update comment about the format of NAME string
9753 passed to handler in attribute_spec.
9755 * config/msp430/msp430.md: Remove erroneous subreg expression from
9756 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
9757 zero_extend{q,h}isi2.
9759 2018-05-30 Borislav Petkov <bp@suse.de>
9761 * doc/extend.texi: Document some architecture specific
9762 constraints and sort entries.
9764 2018-05-30 Martin Sebor <msebor@redhat.com>
9767 * builtins.c (expand_builtin_stpcpy_1): New function.
9768 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
9769 only if the former succeeds.
9771 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9773 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
9776 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9778 * doc/invoke.texi (-flinker-output): Document
9780 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9782 * passes.c (ipa_write_summaries): Only modify statements if body
9784 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
9785 incrementally linking.
9786 (ipa_passes): Likewise.
9787 * lto-cgraph.c (lto_output_node): When incrementally linking do not
9788 pass down resolution info.
9789 * common.opt (flag_incremental_link): Update info.
9790 * gcc.c (plugin specs): Turn flinker-output=* to
9791 -plugin-opt=-linker-output-known
9792 * toplev.c (compile_file): Also cut compilation when doing incremental
9794 * flag-types. (enum lto_partition_model): Add
9795 LTO_LINKER_OUTPUT_NOLTOREL.
9796 (invoke.texi): Add -flinker-output docs.
9797 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
9798 link same way as WPA; do not stream in dead initializers.
9800 * dwarf2out.c (dwarf2out_die_ref_for_decl,
9801 darf2out_register_external_decl): Support incremental link.
9803 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9805 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
9807 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9809 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
9810 it down to simple_object_copy_lto_debug_sections.
9811 (run_gcc): Determine incremental LTO link time and configure
9812 lto1 into non-wpa mode, disable renaming of debug sections.
9814 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
9816 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9817 descriptions of various incorrectly documented functions.
9819 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
9822 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9823 address check not strict.
9825 2018-05-30 Richard Biener <rguenther@suse.de>
9827 PR tree-optimization/85964
9828 * tracer.c (better_p): Drop initialized count check, we only
9829 call the function with initialized counts now.
9830 (find_best_successor): Do find a best edge if one
9831 has uninitialized count.
9832 (find_best_predecessor): Likewise. Do BB frequency check only
9833 if count is initialized.
9835 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
9837 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
9838 (aarch64_ldrstr_offset_compare): New.
9839 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
9840 load/store orderings.
9841 (aarch64_gen_adjusted_ldpstp): Likewise.
9843 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
9845 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9846 Check for subset of GENERAL_REGS and FP_REGS.
9847 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
9850 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
9852 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
9853 and wi::to_poly_offset. Add the current offset and then check
9854 whether the sum fits, rather than using an unchecked addition of
9855 a checked term. Check for a shwi rather than a uhwi.
9856 * expr.c (get_bit_range): Use tree_to_poly_uint64.
9857 (store_constructor): Use poly_int_tree_p.
9858 (expand_expr_real_1): Likewise.
9859 * function.c (assign_temp): Likewise.
9860 * fold-const.c (const_binop): Use poly_int_tree_p and
9862 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
9864 * ipa-icf-gimple.c (func_checker::compare_operand): Use
9865 to_poly_offset for MEM offsets.
9866 * ipa-icf.c (sem_variable::equals): Likewise.
9867 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
9868 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
9869 wi::to_poly_offset for BIT_FIELD_REF offsets.
9870 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
9872 * var-tracking.c (emit_note_insn_var_location): Use
9873 tree_to_poly_uint64.
9875 2018-05-29 Jim Wilson <jimw@sifive.com>
9877 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
9879 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
9882 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
9883 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
9885 (sse4_1_round<mode>2): Use nonimmediate_operand
9886 for operand 1 predicate.
9888 2018-05-29 Martin Sebor <msebor@redhat.com>
9889 Richard Biener <rguenther@suse.de>
9892 * calls.c (get_size_range): Call determine_value_range instead
9893 of get_value_range..
9894 * tree-vrp.h (determine_value_range): Declared new function.
9895 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
9897 2018-05-29 Richard Biener <rguenther@suse.de>
9899 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
9900 sure to use non-pattern stmts for get_earlier_stmt arguments.
9901 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
9902 called on pattern stmts.
9903 (get_later_stmt): Likewise.
9905 2018-05-29 Martin Liska <mliska@suse.cz>
9907 PR gcov-profile/85759
9908 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
9911 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9913 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
9915 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
9919 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9920 VEC_PACK_FLOAT_EXPR): New tree codes.
9921 * tree-pretty-print.c (op_code_prio): Handle
9922 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9923 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9924 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9925 * tree-inline.c (estimate_operator_cost): Likewise.
9926 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
9927 * fold-const.c (const_binop): Likewise.
9928 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
9929 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9930 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9931 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
9932 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9933 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9934 * expr.c (expand_expr_real_2): Likewise.
9935 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
9936 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
9937 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
9939 * optabs.c (expand_widen_pattern_expr): For
9940 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
9941 sign from result type rather than operand's type.
9942 (expand_binop_directly): For vec_packu_float_optab and
9943 vec_packs_float_optab allow result type to be different from operand's
9945 * optabs-tree.c (optab_for_tree_code): Handle
9946 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9947 VEC_PACK_FLOAT_EXPR. Formatting fixes.
9948 * tree-vect-generic.c (expand_vector_operations_1): Handle
9949 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9950 VEC_PACK_FLOAT_EXPR.
9951 * tree-vect-stmts.c (supportable_widening_operation): Handle
9953 (supportable_narrowing_operation): Handle FLOAT_EXPR.
9954 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
9955 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
9956 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
9957 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
9959 (vec_pack<floatprefix>_float_<mode>): New expander.
9960 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
9962 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
9963 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
9964 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
9965 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
9966 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
9968 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
9969 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
9970 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9971 VEC_PACK_FLOAT_EXPR): Document.
9973 2018-05-29 Richard Biener <rguenther@suse.de>
9975 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
9977 (stmt_vec_info_vec): Make pointer.
9978 (init_stmt_vec_info_vec): Remove.
9979 (free_stmt_vec_info_vec): Likewise.
9980 (set_stmt_vec_info_vec): New function.
9981 (free_stmt_vec_infos): Likewise.
9982 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
9983 (set_vinfo_for_stmt): Likewise.
9984 (get_earlier_stmt): Likewise.
9985 (get_later_stmt): Likewise.
9986 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
9987 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
9988 (vec_info::~vec_info): Free stmt_vec_infos.
9989 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
9990 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
9991 (pass_slp_vectorize::execute): Likewise.
9992 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
9993 (free_stmt_vec_info_vec): Likewise.
9994 (set_stmt_vec_info_vec): New function.
9995 (free_stmt_vec_infos): Likewise.
9996 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
9997 the global stmt_vec_info_vec.
9998 * tree-parloops.c (gather_scalar_reductions): Use
9999 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
10002 2018-05-29 Richard Biener <rguenther@suse.de>
10004 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
10006 2018-05-29 Martin Liska <mliska@suse.cz>
10007 David Malcolm <dmalcolm@redhat.com>
10009 * vec.c (test_reverse): New.
10010 (vec_c_tests): Add new test.
10011 * vec.h (vl_ptr>::reverse): New function.
10013 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
10015 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
10017 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
10020 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
10022 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
10024 2018-05-28 Richard Biener <rguenther@suse.de>
10026 PR tree-optimization/85933
10027 * tree-vect-data-refs.c (vect_record_base_alignments): Only
10028 look at stmts marked as vectorizable.
10030 2018-05-28 Richard Biener <rguenther@suse.de>
10032 PR tree-optimization/85934
10033 * tree-vect-generic.c (expand_vector_operations_1): Hoist
10034 vector boolean check before scalar optimization.
10036 2018-05-28 Jakub Jelinek <jakub@redhat.com>
10038 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
10041 2018-05-28 Mark Wielaard <mark@klomp.org>
10043 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
10044 if it is an expression containing a minus sign.
10046 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
10048 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
10050 2018-05-27 Paul Koning <ni1d@arrl.net>
10052 * config/pdp11/pdp11.md (truncsihi2): Remove.
10054 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10055 Chung-Ju Wu <jasonwucj@gmail.com>
10057 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
10059 (unaligned_store_dw): Ditto.
10060 * config/nds32/nds32-memory-manipulation.c
10061 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
10062 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
10063 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
10064 (emit_setmem_word_loop): Rename to ...
10065 (emit_setmem_doubleword_loop): ... this.
10066 (nds32_gen_dup_4_byte_to_word_value): New function.
10067 (nds32_gen_dup_8_byte_to_double_word_value): New function.
10068 (nds32_expand_setmem_loop): Refine implementation.
10069 (nds32_expand_setmem_loop_v3m): Ditto.
10070 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
10073 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10075 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
10077 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10079 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
10080 (nds32_init_machine_status): Initialize machine->attr_naked_p and
10081 machine->attr_no_prologue_p.
10082 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
10083 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
10084 (nds32_expand_epilogue): Consider attr_naked_p.
10085 (nds32_expand_epilogue_v3pop): Likewise.
10086 (nds32_can_use_return_insn): Likewise.
10087 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
10088 attr_no_prologue_p fields.
10089 * config/nds32/nds32.opt (mret-in-naked-func): New option.
10091 2018-05-27 Jakub Jelinek <jakub@redhat.com>
10094 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
10096 * config/i386/sse.md
10097 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
10099 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
10101 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
10103 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
10105 (*<floatsuffix>floatv2div2sf2): Rename to ...
10106 (*float<floatunssuffix>v2div2sf2): ... this.
10107 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
10108 (float<floatunssuffix>v2div2sf2_mask): ... this.
10109 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
10110 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
10111 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
10113 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
10115 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10117 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10119 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10121 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10123 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
10124 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
10125 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
10126 gen_ufix_truncv8dfv8si2.
10127 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
10128 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
10129 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
10130 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
10131 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
10132 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
10133 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
10134 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
10136 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10140 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
10142 2018-05-25 Jim Wilson <jimw@sifive.com>
10144 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
10145 * config/riscv/riscv.c (struct machine_function): Add
10146 interrupt_handler_p and attribute_checked_p fields.
10147 (riscv_attribute_table): Add interrupt.
10148 (riscv_interrupt_type_p): New.
10149 (riscv_save_reg_p): Save extra regs for interrupt handler.
10150 (riscv_use_save_libcall): Return false for interrupt handler.
10151 (riscv_first_stack_step): Add forward declaration.
10152 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
10153 for interrupt handler with large frame. Use it for saved reg list.
10154 (riscv_expand_prologue): Move flag_stack_usage_info support to
10155 eliminate duplication.
10156 (riscv_expand_epilogue): Generate mret for interrupt handler.
10157 (riscv_epilogue_uses): New.
10158 (riscv_can_use_return_insn): Return false for interrupt handler.
10159 (riscv_function_ok_for_sibcall): Likewise.
10160 (riscv_set_current_function): Add interrupt handler support.
10161 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
10162 * config/riscv/riscv.md (UNSPECV_MRET): New.
10164 (riscv_frflags, riscv_fsflags): Use tab after opcode.
10166 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
10168 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
10170 PR tree-optimization/85712
10171 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
10172 this candidate has already been replaced in-situ by a copy.
10174 2018-05-25 Jason Merrill <jason@redhat.com>
10176 PR c++/80485 - inline function non-zero address.
10177 * symtab.c (nonzero_address): Check DECL_COMDAT.
10179 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
10182 * config/alpha/alpha.md (ashlsi3): New insn pattern.
10183 (*ashlsi_se): Rename from *ashldi_se. Define as sign
10184 extension of SImode operation. Use const123_operand predicate.
10185 (*saddsi_1): Remove.
10186 (*saddl_se_1): Ditto.
10187 (*ssubsi_1): Ditto.
10188 (*ssubl_se_1): Ditto.
10189 * config/alpha/predicates.md (const123_operand): New predicate.
10190 * config/alpha/constraints.md (P): Use IN_RANGE.
10192 2018-05-25 Richard Biener <rguenther@suse.de>
10194 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10196 (ref_maybe_used_by_stmt_p): Likewise.
10197 (stmt_may_clobber_ref_p): Likewise.
10198 (stmt_may_clobber_ref_p_1): Likewise.
10199 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10201 (ref_maybe_used_by_stmt_p): Likewise.
10202 (stmt_may_clobber_ref_p): Likewise.
10203 (stmt_may_clobber_ref_p_1): Likewise.
10204 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10205 the alias oracle to disambiguate DRs with stmts DR analysis
10207 (vect_analyze_data_refs): Do not give up on not analyzable
10208 DRs for BB vectorization. Remove code truncating the dataref
10211 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10214 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10215 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10216 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10218 2018-05-25 Richard Biener <rguenther@suse.de>
10220 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10221 function, combining stmt data ref gathering and fatal analysis
10223 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10224 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10225 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10226 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10227 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10229 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10231 PR tree-optimization/85720
10232 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10233 SCC if all partitions are builtins.
10234 (version_loop_by_alias_check): New parameter. Generate cancelable
10235 runtime alias check if all partitions are builtins.
10236 (distribute_loop): Update call to above function.
10238 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10240 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10241 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10242 (parm_default_def_partition_arg): Ditto.
10243 (set_parm_default_def_partition): Ditto.
10244 (get_parm_default_def_partitions): Ditto and make it static.
10245 (get_undefined_value_partitions): Ditto and make it static.
10246 (remove_ssa_form): Refactor call to init_var_map here.
10247 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10248 computation for loop region.
10249 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10250 (register_default_def): Delete.
10251 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10252 (parm_default_def_partition_arg): Ditto.
10253 (set_parm_default_def_partition): Ditto.
10254 (get_parm_default_def_partitions): Ditto and make it static.
10255 (get_undefined_value_partitions): Ditto and make it static.
10256 (coalesce_with_default, coalesce_with_default): Update comment.
10257 (create_coalesce_list_for_region): New func factored out from
10258 create_outofssa_var_map.
10259 (populate_coalesce_list_for_outofssa): New func factored out from
10260 create_outofssa_var_map and coalesce_ssa_name.
10261 (create_outofssa_var_map): Delete.
10262 (coalesce_ssa_name): Refactor to support live range computation.
10263 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10264 (get_parm_default_def_partitions): Delete.
10265 (get_undefined_value_partitions): Ditto.
10266 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10267 computation for loop region.
10268 (new_tree_live_info, loe_visit_block): Ditto.
10269 (live_worklist, set_var_live_on_entry): Ditto.
10270 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10271 * tree-ssa-live.h (struct _var_map): New fields.
10272 (init_var_map): Change decl.
10273 (region_contains_p): New.
10275 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10277 * tree-ssa-live.h (live_merge_and_clear): Delete.
10279 2018-05-25 Richard Biener <rguenther@suse.de>
10282 * tree-dump.c (dequeue_and_dump): Remove access to removed
10283 operand 2 of a SWITCH_EXPR.
10285 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10287 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10288 multiplication and division.
10289 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10290 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10291 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10292 (cond_udiv_optab, cond_umod_optab): New optabs.
10293 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10294 (IFN_COND_RDIV): New internal functions.
10295 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10296 TRUNC_MOD_EXPR and RDIV_EXPR.
10297 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10298 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10300 (SVE_INT_BINARY): Include mult.
10301 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10302 (optab, sve_int_op): Handle mult.
10303 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10305 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10306 for SVE_INT_BINARY_SD.
10308 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10310 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10311 (optab, sve_int_op): Handle div and udiv.
10312 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10313 for SVE_INT_BINARY_SD.
10314 (*<optab><mode>3): New insn for the same.
10316 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10318 * tree-vect-patterns.c: Include predict.h.
10319 (vect_recog_divmod_pattern): Restrict check for division support
10320 to when optimizing for size.
10322 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10324 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10325 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10326 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10327 (gimple_match_op::set_op): Likewise.
10328 (gimple_resimplify4): Declare.
10329 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10330 (expr::gen_transform): Likewise.
10331 (decision_tree::gen): Generate a simplification routine for 4 operands.
10332 * gimple-match-head.c (gimple_simplify): Add an overload for
10333 4 operands. In the top-level function, handle up to 4 call
10334 arguments and call gimple_resimplify4.
10335 (gimple_resimplify4): New function.
10336 (build_call_internal): Pass a fourth operand.
10337 (maybe_push_to_seq): Likewise.
10338 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10339 Fold VEC_COND_EXPRs of an operation and a default value into
10340 an IFN_COND_* function if possible.
10341 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10343 (SVE_COND_FP_BINARY): Include them.
10344 (optab, sve_fp_op): Handle them.
10345 (SVE_INT_BINARY_REV): New code iterator.
10346 (SVE_COND_FP_BINARY_REV): New int iterator.
10347 (commutative): New int attribute.
10348 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10350 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10352 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10353 (*cond_<optab><mode>): New patterns for reversed operands.
10355 2018-05-25 Richard Biener <rguenther@suse.de>
10357 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10358 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10359 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10360 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10361 * tree-vect-data-refs.c (everywhere): Adjust users.
10362 * tree-vect-loop.c (everywhere): Likewise.
10363 * tree-vect-slp.c (everywhere): Likewise.
10364 * tree-vect-stmts.c (everywhere): Likewise.
10365 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10367 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10369 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10371 (gcc_cv_as_section_exclude): ... this.
10372 Try Solaris as #exclude syntax.
10373 * configure: Regenerate.
10374 * config.in: Regenerate.
10375 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
10377 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
10378 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
10380 * varasm.c (default_elf_asm_named_section): Don't check if
10381 HAVE_GAS_SECTION_EXCLUDE is defined.
10383 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10385 * doc/md.texi: Update the documentation of the cond_* optabs
10386 to mention the new final operand. Fix GET_MODE_NUNITS call.
10387 Describe the scalar case too.
10388 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
10389 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
10391 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
10392 (get_conditional_internal_fn): Update comment.
10393 * tree-vect-loop.c (vectorizable_reduction): Pass the original
10394 accumulator value as a final argument to conditional functions.
10395 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
10396 a define_expand and add an "else" operand. Assert for now that
10397 the else operand is equal to operand 2. Use SVE_INT_BINARY and
10398 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
10399 (*cond_<optab><mode>): New patterns.
10400 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
10401 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10402 (UNSPEC_COND_EOR): Delete.
10403 (optab): Remove associated mappings.
10404 (SVE_INT_BINARY): New code iterator.
10405 (sve_int_op): Remove int attribute and add "minus" to the code
10407 (SVE_COND_INT_OP): Delete.
10408 (SVE_COND_FP_OP): Rename to...
10409 (SVE_COND_FP_BINARY): ...this.
10411 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10413 * optabs.c (can_reuse_operands_p): New function.
10414 (maybe_legitimize_operands): Try to reuse the results for
10417 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10419 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10420 Add {q} suffix to insn mnemonic.
10422 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10424 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
10425 (msp430_warn_func_return): New.
10427 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
10429 * fold-const.c (tree_nonzero_bits): New function.
10430 * fold-const.h (tree_nonzero_bits): Likewise.
10431 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
10432 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
10434 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10438 * varasm.c (assemble_alias): Check ifunc_resolver only on
10441 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10444 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
10445 when memory input operand is handled.
10447 2018-05-24 Luis Machado <luis.machado@linaro.org>
10449 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
10451 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
10453 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10455 * match.pd: Delay FMA folds until after vectorization.
10457 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
10460 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10461 address check not strict.
10463 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10465 * gimple-match.h (gimple_match_op): New class.
10466 (mprts_hook): Replace parameters with a gimple_match_op *.
10467 (maybe_build_generic_op): Likewise.
10468 (gimple_simplified_result_is_gimple_val): Replace parameters with
10469 a const gimple_match_op *.
10470 (gimple_simplify): Replace code_helper * and tree * parameters with
10471 a gimple_match_op * parameter.
10472 (gimple_resimplify1): Replace code_helper *, tree and tree *
10473 parameters with a gimple_match_op * parameter.
10474 (gimple_resimplify2): Likewise.
10475 (gimple_resimplify3): Likewise.
10476 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
10477 parameters with a gimple_match_op * parameter.
10478 * gimple-match-head.c (gimple_simplify): Change prototypes of
10479 auto-generated functions to take a gimple_match_op * instead of
10480 separate code_helper * and tree * parameters. Make the same
10481 change in the top-level overload and update calls to the
10482 gimple_resimplify routines. Update calls to the auto-generated
10483 functions and to maybe_push_res_to_seq in the publicly-facing
10484 operation-specific gimple_simplify overloads.
10485 (gimple_match_op::MAX_NUM_OPS): Define.
10486 (gimple_resimplify1): Replace rcode and ops with a single res_op
10487 parameter. Update call to gimple_simplify.
10488 (gimple_resimplify2): Likewise.
10489 (gimple_resimplify3): Likewise.
10490 (mprts_hook): Replace parameters with a gimple_match_op *.
10491 (maybe_build_generic_op): Likewise.
10492 (build_call_internal): Replace type, nargs and ops with
10493 a gimple_match_op *.
10494 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
10495 with a single gimple_match_op *. Update calls to mprts_hook,
10496 build_call_internal and gimple_simplified_result_is_gimple_val.
10497 Factor out code that is common to the tree_code and combined_fn cases.
10498 * genmatch.c (expr::gen_transform): Replace tem_code and
10499 tem_ops with a gimple_match_op called tem_op. Update calls
10500 to the gimple_resimplify functions and maybe_push_res_to_seq.
10501 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
10502 res_ops. Update call to the gimple_resimplify functions.
10503 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
10504 (decision_tree::gen): Make the functions take a gimple_match_op *
10505 called res_op instead of separate res_code and res_ops parameters.
10506 Update call accordingly.
10507 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
10508 and ops with a single res_op parameter. Update calls to
10509 maybe_build_generic_op and maybe_push_res_to_seq.
10510 (fold_stmt_1): Update calls to gimple_simplify and
10511 replace_stmt_with_simplification.
10512 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
10513 and gimple_simplified_result_is_gimple_val.
10514 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
10516 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
10517 with a gimple_match_op *.
10518 (vn_nary_build_or_lookup): Likewise. Update call to
10519 vn_nary_build_or_lookup_1.
10520 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
10521 gimple_match_op *. Update calls to the gimple_resimplify routines
10522 and to gimple_simplified_result_is_gimple_val.
10523 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
10524 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
10525 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
10526 (visit_nary_op): Likewise.
10527 (visit_reference_op_load): Likewise.
10529 2018-05-23 Luis Machado <luis.machado@linaro.org>
10531 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
10532 modifier for printing the step amount.
10534 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
10537 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
10540 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
10542 * doc/sourcebuild.texi (Endianness): New subsubsection.
10544 2018-05-23 Luis Machado <luis.machado@linaro.org>
10546 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10547 <prefetch_dynamic_strides>: New const bool field.
10548 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10549 prefetch_dynamic_strides.
10550 (exynosm1_prefetch_tune): Likewise.
10551 (thunderxt88_prefetch_tune): Likewise.
10552 (thunderx_prefetch_tune): Likewise.
10553 (thunderx2t99_prefetch_tune): Likewise.
10554 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
10556 (aarch64_override_options_internal): Update to set
10557 PARAM_PREFETCH_DYNAMIC_STRIDES.
10558 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10559 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10560 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10561 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10562 prefetch-dynamic-strides setting.
10564 2018-05-23 Luis Machado <luis.machado@linaro.org>
10566 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10567 <minimum_stride>: New const int field.
10568 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10569 minimum_stride field defaulting to -1.
10570 (exynosm1_prefetch_tune): Likewise.
10571 (thunderxt88_prefetch_tune): Likewise.
10572 (thunderx_prefetch_tune): Likewise.
10573 (thunderx2t99_prefetch_tune): Likewise.
10574 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
10575 <default_opt_level>: Set to 3.
10576 (aarch64_override_options_internal): Update to set
10577 PARAM_PREFETCH_MINIMUM_STRIDE.
10578 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10579 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10580 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10581 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10582 stride is constant and is below the minimum stride threshold.
10584 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10586 * config/arm/arm-cpus.in (mode26): Delete.
10587 (armv4): Delete mode26 reference.
10588 * config/arm/arm.c (arm_configure_build_target): Delete use of
10591 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10593 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
10595 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
10596 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
10598 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
10599 (floatunsdidf2): Ditto.
10601 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10603 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
10604 (fixuns_trunc<mode>si2_avx512f): Ditto.
10605 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
10606 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
10607 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
10609 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10611 PR rtl-optimization/79985
10612 * df-scan.c (df_insn_refs_collect): Remove special case for
10613 global registers and asm statements.
10615 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10617 * extend.texi (Global Register Variables): Rewrite the bullet list.
10618 Note that the register is available for allocation. Note that access
10619 via inline asm must use constraints. Add note about async-signal
10620 handlers. Remove paragraph about automagic register selection.
10622 2018-05-23 Richard Biener <rguenther@suse.de>
10624 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
10625 of fixed offset from memset VN.
10627 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
10629 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
10630 first_interp field.
10631 (alloc_cand_and_find_basis): Initialize first_interp field.
10632 (slsr_process_mul): Modify first_interp field.
10633 (slsr_process_add): Likewise.
10634 (slsr_process_cast): Modify first_interp field for each new
10636 (slsr_process_copy): Likewise.
10637 (dump_candidate): Dump first_interp field.
10638 (replace_mult_candidate): Process all interpretations, not just
10640 (replace_rhs_if_not_dup): Likewise.
10641 (replace_one_candidate): Likewise.
10643 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
10645 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
10647 (aarch64_needs_frame_chain): New function.
10648 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
10650 2018-05-23 Sudakshina Das <sudi.das@arm.com>
10653 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
10654 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
10655 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
10656 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
10657 as true for strict-align.
10658 (aarch64_can_inline_p): Perform checks even when callee has no
10659 attributes to check for strict alignment.
10660 * doc/extend.texi (AArch64 Function Attributes): Document
10662 * doc/invoke.texi: (AArch64 Options): Likewise.
10664 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
10666 PR tree-optimization/85853
10667 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
10668 the handling of the root of the node to...
10669 (vect_slp_analyze_node_operations_1): ...this new function,
10670 and run the whole thing with the child nodes' def types
10671 set according to their SLP node's def type.
10673 2018-05-23 Richard Biener <rguenther@suse.de>
10675 PR middle-end/85874
10676 * tree-data-ref.c (create_runtime_alias_checks): Defer
10677 and ignore overflow warnings.
10679 2018-05-23 Yury Gribov <tetra2005@gmail.com>
10681 PR tree-optimization/85822
10682 * tree-vrp.c (is_masked_range_test): Fix handling of negative
10685 2018-05-23 Richard Biener <rguenther@suse.de>
10687 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10688 memset constants via native_interpret_expr.
10690 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
10693 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
10695 (cgraph_node::create_alias): Likewise.
10696 (cgraph_node::get_availability): Check ifunc_resolver instead
10697 of looking up ifunc attribute.
10698 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10699 * varasm.c (do_assemble_alias): Likewise.
10700 (assemble_alias): Likewise.
10701 (default_binds_local_p_3): Likewise.
10702 * cgraph.h (cgraph_node): Add ifunc_resolver.
10703 (cgraph_node::only_called_directly_or_aliased_p): Return false
10704 for IFUNC resolver.
10705 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
10707 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
10708 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
10709 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
10710 instead of looking up ifunc attribute.
10712 2018-05-22 Luis Machado <luis.machado@linaro.org>
10714 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
10716 2018-05-22 Martin Sebor <msebor@redhat.com>
10718 PR middle-end/85359
10719 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
10720 only when expasion succeeds.
10721 (expand_builtin_strcmp): Same.
10722 (expand_builtin_strncmp): Same.
10724 2018-05-22 Martin Sebor <msebor@redhat.com>
10726 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
10728 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
10729 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10731 * config/aarch64/aarch64-ldpstp.md: Replace uses of
10732 aarch64_mem_pair_operand with memory_operand and delete operand swapping
10734 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10735 Add check for legitimate_address.
10736 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
10737 (aarch64_swap_ldrstr_operands): New.
10738 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
10741 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
10742 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10744 * config/aarch64/aarch64.md: New patterns to generate stp
10746 (store_pair_sw, store_pair_dw): New patterns to generate stp for
10747 single words and double words.
10748 (load_pair_sw, load_pair_dw): Likewise.
10749 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
10751 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
10753 * config/aarch64/aarch64-ldpstp.md: Modify peephole
10754 for different mode ldpstp and add peephole for merged zero stores.
10755 Likewise for loads.
10756 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10758 (aarch64_gen_store_pair): Rename calls to match new patterns.
10759 (aarch64_gen_load_pair): Rename calls to match new patterns.
10760 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
10761 (load_pair<DREG:mode><DREG2:mode>): ... This.
10762 (store_pair<mode>): Rename to...
10763 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
10764 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
10765 New mode iterators.
10766 (V_INT_EQUIV): Handle SImode.
10767 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
10770 2018-05-22 Martin Sebor <msebor@redhat.com>
10773 * calls.c (maybe_warn_nonstring_arg): Use string length to set
10774 or ajust the presumed bound on an operation to avoid unnecessary
10777 2018-05-22 Martin Sebor <msebor@redhat.com>
10779 PR tree-optimization/85826
10780 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
10781 assuming that a DECL necesarily has a constant size.
10783 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
10785 PR middle-end/85862
10786 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
10788 2018-05-22 Richard Biener <rguenther@suse.de>
10790 PR tree-optimization/85834
10791 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
10792 non-constant and non-zero memset arguments.
10794 2018-05-22 Martin Liska <mliska@suse.cz>
10797 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
10799 2018-05-22 Richard Biener <rguenther@suse.de>
10801 PR tree-optimization/85863
10802 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
10803 comparisons when vectype is specified.
10804 (vectorizable_condition): Do not specify vectype for
10805 vect_is_simple_cond when SLP vectorizing.
10807 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
10810 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
10811 define __ibm128 as long double.
10812 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
10813 as a distinct type when IEEE 128-bit support is enabled.
10814 (init_float128_ieee): Fix up conversions between IFmode and IEEE
10815 128-bit types to use the correct functions.
10816 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
10817 convert between 128-bit floating point types that have different
10818 modes but the same representation, instead of using gen_lowpart to
10820 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
10822 (IFKF_reg): New attributes to give the register constraints for
10824 (extend<mode>tf2_internal): New insns to mark an explicit
10825 conversion between 128-bit floating point types that have a
10826 different mode but share the same representation.
10828 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
10830 PR tree-optimization/85814
10831 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
10832 a null return from get_strinfo when unsharing the next
10833 strinfo in the chain.
10835 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10838 * varasm.c (weak_finish): Clean up weak_decls.
10840 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10842 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
10843 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
10844 UNSPEC_UADALP values.
10845 * config/aarch64/iterators.md (ABAL): New int iterator.
10848 (sur): Add mappings for the above.
10849 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
10851 (aarch64_<sur>abal<mode>_4): Likewise.
10852 (aarch64_<sur>adalp<mode>_3): Likewise.
10853 (<sur>sadv16qi): New define_expand.
10855 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
10857 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
10858 (*movdf_internal): Ditto.
10859 (*rcpsf2_sse): Ditto.
10860 (*rsqrtsf2_sse): Ditto.
10861 (*sqrt<mode>2_sse): Ditto.
10863 2018-05-21 Tamar Christina <tamar.christina@arm.com>
10865 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
10867 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
10868 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
10869 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
10870 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
10872 * config/aarch64/arm_neon.h: Likewise.
10873 * config/aarch64/iterators.md (VQ_I): New.
10875 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
10877 * config.gcc: Add arc/t-multilib-linux to tmake_file for
10879 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
10882 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10884 * config/nds32/constraints.md (S): New constraint.
10885 * config/nds32/nds32.md (call_internal): Use constraint S.
10886 (call_value_internal): Likewise.
10887 (sibcall_internal): Likewise.
10888 (sibcall_value_internal): Likewise.
10890 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10891 Chung-Ju Wu <jasonwucj@gmail.com>
10893 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
10894 into consideration.
10896 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10897 Chung-Ju Wu <jasonwucj@gmail.com>
10899 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
10900 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
10901 (nds32_rtx_costs_impl): Simplify.
10902 (nds32_address_cost_impl): Simplify.
10903 (nds32_init_rtx_costs): New function.
10904 (nds32_rtx_costs_speed_prefer): Likewise.
10905 (nds32_rtx_costs_size_prefer): Likewise.
10906 (nds32_address_cost_speed_prefer): Likewise.
10907 (nds32_address_cost_speed_fwprop): Likewise.
10908 (nds32_address_cost_size_prefer): Likewise.
10909 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
10910 * config/nds32/nds32.c (nds32_option_override): Use
10911 nds32_init_rtx_costs function.
10913 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10915 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
10916 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
10917 (TARGET_PIPELINE_N8): Likewise.
10918 (TARGET_PIPELINE_N10): Likewise.
10919 (TARGET_PIPELINE_N13): Likewise.
10920 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10922 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
10924 * config/nds32/nds32-fpu.md: Update copyright year.
10926 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10928 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
10930 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10932 * config/nds32/nds32.c
10933 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
10934 * config/nds32/nds32.opt (minline-asm-r15): New option.
10936 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10938 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
10940 * config/nds32/nds32.md (abssi2): New pattern.
10942 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
10944 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10945 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
10946 Merge insn pattern from sse_cvtsi2ss<round_name> and
10947 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
10948 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
10949 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
10950 using SWI48 mode iterator.
10951 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
10952 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
10953 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
10954 pattern from sse_cvttss2si<round_saeonly_name>
10955 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
10956 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
10957 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
10958 using SWI48 mode iterator.
10959 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
10960 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
10961 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
10962 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
10963 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
10964 using SWI48 mode iterator.
10965 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
10966 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
10967 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
10968 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
10969 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
10970 SWI48 mode iterator.
10971 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
10972 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
10973 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
10974 pattern from sse_cvttsd2si<round_saeonly_name>
10975 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
10977 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10979 * config/nds32/nds32-md-auxiliary.c
10980 (nds32_valid_smw_lwm_base_p): Refine.
10981 (nds32_output_smw_single_word): Refine.
10982 (nds32_output_smw_double_word): New.
10983 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
10985 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10987 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
10988 (nds32_output_stack_pop): Refine.
10989 (nds32_expand_unaligned_load): Refine.
10990 (nds32_expand_unaligned_store): Refine.
10992 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10993 Chung-Ju Wu <jasonwucj@gmail.com>
10995 * config/nds32/constants.md: Add TP_REGNUM constant.
10996 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
10997 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
10999 * config/nds32/nds32-doubleword.md: Consider flag_pic.
11000 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
11001 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11002 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
11003 and PIC code generation.
11004 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
11006 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
11008 * config/nds32/nds32.md: Support TLS and PIC.
11009 * config/nds32/nds32.c: Support TLS and PIC.
11010 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
11011 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
11014 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11016 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
11017 mode with E_ prefix.
11019 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11020 Chung-Ju Wu <jasonwucj@gmail.com>
11022 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
11023 * config/nds32/nds32-md-auxiliary.c
11024 (symbolic_reference_mentioned_p): New.
11025 (nds32_legitimize_ict_address): New.
11026 (nds32_expand_ict_move): New.
11027 (nds32_indirect_call_referenced_p): New.
11028 (nds32_symbol_binds_local_p): Delete.
11029 (nds32_long_call_p): Modify.
11030 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
11031 * config/nds32/nds32-protos.h
11032 (symbolic_reference_mentioned_p): Declare.
11033 (nds32_legitimize_ict_address): Declare.
11034 (nds32_expand_ict_move): Declare.
11035 (nds32_indirect_call_referenced_p): Declare.
11036 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
11037 (nds32_relax_group): Use nds32_ict_const_p as condition.
11038 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
11039 (nds32_asm_file_start): Output ict_model directive in asm code.
11040 (nds32_legitimate_address_p): Consider indirect call.
11041 (nds32_print_operand): Consider indirect call.
11042 (nds32_print_operand_address): Consider indirect call.
11043 (nds32_insert_attributes): Handle "indirect_call" attribute.
11044 (TARGET_LEGITIMATE_ADDRESS_P): Define.
11045 (TARGET_LEGITIMATE_CONSTANT_P): Define.
11046 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
11047 (TARGET_DELEGITIMIZE_ADDRESS): Define.
11048 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
11049 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
11050 (TARGET_ICT_MODEL_SMALL): Define.
11051 (TARGET_ICT_MODEL_LARGE): Define.
11052 * config/nds32/nds32.md (movsi): Consider ict model.
11053 (call, call_value): Consider ict model.
11054 (sibcall, sibcall_value): Consider ict model.
11055 * config/nds32/nds32.opt (mict-model): New option.
11056 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
11059 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
11060 Monk Chiang <sh.chiang04@gmail.com>
11061 Jim Wilson <jimw@sifive.com>
11063 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
11064 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
11065 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
11066 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
11067 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
11068 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
11069 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
11070 compute save_libcall_adjustment properly.
11071 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
11072 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
11073 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
11074 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
11075 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
11076 (ABI_SPEC): Handle mabi=ilp32e.
11077 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
11078 (RVE): Add RVE mask.
11079 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
11080 <-march>: Add rv32e as an example.
11082 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
11085 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
11086 (intra_create_variable_infos): Handle C++ constructors.
11088 2018-05-18 Martin Liska <mliska@suse.cz>
11090 * passes.def: Remove a redundant pass.
11092 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
11095 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
11097 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11099 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
11101 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
11103 (armv2, armv2a, armv3, armv3m): Delete architectures.
11104 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
11105 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
11106 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
11108 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
11109 (*mulsidi3adddi): Likewise.
11110 (mulsidi3): Likewise.
11111 (*mulsidi3_nov6): Likewise.
11112 (umulsidi3): Likewise.
11113 (umulsidi3_nov6): Likewise.
11114 (umaddsidi4): Likewise.
11115 (*umulsidi3adddi): Likewise.
11116 (smulsi3_highpart): Likewise.
11117 (*smulsi3_highpart_nov6): Likewise.
11118 (umulsi3_highpart): Likewise.
11119 (*umulsi3_highpart_nov6): Likewise.
11120 * config/arm/arm.h (arm_arch3m): Delete.
11121 * config/arm/arm.c (arm_arch3m): Delete.
11122 (arm_option_override_internal): Update armv3-related comment.
11123 (arm_configure_build_target): Delete use of isa_bit_mode32.
11124 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
11125 (arm_rtx_costs_internal): Delete check of arm_arch3m.
11126 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
11127 (mulsa3): Likewise.
11128 (mulusa3): Likewise.
11129 * config/arm/arm-protos.h (arm_arch3m): Delete.
11130 * config/arm/arm-tables.opt: Regenerate.
11131 * config/arm/arm-tune.md: Likewise.
11132 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
11133 deleted architectures.
11135 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11137 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
11138 (armv5t, armv5te): New features.
11139 (ARMv5, ARMv5e): Delete fgroups.
11140 (ARMv5t, ARMv5te): Adjust for above changes.
11141 (ARMv6m): Likewise.
11142 (armv5, armv5e): Delete arches.
11143 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
11145 (*call_reg_arm): Likewise.
11146 (*call_value_reg_armv5): Likewise.
11147 (*call_value_reg_arm): Likewise.
11148 (*call_symbol): Likewise.
11149 (*call_value_symbol): Likewise.
11150 (*sibcall_insn): Likewise.
11151 (*sibcall_value_insn): Likewise.
11152 (clzsi2): Likewise.
11153 (prefetch): Likewise.
11154 (define_split and define_peephole2 dependent on arm_arch5):
11156 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
11158 (TARGET_ARM_QBIT): Likewise.
11159 (TARGET_DSP_MULTIPLY): Likewise.
11160 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
11161 (arm_arch5, arm_arch5e): Delete.
11162 (arm_arch5t, arm_arch5te): Declare.
11163 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
11164 (arm_arch5t): Declare.
11165 (arm_option_reconfigure_globals): Update for the above.
11166 (arm_options_perform_arch_sanity_checks): Update comment, replace
11167 use of arm_arch5 with arm_arch5t.
11168 (use_return_insn): Likewise.
11169 (arm_emit_call_insn): Likewise.
11170 (output_return_instruction): Likewise.
11171 (arm_final_prescan_insn): Likewise.
11172 (arm_coproc_builtin_available): Likewise.
11173 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
11174 arm_arch5e with arm_arch5t and arm_arch5te.
11175 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
11176 (arm_arch5t, arm_arch5te): Declare.
11177 * config/arm/arm-tables.opt: Regenerate.
11178 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
11179 * config/arm/t-multilib: Likewise.
11180 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
11181 instead of arm_arch5.
11182 (*call_reg_thumb1): Likewise.
11183 (*call_value_reg_thumb1_v5): Likewise.
11184 (*call_value_reg_thumb1): Likewise.
11185 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
11187 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11189 2018-05-18 Martin Liska <mliska@suse.cz>
11191 PR gcov-profile/84846
11192 * doc/gcov.texi: Document -t option of gcov tool.
11194 2018-05-18 Martin Liska <mliska@suse.cz>
11196 PR gcov-profile/84846
11197 * gcov.c (print_usage): Add new -t option.
11198 (process_args): Handle the option.
11199 (generate_results): Use stdout as output when requested by
11202 2018-05-18 Martin Liska <mliska@suse.cz>
11204 PR gcov-profile/84846
11205 * coverage.c (coverage_init): Write PWD to .gcno file.
11206 * doc/gcov.texi: Document how working directory is printed.
11207 * gcov-dump.c (dump_gcov_file): Print PWD.
11208 * gcov.c (output_intermediate_file): Likewise.
11209 (read_graph_file): Read PWD string.
11210 (output_lines): Print PWD.
11212 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11214 PR middle-end/85817
11215 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11216 for retval and return false if all args to phi are zero.
11218 2018-05-18 Richard Biener <rguenther@suse.de>
11220 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11222 (evrp_dom_walker::before_dom_children): Call it.
11224 2018-05-18 Richard Biener <rguenther@suse.de>
11226 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11227 results when processing array refs with variable index.
11229 2018-05-18 Toon Moene <toon@moene.org>
11231 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11232 directly after that of -floop-interchange. Indicate that both
11233 options are enabled by default when specifying -O3.
11235 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11237 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11238 iterator. Delete separate integer-mode vec_set<mode> expander.
11239 (aarch64_simd_vec_setv2di): Delete.
11240 (vec_setv2di): Delete.
11241 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11242 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11243 the "w, r" alternative.
11245 2018-05-18 Martin Liska <mliska@suse.cz>
11247 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11248 * tree-pass.h (make_pass_lower_switch_O0): New function.
11249 * tree-switch-conversion.c (node_has_low_bound): Remove.
11250 (node_has_high_bound): Likewise.
11251 (node_is_bounded): Likewise.
11252 (class pass_lower_switch): Make it a template type and create
11254 (pass_lower_switch::execute): Add template argument.
11255 (make_pass_lower_switch): New function.
11256 (make_pass_lower_switch_O0): New function.
11257 (do_jump_if_equal): Remove.
11258 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11259 all the hard work to tree optimization passes.
11261 2018-05-18 Martin Liska <mliska@suse.cz>
11263 * dbgcnt.c (limit_low): Renamed from limit.
11264 (limit_high): New variable.
11265 (dbg_cnt_is_enabled): Check for upper limit.
11266 (dbg_cnt): Adjust dumping.
11267 (dbg_cnt_set_limit_by_index): Add new argument for high
11269 (dbg_cnt_set_limit_by_name): Likewise.
11270 (dbg_cnt_process_single_pair): Parse new format.
11271 (dbg_cnt_process_opt): Use strtok.
11272 (dbg_cnt_list_all_counters): Remove 'value' and add
11274 * doc/invoke.texi: Document changes.
11276 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11278 * doc/sourcebuild.texi (scalar_all_fma): Document.
11279 * tree.def (FMA_EXPR): Delete.
11280 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11281 * internal-fn.c (ternary_direct): New macro.
11282 (expand_ternary_optab_fn): Likewise.
11283 (direct_ternary_optab_supported_p): Likewise.
11284 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11285 * builtins.c (fold_builtin_fma): Delete.
11286 (fold_builtin_3): Don't call it.
11287 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11288 * expr.c (expand_expr_real_2): Likewise.
11289 * fold-const.c (operand_equal_p): Likewise.
11290 (fold_ternary_loc): Likewise.
11291 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11292 * gimple.c (DEFTREECODE): Likewise.
11293 * gimplify.c (gimplify_expr): Likewise.
11294 * optabs-tree.c (optab_for_tree_code): Likewise.
11295 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11296 * tree-eh.c (operation_could_trap_p): Likewise.
11297 (stmt_could_throw_1_p): Likewise.
11298 * tree-inline.c (estimate_operator_cost): Likewise.
11299 * tree-pretty-print.c (dump_generic_node): Likewise.
11300 (op_code_prio): Likewise.
11301 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11302 * tree-ssa-operands.c (get_expr_operands): Likewise.
11303 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11304 * fold-const-call.h (fold_fma): Delete.
11305 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11306 CFN_FNMA and CFN_FNMS.
11307 (fold_fma): Delete.
11308 * genmatch.c (combined_fn): New enum.
11309 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11310 (commutative_op): New function.
11311 (commutate): Use it. Handle more than 2 operands.
11312 (dt_operand::gen_gimple_expr): Use commutative_op.
11313 (parser::parse_expr): Allow :c to be used with non-binary
11314 operators if the commutative operand is known.
11315 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11316 CFN_FMS, CFN_FNMA and CFN_FNMS.
11317 (backprop::process_assign_use): Remove FMA_EXPR handling.
11318 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11319 (gen_hsa_fma): New function.
11320 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11321 IFN_FNMA and IFN_FNMS.
11322 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11323 * gimple-fold.h (follow_all_ssa_edges): Declare.
11324 * gimple-fold.c (follow_all_ssa_edges): New function.
11325 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11326 gimple_build interface and use follow_all_ssa_edges to fold the result.
11327 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11328 instead of checking for optabs directly.
11329 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11330 rather than FMA_EXPRs.
11331 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11332 call to IFN_FMA instead of an FMA_EXPR.
11334 2018-05-17 Jim Wilson <jimw@sifive.com>
11336 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11337 sign extended promoted subreg, and the range does not have the sign bit
11338 set, then do a sign extend.
11340 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11341 test, check for sign extended subreg and/or constant operands, and
11342 do a sign extend in that case.
11344 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11346 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11347 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11349 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11350 Change logics_shift_reg to logics_shift_imm.
11351 (thunderx2t99_fp_loadpair_basic): Delete.
11352 (thunderx2t99_fp_storepair_basic): Delete.
11353 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11354 (thunderx2t99_asimd_polynomial): Delete.
11355 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11356 and neon_fp_mul_d_scalar_q.
11357 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11358 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11359 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11360 (thunderx2t99_asimd_lut): Add missing tbl types.
11361 (thunderx2t99_asimd_ext): Delete.
11362 (thunderx2t99_asimd_load1_1_mult): Delete.
11363 (thunderx2t99_asimd_load1_2_mult): Delete.
11364 (thunderx2t99_asimd_load1_ldp): New.
11365 (thunderx2t99_asimd_load1): New.
11366 (thunderx2t99_asimd_load2): Add missing *load2* types.
11367 (thunderx2t99_asimd_load3): New.
11368 (thunderx2t99_asimd_load4): New.
11369 (thunderx2t99_asimd_store1_1_mult): Delete.
11370 (thunderx2t99_asimd_store1_2_mult): Delete.
11371 (thunderx2t99_asimd_store2_mult): Delete.
11372 (thunderx2t99_asimd_store2_onelane): Delete.
11373 (thunderx2t99_asimd_store_stp): New.
11374 (thunderx2t99_asimd_store1): New.
11375 (thunderx2t99_asimd_store2): New.
11376 (thunderx2t99_asimd_store3): New.
11377 (thunderx2t99_asimd_store4): New.
11379 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
11381 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
11382 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
11384 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
11385 Segher Boessenkool <segher@kernel.crashing.org>
11388 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
11391 2018-05-17 Richard Biener <rguenther@suse.de>
11393 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11394 for pruning loop and prune defs feeding only already visited PHIs.
11396 2018-05-17 Richard Biener <rguenther@suse.de>
11398 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
11400 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11401 Richard Biener <rguenther@suse.de>
11403 PR tree-optimization/85793
11404 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
11405 for VMAT_ELEMENTWISE.
11407 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11409 * internal-fn.h (lookup_internal_fn): Declare
11410 * internal-fn.c (lookup_internal_fn): New function.
11411 * gimple.c (gimple_build_call_from_tree): Handle calls to
11412 internal functions.
11413 * gimple-pretty-print.c (dump_gimple_call): Print "." before
11414 internal function names.
11415 * tree-pretty-print.c (dump_generic_node): Likewise.
11416 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
11418 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11420 * gimple-fold.h (gimple_build): Make the function forms take
11421 combined_fn rather than built_in_function.
11422 (gimple_simplify): Likewise.
11423 * gimple-match-head.c (gimple_simplify): Likewise.
11424 * gimple-fold.c (gimple_build): Likewise.
11425 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
11426 rather than gimple_build_call_internal.
11427 (get_initial_defs_for_reduction): Likewise.
11428 (vect_create_epilog_for_reduction): Likewise.
11429 (vectorizable_live_operation): Likewise.
11431 2018-05-17 Martin Liska <mliska@suse.cz>
11433 * gimple-ssa-sprintf.c (format_directive): Do not use
11434 space in between 'G_' and '('.
11436 2018-05-17 Jakub Jelinek <jakub@redhat.com>
11439 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
11440 even if the mask is not all ones.
11443 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
11445 (ix86_gimple_fold_builtin): Likewise.
11448 * config/i386/i386.c: Include tree-vector-builder.h.
11449 (ix86_vector_shift_count): New function.
11450 (ix86_fold_builtin): Fold shift builtins by scalar count.
11451 (ix86_gimple_fold_builtin): Likewise.
11453 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
11454 _mm512_setzero): New intrinsics.
11456 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
11457 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11459 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
11460 code generation for cases where splatting a value is not useful.
11461 * simplify-rtx.c (simplify_ternary_operation): Simplify
11462 vec_merge across a vec_duplicate and a paradoxical subreg forming
11463 a vector mode to a vec_concat.
11465 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
11467 * config.gcc: Support "goldmont-plus".
11468 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11470 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11471 PROCESSOR_GOLDMONT_PLUS.
11472 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
11473 (processor_target_table): Add "goldmont-plus".
11474 (PTA_GOLDMONT_PLUS): Define.
11475 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
11476 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
11477 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
11478 (fold_builtin_cpu): Add "goldmont-plus".
11479 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
11480 (ix86_option_override_internal): Add "goldmont-plus".
11481 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
11482 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
11483 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
11484 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
11486 2018-05-17 Richard Biener <rguenther@suse.de>
11488 PR tree-optimization/85757
11489 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
11490 remove defs that only feed that PHI from further processing.
11492 2018-05-16 Jim Wilson <jimw@sifive.com>
11494 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11496 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
11498 2018-05-16 Mark Wielaard <mark@klomp.org>
11500 * dwarf2out.c (count_index_strings): New function.
11501 (output_indirect_strings): Call count_index_strings and generate
11502 header for dwarf_version >= 5.
11504 2018-05-16 Mark Wielaard <mark@klomp.org>
11506 * dwarf2out.c (dwarf_FORM): New function.
11507 (set_indirect_string): Use dwarf_FORM.
11508 (reset_indirect_string): Likewise.
11509 (size_of_die): Likewise.
11510 (value_format): Likewise.
11511 (output_die): Likewise.
11512 (add_skeleton_AT_string): Likewise.
11513 (output_macinfo_op): Likewise.
11514 (index_string): Likewise.
11515 (output_index_string_offset): Likewise.
11516 (output_index_string): Likewise.
11517 (count_index_strings): Likewise.
11519 2018-05-16 Carl Love <cel@us.ibm.com>
11521 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11522 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
11524 2018-05-16 Martin Jambor <mjambor@suse.cz>
11526 * ipa-prop.c (ipa_free_all_edge_args): Remove.
11527 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
11529 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
11531 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
11532 (fnma<mode>4): Likewise.
11533 (fms<mode>4): Likewise.
11534 (fnms<mode>4): Likewise.
11535 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
11536 (aarch64_fnma<mode>4): Likewise.
11537 (aarch64_fms<mode>4): Likewise.
11538 (aarch64_fnms<mode>4): Likewise.
11539 (aarch64_fnmadd<mode>4): Likewise.
11541 2018-05-16 Jason Merrill <jason@redhat.com>
11543 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
11545 2018-05-16 Richard Biener <rguenther@suse.de>
11547 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
11548 (dump_stmt_cost): Declare.
11549 (add_stmt_cost): Dump cost we add.
11550 (add_stmt_costs): New function.
11551 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
11552 No longer exported.
11553 (vect_analyze_stmt): Adjust prototype.
11554 (vectorizable_condition): Likewise.
11555 (vectorizable_live_operation): Likewise.
11556 (vectorizable_reduction): Likewise.
11557 (vectorizable_induction): Likewise.
11558 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
11559 cost vector to pass to vectorizable_ and record afterwards.
11560 (vect_model_reduction_cost): Take cost vector argument and adjust.
11561 (vect_model_induction_cost): Likewise.
11562 (vectorizable_reduction): Likewise.
11563 (vectorizable_induction): Likewise.
11564 (vectorizable_live_operation): Likewise.
11565 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
11566 SLP_TREE_NUMBER_OF_VEC_STMTS.
11567 (vect_analyze_slp_cost_1): Remove.
11568 (vect_analyze_slp_cost): Likewise.
11569 (vect_slp_analyze_node_operations): Take visited args and
11570 a target cost vector. Avoid processing already visited stmt sets.
11571 (vect_slp_analyze_operations): Use a local cost vector to gather
11572 costs and register those of non-discarded instances.
11573 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
11574 (vect_schedule_slp_instance): Remove copying of
11575 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
11577 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
11578 adding cost. Record cost entry location.
11579 (vect_prologue_cost_for_slp_op): Function to compute cost of
11580 a constant or invariant generated for SLP vect in the prologue,
11581 split out from vect_analyze_slp_cost_1.
11582 (vect_model_simple_cost): Make static. Adjust for SLP costing.
11583 (vect_model_promotion_demotion_cost): Likewise.
11584 (vect_model_store_cost): Likewise, make static.
11585 (vect_model_load_cost): Likewise.
11586 (vectorizable_bswap): Add cost vector arg and adjust.
11587 (vectorizable_call): Likewise.
11588 (vectorizable_simd_clone_call): Likewise.
11589 (vectorizable_conversion): Likewise.
11590 (vectorizable_assignment): Likewise.
11591 (vectorizable_shift): Likewise.
11592 (vectorizable_operation): Likewise.
11593 (vectorizable_store): Likewise.
11594 (vectorizable_load): Likewise.
11595 (vectorizable_condition): Likewise.
11596 (vectorizable_comparison): Likewise.
11597 (can_vectorize_live_stmts): Likewise.
11598 (vect_analyze_stmt): Likewise.
11599 (vect_transform_stmt): Adjust calls to vectorizable_*.
11600 * tree-vectorizer.c: Include gimple-pretty-print.h.
11601 (dump_stmt_cost): New function.
11603 2018-05-16 Richard Biener <rguenther@suse.de>
11605 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
11606 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
11607 * tree-ssa-dse.c: Include tree-ssa-loop.h.
11608 (check_name): New callback.
11609 (dse_classify_store): Track cycles via a visited bitmap of PHI
11610 defs and simplify handling of in-loop and across loop dead stores
11611 and properly fail for loop-variant refs. Handle byte-tracking with
11612 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
11615 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
11617 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
11618 (vect_get_mask_type_for_stmt): Likewise.
11619 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
11621 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
11622 to determine the statement's vector type and the vector type that
11623 should be used for calculating nunits. Deal with cases in which
11624 the type has to be deferred.
11625 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
11626 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
11627 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
11628 (vect_determine_vf_for_stmt): New functions, split out from...
11629 (vect_determine_vectorization_factor): ...here.
11630 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
11631 (vect_get_mask_type_for_stmt): New functions, split out from
11632 vect_determine_vectorization_factor.
11634 2018-05-16 Richard Biener <rguenther@suse.de>
11636 * tree-cfg.c (verify_gimple_assign_ternary): Properly
11637 verify the [VEC_]COND_EXPR embedded comparison.
11639 2018-05-15 Martin Sebor <msebor@redhat.com>
11641 PR tree-optimization/85753
11642 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
11643 RECORD_TYPE in addition to ARRAY_TYPE.
11645 2018-05-15 Martin Sebor <msebor@redhat.com>
11647 PR middle-end/85643
11648 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
11650 2018-05-15 Richard Biener <rguenther@suse.de>
11652 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
11653 add by_clobber_p one. Change algorithm to collect all defs
11654 representing uses we need to walk and try reducing them to
11655 a single one before failing.
11656 (dse_dom_walker::dse_optimize_stmt): Adjust.
11658 2018-05-13 Mark Wielaard <mark@klomp.org>
11660 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
11661 (size_of_loc_descr): Likewise.
11662 (output_loc_operands): Likewise.
11663 (output_loc_operands_raw): Likewise.
11664 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
11665 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
11666 (hash_loc_operands): Likewise.
11667 (compare_loc_operands): Likewise.
11669 2018-05-14 Mark Wielaard <mark@klomp.org>
11671 * dwarf2out.c (count_index_addrs): New function.
11672 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
11674 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11676 PR tree-optimization/83648
11677 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
11678 return value as malloc candidate.
11680 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11683 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
11684 param as true in call to suggest_attribute.
11686 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
11688 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
11689 -mreadonly-in-sdata.
11691 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11693 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
11695 (aarch64_crypto_aesd_fused): Likewise.
11697 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
11699 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
11700 (movsi_aarch64): Likewise.
11701 (load_pairsi): Likewise.
11702 (load_pairdi): Likewise.
11703 (store_pairsi): Likewise.
11704 (store_pairdi): Likewise.
11705 (load_pairsf): Likewise.
11706 (load_pairdf): Likewise.
11707 (store_pairsf): Likewise.
11708 (store_pairdf): Likewise.
11709 (zero_extend): Likewise.
11710 (trunc): Swap alternatives.
11711 (fcvt_target): Add '?' to prefer w over r.
11713 2018-05-14 Jakub Jelinek <jakub@redhat.com>
11716 * config/i386/i386.md: Disallow non-commutative arithmetics in
11717 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
11718 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
11719 in the peephole2 before it.
11721 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
11723 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
11724 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
11725 (ix86_handle_option): Handle -mcldemote.
11726 * config.gcc: New header.
11727 * config/i386/cldemoteintrin.h: New file.
11728 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
11729 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11731 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11732 OPTION_MASK_ISA_CLDEMOTE.
11733 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
11734 (ix86_valid_target_attribute_inner_p): Ditto.
11735 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
11736 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
11737 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
11738 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
11739 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
11741 * config/i386/i386.opt: Add -mcldemote.
11742 * config/i386/x86intrin.h: New header.
11743 * doc/invoke.texi: Add -mcldemote.
11745 2018-05-14 Richard Biener <rguenther@suse.de>
11747 * doc/match-and-simplify.texi: Adjust :s documentation.
11749 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
11751 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
11752 intended memcpy size.
11753 (REORDER_45): Likewise.
11755 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
11757 * sort.cc: New file.
11758 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
11759 * vec.c (qsort_chk): Use gcc_qsort.
11760 * Makefile.in (OBJS-libcommon): Add sort.o.
11761 (build/sort.o): New target. Use it...
11762 (BUILD_RTL): ... here, and...
11763 (build/gencfn-macros): ... here, and...
11764 (build/genmatch): ... here.
11766 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11767 Chung-Ju Wu <jasonwucj@gmail.com>
11769 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
11770 * config/nds32/nds32-graywolf.md: New file.
11771 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
11772 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
11774 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
11775 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
11776 * config/nds32/nds32.md (pipeline_model): Add graywolf.
11777 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
11778 * config/nds32/pipelines.md: Include n15 settings.
11780 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11781 Chung-Ju Wu <jasonwucj@gmail.com>
11783 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
11784 * config/nds32/nds32-n13.md: New file.
11785 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
11786 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
11788 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
11789 * config/nds32/nds32.md (pipeline_model): Add n13.
11790 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
11791 * config/nds32/pipelines.md: Include n13 settings.
11793 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11794 Chung-Ju Wu <jasonwucj@gmail.com>
11796 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
11797 * config/nds32/nds32-n10.md: New file.
11798 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
11799 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
11801 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
11802 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
11803 * config/nds32/nds32.md (pipeline_model): Add n10.
11804 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
11805 * config/nds32/pipelines.md: Include n10 settings.
11807 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
11808 Kito Cheng <kito.cheng@gmail.com>
11809 Chung-Ju Wu <jasonwucj@gmail.com>
11811 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
11812 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11813 Add enum values for DSP extension instructions.
11814 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
11816 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
11817 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
11818 New code iterators.
11819 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
11820 * config/nds32/nds32-dspext.md: New file for DSP implementation.
11821 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
11822 * config/nds32/nds32-intrinsic.md: Likewise.
11823 * config/nds32/nds32_intrinsic.h: Likewise.
11824 * config/nds32/nds32-md-auxiliary.c: Likewise.
11825 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
11826 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
11827 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
11828 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
11829 * config/nds32/nds32-protos.h: New declarations for DSP extension.
11830 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
11831 TYPE_DMAC in switch statement.
11832 * config/nds32/nds32.c: New checking and implementation for DSP
11833 extension instructions.
11834 * config/nds32/nds32.h: Likewise.
11835 * config/nds32/nds32.md: Likewise.
11836 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
11837 * config/nds32/predicates.md: Implement new predicates for DSP
11840 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11842 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
11843 Reformat alternatives and attributes so it is easier to identify
11844 which constraints/attributes go with which instruction.
11845 (mov<mode>_hardfloat32, FMOVE64): Likewise.
11846 (mov<mode>_softfloat32, FMOVE64): Likewise.
11847 (mov<mode>_hardfloat64, FMOVE64): Likewise.
11848 (mov<mode>_softfloat64, FMOVE64): Likewise.
11850 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
11852 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11854 (Basic PowerPC Built-in Functions): The new name of the
11855 subsection previously known as "PowerPC Built-in Functions".
11856 (Basic PowerPC Built-in Functions Available on all Configurations):
11858 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
11859 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
11860 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
11861 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
11863 2018-05-11 Martin Jambor <mjambor@suse.cz>
11866 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
11869 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11872 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
11874 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
11876 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
11877 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
11878 (ix86_handle_option): Handle -mwaitpkg.
11879 * config.gcc: New header.
11880 * config/i386/cpuid.h (bit_WAITPKG): New bit.
11881 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
11882 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
11884 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11885 OPTION_MASK_ISA_WAITPKG.
11886 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
11887 (ix86_option_override_internal): Add PTA_WAITPKG.
11888 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
11889 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
11890 IX86_BUILTIN_TPAUSE.
11891 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
11892 __builtin_ia32_umwait and __builtin_ia32_tpause.
11893 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
11894 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
11895 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
11896 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
11897 UNSPECV_TPAUSE): New.
11898 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
11899 * config/i386/i386.opt: Add -mwaitpkg.
11900 * config/i386/waitpkgintrin.h: New file.
11901 * config/i386/x86intrin.h: New header.
11902 * doc/invoke.texi: Add -mwaitpkg.
11904 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11907 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
11909 (cortex-m0): Use armv6s-m isa.
11910 (cortex-m0plus): Likewise.
11911 (cortex-m1): Likewise.
11912 (cortex-m0.small-multiply): Likewise.
11913 (cortex-m0plus.small-multiply): Likewise.
11914 (cortex-m1.small-multiply): Likewise.
11916 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
11917 Jakub Jelinek <jakub@redhat.com>
11919 PR tree-optimization/85692
11920 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
11921 source permute as well.
11923 2018-05-11 Martin Liska <mliska@suse.cz>
11926 * doc/extend.texi: Document LLVM style format for no_sanitize
11929 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11931 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
11932 mode_supports_vsx_dform_quad to mode_supports_dq_form.
11933 (mode_supports_vsx_dform_quad): Likewise.
11934 (mode_supports_vmx_dform): Move these functions to be next to the
11935 other mode_supports functions.
11936 (mode_supports_dq_form): Likewise.
11937 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
11938 mode_supports_dq_form.
11939 (reg_offset_addressing_ok_p): Likewise.
11940 (offsettable_ok_by_alignment): Likewise.
11941 (rs6000_legitimate_offset_address_p): Likewise.
11942 (legitimate_lo_sum_address_p): Likewise.
11943 (rs6000_legitimize_address): Likewise.
11944 (rs6000_legitimize_reload_address): Likewise.
11945 (rs6000_secondary_reload_inner): Likewise.
11946 (rs6000_preferred_reload_class): Likewise.
11947 (rs6000_output_move_128bit): Likewise.
11949 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11951 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
11952 Generate SImode target register for null target.
11953 <case IX86_BUILTIN_XGETBV>: Ditto.
11954 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
11955 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
11957 2018-05-10 Carl Love <cel@us.ibm.com>
11959 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11960 dcbtt and dcbtstt if operands[2] is 0.
11962 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11965 * config/i386/sse.md (usadv64qi): New expander.
11967 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
11969 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
11970 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
11971 -maltivec=be support.
11972 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
11973 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
11974 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
11975 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
11976 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
11977 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
11978 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
11979 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
11980 altivec_vsumsws): Adjust.
11981 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
11982 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
11983 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
11985 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
11986 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
11987 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
11988 (altivec_lve<VI_char>x): Delete expand.
11989 (*altivec_lve<VI_char>x_internal): Rename to...
11990 (altivec_lve<VI_char>x): ... this.
11991 (altivec_lvxl_<mode>): Delete expand.
11992 (*altivec_lvxl_<mode>_internal): Rename to ...
11993 (altivec_lvxl_<mode>): ... this.
11994 (altivec_stvxl_<mode>): Delete expand.
11995 (*altivec_stvxl_<mode>_internal): Rename to ...
11996 (altivec_stvxl_<mode>): ... this.
11997 (altivec_stve<VI_char>x): Delete expand.
11998 (*altivec_stve<VI_char>x_internal): Rename to ...
11999 (altivec_stve<VI_char>x): ... this.
12000 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
12001 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
12002 reduc_plus_scal_<mode>): Adjust.
12003 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
12005 (rs6000_cpu_cpp_builtins): Adjust.
12006 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
12007 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
12008 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
12009 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
12010 -maltivec=be support.
12011 (rs6000_split_vec_extract_var): Adjust.
12012 (rs6000_split_v4si_init): Adjust.
12013 (swap_selector_for_mode): Delete.
12014 (altivec_expand_lvx_be, altivec_expand_stvx_be,
12015 altivec_expand_stvex_be): Delete.
12016 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
12017 -maltivec=be support.
12018 (rs6000_gimple_fold_builtin): Ditto.
12019 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
12021 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
12022 (TARGET_DIRECT_MOVE_64BIT): Adjust.
12023 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
12024 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
12025 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
12026 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
12027 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
12028 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
12029 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
12030 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
12031 anonymous split): Adjust.
12032 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
12033 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
12035 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
12037 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
12038 when --with-gxx-include-dir is also specified.
12039 * configure: Regenerate.
12041 2018-05-09 Jim Wilson <jimw@sifive.com>
12044 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
12045 * config/riscv/t-withmultilib: New.
12046 * config/riscv/withmultilib.h: New.
12047 * doc/install.texi: Document RISC-V --with-multilib-list support.
12049 2018-05-09 Richard Biener <rguenther@suse.de>
12051 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
12053 (vect_bb_vectorization_profitable_p): Adjust. Compute
12054 actual scalar cost using the cost vector and the add_stmt_cost
12057 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12059 PR rtl-optimization/85645
12060 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
12061 in the REG_CFA_REGISTER note for LR, don't leave it empty.
12063 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12065 PR rtl-optimization/85645
12066 * shrink-wrap.c (spread_components): Return a boolean saying if
12067 anything was changed.
12068 (try_shrink_wrapping_separate): Iterate spread_components until
12069 nothing changes anymore.
12071 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12073 PR rtl-optimization/85645
12074 * regrename.c (build_def_use): Also kill the chains that include the
12075 destination of a REG_CFA_REGISTER note.
12077 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12079 PR rtl-optimization/85645
12080 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
12081 insn that has a REG_CFA_REGISTER note.
12083 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
12085 * cfgexpand.c (expand_clobber): New function.
12086 (expand_gimple_stmt_1): Use it.
12087 * tree-vect-stmts.c (vect_clobber_variable): New function,
12089 (vectorizable_simd_clone_call): ...here.
12090 (vectorizable_store): Emit a clobber either side of an
12091 IFN_STORE_LANES sequence.
12092 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
12094 2018-05-09 Tom de Vries <tom@codesourcery.com>
12097 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
12098 (define_insn "trap_if_false"): Add exit after trap.
12100 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
12102 PR rtl-optimization/85638
12103 * bb-reorder.c: Include common/common-target.h.
12104 (create_forwarder_block): New function extracted from...
12105 (fix_up_crossing_landing_pad): ...here. Rename into...
12106 (dw2_fix_up_crossing_landing_pad): ...this.
12107 (sjlj_fix_up_crossing_landing_pad): New function.
12108 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
12109 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
12110 from both partitions and exit the loop after one iteration.
12112 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12115 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12117 (Basic PowerPC Built-in Functions): The new name of the
12118 subsection previously known as "PowerPC Built-in Functions".
12119 (Basic PowerPC Built-in Functions Available on all Configurations):
12121 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12123 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12125 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12127 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12130 2018-05-08 Jim Wilson <jimw@sifive.com>
12132 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
12133 (LD_EMUL_SUFFIX): New.
12134 (LINK_SPEC): Use it.
12136 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12138 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12140 (Basic PowerPC Built-in Functions): The new name of the
12141 subsection previously known as "PowerPC Built-in Functions".
12142 (Basic PowerPC Built-in Functions Available on all Configurations):
12144 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12146 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12148 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12150 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12153 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12156 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
12157 after cmpelim optimization.
12159 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
12161 * config.gcc: Support "goldmont".
12162 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
12163 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12164 PROCESSOR_GOLDMONT.
12165 * config/i386/i386.c (m_GOLDMONT): Define.
12166 (processor_target_table): Add "goldmont".
12167 (PTA_GOLDMONT): Define.
12168 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
12169 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
12170 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
12171 (fold_builtin_cpu): Add "goldmont".
12172 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
12173 (ix86_option_override_internal): Add "goldmont".
12174 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
12175 (processor_type): Add PROCESSOR_GOLDMONT.
12176 * config/i386/i386.md: Add CPU "glm".
12177 * config/i386/glm.md: New file.
12178 * config/i386/x86-tune.def: Add m_GOLDMONT.
12179 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
12181 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12184 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
12186 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
12187 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12188 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12191 * config/i386/i386.c (ix86_fold_builtin): Handle
12192 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12195 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12196 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12198 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12201 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12202 (check_arch): Likewise.
12203 (check_fpu): Return the result rather than printing it.
12204 (end arch): Fix operator precedence.
12205 (end cpu): Likewise.
12206 (END): Print the result from check_fpu.
12208 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12209 Alan Hayward <alan.hayward@arm.com>
12210 David Sherwood <david.sherwood@arm.com>
12212 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12213 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12214 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12215 (*fcmuo<mode>_and): New patterns.
12217 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12219 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12220 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12221 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12222 (cmp_op, sve_imm_con): New code attributes.
12223 (SVE_COND_INT_CMP, imm_con): Delete.
12224 (cmp_op): Remove above unspecs from int attribute.
12225 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12227 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12228 comparison-specific unspecs.
12229 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12230 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12231 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12232 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12233 (*vec_fcm<cmp_op><mode>): Rename to...
12234 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12235 (*vec_fcmuo<mode>): Rename to...
12236 (*fcmuo<mode>): ...this and adjust likewise.
12237 (*pred_fcm<cmp_op><mode>): New pattern.
12238 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12239 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12241 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12243 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12244 (aarch64_emit_sve_predicated_cond): New function.
12245 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12246 (aarch64_emit_unspec_cond_or): Replace with...
12247 (aarch64_emit_sve_or_conds): ...this new function. Use
12248 aarch64_emit_sve_ptrue_op for the individual comparisons and
12249 aarch64_emit_binop to OR them together.
12250 (aarch64_emit_inverted_unspec_cond): Replace with...
12251 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12252 aarch64_emit_sve_ptrue_op for the comparison and
12253 aarch64_emit_unop to invert the result.
12254 (aarch64_expand_sve_vec_cmp_float): Update after the above
12255 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12257 2018-05-07 Nathan Sidwell <nathan@acm.org>
12259 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12260 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12261 (Backwards Compatibility): Likewise.
12263 2018-05-07 Luis Machado <luis.machado@linaro.org>
12267 2018-05-07 Luis Machado <luis.machado@linaro.org>
12269 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12270 <prefetch_dynamic_strides>: New const bool field.
12271 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12272 prefetch_dynamic_strides.
12273 (exynosm1_prefetch_tune): Likewise.
12274 (thunderxt88_prefetch_tune): Likewise.
12275 (thunderx_prefetch_tune): Likewise.
12276 (thunderx2t99_prefetch_tune): Likewise.
12277 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12279 (aarch64_override_options_internal): Update to set
12280 PARAM_PREFETCH_DYNAMIC_STRIDES.
12281 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12282 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12283 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12284 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12285 prefetch-dynamic-strides setting.
12287 2018-05-07 Luis Machado <luis.machado@linaro.org>
12289 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12290 <minimum_stride>: New const int field.
12291 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12292 minimum_stride field.
12293 (exynosm1_prefetch_tune): Likewise.
12294 (thunderxt88_prefetch_tune): Likewise.
12295 (thunderx_prefetch_tune): Likewise.
12296 (thunderx2t99_prefetch_tune): Likewise.
12297 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12298 (aarch64_override_options_internal): Update to set
12299 PARAM_PREFETCH_MINIMUM_STRIDE.
12300 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12301 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12302 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12303 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12304 stride is constant and is below the minimum stride threshold.
12306 2018-05-07 Luis Machado <luis.machado@linaro.org>
12308 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12311 2018-05-07 Luis Machado <luis.machado@linaro.org>
12313 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12314 <prefetch_dynamic_strides>: New const bool field.
12315 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12316 prefetch_dynamic_strides.
12317 (exynosm1_prefetch_tune): Likewise.
12318 (thunderxt88_prefetch_tune): Likewise.
12319 (thunderx_prefetch_tune): Likewise.
12320 (thunderx2t99_prefetch_tune): Likewise.
12321 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12323 (aarch64_override_options_internal): Update to set
12324 PARAM_PREFETCH_DYNAMIC_STRIDES.
12325 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12326 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12327 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12328 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12329 prefetch-dynamic-strides setting.
12331 2018-05-07 Luis Machado <luis.machado@linaro.org>
12333 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12334 <minimum_stride>: New const int field.
12335 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12336 minimum_stride field.
12337 (exynosm1_prefetch_tune): Likewise.
12338 (thunderxt88_prefetch_tune): Likewise.
12339 (thunderx_prefetch_tune): Likewise.
12340 (thunderx2t99_prefetch_tune): Likewise.
12341 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12342 (aarch64_override_options_internal): Update to set
12343 PARAM_PREFETCH_MINIMUM_STRIDE.
12344 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12345 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12346 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12347 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12348 stride is constant and is below the minimum stride threshold.
12350 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12353 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12354 the type is addressable. Don't force op into register if it has
12357 2018-05-05 Roland McGrath <mcgrathr@google.com>
12360 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12361 any section for which we don't know a specific type it should have,
12362 regardless of name. Previously this was done only for the exact
12363 names ".init_array", ".fini_array", and ".preinit_array".
12364 (default_elf_asm_named_section): Add comment about
12365 relationship with default_section_type_flags and SECTION_NOTYPE.
12366 (get_section): Don't consider it a type conflict if one side has
12367 SECTION_NOTYPE and the other doesn't, as long as neither has the
12368 SECTION_BSS et al used in the default_section_type_flags logic.
12370 2018-05-05 Tom de Vries <tom@codesourcery.com>
12373 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12374 (workaround_barsyncs): New function.
12375 (nvptx_reorg): Use workaround_barsyncs.
12376 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12377 (define_expand "nvptx_membar_cta"): New define_expand.
12378 (define_insn "*nvptx_membar_cta"): New insn.
12380 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
12382 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
12383 To improve optimization opportunities.
12384 * builtin-types.def: The new needed builtin types for the above.
12386 2018-05-04 Richard Biener <rguenther@suse.de>
12388 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
12389 * gimple-ssa-store-merging.c
12390 (imm_store_chain_info::output_merged_store): Remove redundant create,
12391 release split_store vector contents on failure.
12392 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
12393 scalar stmt vector on cache hit.
12395 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
12397 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
12399 * config.gcc (powerpc-xilinx-eabi*): Remove.
12400 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
12402 (fusion_addis_mem_combo_load): Ditto.
12403 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
12405 (rs6000_cpu_cpp_builtins): Ditto.
12406 * config/rs6000/rs6000-linux.c
12407 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
12408 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
12409 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
12411 (rs6000_setup_reg_addr_masks): Ditto.
12412 (rs6000_init_hard_regno_mode_ok): Ditto.
12413 (rs6000_option_override_internal): Ditto.
12414 (legitimate_lo_sum_address_p): Ditto.
12415 (rs6000_legitimize_address): Ditto.
12416 (rs6000_legitimize_reload_address): Ditto.
12417 (rs6000_legitimate_address_p): Ditto.
12418 (abi_v4_pass_in_fpr): Ditto.
12419 (setup_incoming_varargs): Ditto.
12420 (rs6000_gimplify_va_arg): Ditto.
12421 (rs6000_split_multireg_move): Ditto.
12422 (rs6000_savres_strategy): Ditto.
12423 (rs6000_emit_prologue_components): Ditto.
12424 (rs6000_emit_epilogue_components): Ditto.
12425 (rs6000_emit_prologue): Ditto.
12426 (rs6000_emit_epilogue): Ditto.
12427 (rs6000_elf_file_end): Ditto.
12428 (rs6000_function_value): Ditto.
12429 (rs6000_libcall_value): Ditto.
12430 * config/rs6000/rs6000.h: Ditto.
12431 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
12432 (TARGET_MINMAX): ... this. New.
12433 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
12434 * config/rs6000/rs6000.md: Remove Xilinx FP support.
12435 (*movsi_internal1_single): Delete.
12436 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
12437 mfpu=, mxilinx-fpu): Delete.
12438 * config/rs6000/singlefp.h: Delete.
12439 * config/rs6000/sysv4.h: Remove Xilinx FP support.
12440 * config/rs6000/t-rs6000: Ditto.
12441 * config/rs6000/t-xilinx: Delete.
12442 * config/rs6000/titan.md: Adjust for fp_type removal.
12443 * config/rs6000/vsx.md: Remove Xilinx FP support.
12444 (VStype_simple): Delete.
12445 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
12446 * config/rs6000/xfpu.h: Delete.
12447 * config/rs6000/xfpu.md: Delete.
12448 * config/rs6000/xilinx.h: Delete.
12449 * config/rs6000/xilinx.opt: Delete.
12450 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
12451 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
12453 2018-05-04 Tom de Vries <tom@codesourcery.com>
12456 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
12459 2018-05-04 Richard Biener <rguenther@suse.de>
12461 PR middle-end/85627
12462 * tree-complex.c (update_complex_assignment): We are always in SSA form.
12463 (expand_complex_div_wide): Likewise.
12464 (expand_complex_operations_1): Likewise.
12465 (expand_complex_libcall): Preserve EH info of the original stmt.
12466 (tree_lower_complex): Handle removed blocks.
12467 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
12468 on complex multiplication and division libcall builtins.
12470 2018-05-04 Richard Biener <rguenther@suse.de>
12472 PR middle-end/85574
12473 * fold-const.c (negate_expr_p): Restrict negation of operand
12474 zero of a division to when we know that can happen without
12476 (fold_negate_expr_1): Likewise.
12478 2018-05-04 Jakub Jelinek <jakub@redhat.com>
12481 * real.h (real_nextafter): Declare.
12482 * real.c (real_nextafter): New function.
12483 * fold-const-call.c (fold_const_nextafter): New function.
12484 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
12485 CASE_CFN_NEXTTOWARD.
12486 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
12487 even when arg1_mode is different from arg0_mode.
12489 2018-05-03 Nathan Sidwell <nathan@acm.org>
12491 * doc/extend.texi (Deprecated Features): Remove
12492 -ffriend-injection.
12493 (Backwards Compatibility): Likewise.
12494 * doc/invoke.texi (C++ Language Options): Likewise.
12495 (C++ Dialect Options): Likewise.
12497 2018-05-03 Jakub Jelinek <jakub@redhat.com>
12500 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
12501 _mm512_mask_mullox_epi64): New intrinsics.
12503 2018-05-03 Tom de Vries <tom@codesourcery.com>
12506 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12507 dump files): Add offload-tree.
12509 2018-05-03 Richard Biener <rguenther@suse.de>
12511 PR tree-optimization/85615
12512 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
12513 to loops not nested in BBs loop father to avoid creating multi-entry
12516 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12518 PR tree-optimization/70291
12519 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
12520 arguments. Change return type to tree. Emit libcall as a new
12521 statement rather than replacing existing one when inplace_p is true.
12522 (expand_complex_multiplication_components): New function.
12523 (expand_complex_multiplication): Expand floating-point complex
12524 multiplication using the above.
12525 (expand_complex_division): Rename inner_type parameter to type.
12526 Update expand_complex_libcall call-site.
12527 (expand_complex_operations_1): Update expand_complex_multiplication
12528 and expand_complex_division call-sites.
12530 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12533 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12534 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12535 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
12536 the highest significant bit of the shift count mask is clear. In
12537 check whether and[sq]i3 is needed verify that all significant bits
12538 of the shift count other than the highest are set.
12540 2018-05-02 Tom de Vries <tom@codesourcery.com>
12543 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
12544 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
12545 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
12546 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
12547 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
12548 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
12549 __builtin_goacc_parlevel_size.
12551 2018-05-02 Richard Biener <rguenther@suse.de>
12553 PR tree-optimization/85597
12554 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
12555 do not use split vect_get_vec_defs call but call vect_get_slp_defs
12558 2018-05-02 Tom de Vries <tom@codesourcery.com>
12561 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12562 dump files): Add ltrans-tree.
12564 2018-05-02 Tom de Vries <tom@codesourcery.com>
12567 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12568 dump files): Add wpa-ipa.
12570 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
12572 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
12573 powerpc*-*-linux*paired* target.
12574 * config/rs6000/750cl.h: Delete.
12575 * config/rs6000/paired.h: Delete.
12576 * config/rs6000/paired.md: Delete.
12577 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
12579 * config/rs6000/rs6000-builtin.def: Remove paired float support.
12580 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
12581 comment. Remove paired float support.
12582 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
12583 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
12585 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
12586 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
12588 * config/rs6000/rs6000.c: Remove paired float support.
12589 (paired_expand_vector_init, paired_expand_vector_move,
12590 paired_emit_vector_compare, paired_emit_vector_cond_expr,
12591 (paired_expand_lv_builtin, paired_expand_stv_builtin,
12592 paired_expand_builtin, paired_expand_predicate_builtin,
12593 paired_init_builtins): Delete.
12594 * config/rs6000/rs6000.h: Remove paired float support.
12595 * config/rs6000/rs6000.md: Remove paired float support.
12596 (move_from_CR_ov_bit): Delete.
12597 * config/rs6000/rs6000.opt (mpaired): Delete.
12598 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
12599 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
12601 2018-05-02 Richard Biener <rguenther@suse.de>
12603 PR middle-end/85567
12604 * gimplify.c (gimplify_save_expr): When in SSA form allow
12605 SAVE_EXPRs to compute to SSA vars.
12607 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12610 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12611 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12612 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
12613 clobber operands[2], instead use a new pseudo. Formatting fixes.
12615 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
12617 PR tree-optimization/85586
12618 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
12619 exit early for statements in the same group if the accesses are
12622 2018-05-02 Tom de Vries <tom@codesourcery.com>
12625 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
12628 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
12630 PR tree-optimization/85143
12631 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
12633 2018-05-01 Tom de Vries <tom@codesourcery.com>
12636 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
12637 not found" error message.
12639 2018-05-01 Tom de Vries <tom@codesourcery.com>
12642 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
12643 * vec.c (test_ordered_remove_if): New function.
12644 (vec_c_tests): Call test_ordered_remove_if.
12645 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
12646 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
12647 * tree-vect-patterns.c (vect_pattern_recog_1): Use
12648 VEC_ORDERED_REMOVE_IF.
12650 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12652 PR tree-optimization/82665
12653 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
12654 pointer subtraction where arguments come from a memchr call.
12656 2018-05-01 Jakub Jelinek <jakub@redhat.com>
12658 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
12659 --push-state --as-needed and --pop-state instead of --as-needed and
12660 --no-as-needed if ld supports it.
12661 * configure: Regenerated.
12664 * doc/install.texi2html: Replace _002d with - and _002a with * in
12665 generated html files using sed.
12667 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12670 * gcc-rich-location.c (blank_line_before_p): New function.
12671 (use_new_line): New function.
12672 (gcc_rich_location::add_fixit_insert_formatted): New function.
12673 * gcc-rich-location.h
12674 (gcc_rich_location::add_fixit_insert_formatted): New function.
12676 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12678 * selftest.c (assert_streq): Rename "expected" and "actual" to
12679 "val1" and "val2". Extend NULL-handling to cover both inputs
12680 symmetrically, while still requiring both to be non-NULL for a pass.
12681 * selftest.h (assert_streq): Rename "expected" and "actual" to
12683 (ASSERT_EQ): Likewise.
12684 (ASSERT_EQ_AT): Likewise.
12685 (ASSERT_KNOWN_EQ): Likewise.
12686 (ASSERT_KNOWN_EQ_AT): Likewise.
12687 (ASSERT_NE): Likewise.
12688 (ASSERT_MAYBE_NE): Likewise.
12689 (ASSERT_MAYBE_NE_AT): Likewise.
12690 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
12691 the assertion to pass.
12692 (ASSERT_STREQ_AT): Likewise.
12694 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12696 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
12697 interaction with -pie.
12699 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12701 * selftest.h: Fix alphabetization of per-source-file selftest
12704 2018-04-30 Jason Merrill <jason@redhat.com>
12706 PR c++/61982 - dead stores to destroyed objects.
12707 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
12710 2018-04-30 Jason Merrill <jason@redhat.com>
12712 * tree.c (build_clobber): New.
12713 * tree.h: Declare it.
12714 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
12716 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12718 * diagnostic-show-locus.c (layout::layout): Update for
12719 location_get_source_line returning a char_span.
12720 (struct char_span): Move to input.h.
12721 (struct correction): Update for fields in char_span becoming
12723 (struct source_line): Update for location_get_source_line
12724 returning a char_span.
12725 (layout::print_line): Likewise.
12726 * edit-context.c (edited_file::print_content): Likewise.
12727 (edited_file::print_diff_hunk): Likewise.
12728 (edited_file::print_run_of_changed_lines): Likewise.
12729 (edited_file::get_num_lines): Likewise.
12730 (edited_line::edited_line): Likewise.
12731 * final.c (asm_show_source): Likewise.
12732 * input.c (location_get_source_line): Convert return type
12733 from const char * to char_span, losing the final "line_len"
12735 (dump_location_info): Update for the above.
12736 (get_substring_ranges_for_loc): Likewise. Use a char_span
12737 when handling the literal within the line.
12738 (test_reading_source_line): Update for location_get_source_line
12739 returning a char_span.
12740 * input.h (class char_span): Move here from
12741 diagnostic-show-locus.c, converting from a struct to a class.
12742 Make data members private.
12743 (char_span::operator bool): New.
12744 (char_span::length): New.
12745 (char_span::get_buffer): New.
12746 (char_span::operator[]): New.
12747 (char_span::subspan): Make const.
12748 (char_span::xstrdup): New.
12749 (location_get_source_line): Convert return type from const char *
12750 to char_span, losing the final "line_size" param.
12752 2018-04-30 Jan Hubicka <jh@suse.cz>
12754 * lto-wrapper.c (ltrans_priorities): New static var.
12755 (cmp_priority): New.
12756 (run_gcc): Read priorities and if doing parallel build order
12757 the Makefile by them.
12759 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12761 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
12763 2018-04-30 Richard Biener <rguenther@suse.de>
12765 * tree-cfg.c (verify_address): Remove base argument, add
12766 flag whether to check TREE_ADDRESSABLE and do that.
12767 (verify_expr): Remove.
12768 (verify_types_in_gimple_reference): Add pieces from verify_expr.
12769 (verify_gimple_assign_single): Likewise.
12770 (verify_gimple_switch): Likewise.
12771 (verify_expr_location_1): Dereference tp once. Add (disabled)
12772 piece from verify_expr.
12773 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
12775 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12777 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
12779 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12781 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
12782 (small_data_pattern): Likewise.
12783 (arc_rewrite_small_data): Likewise.
12784 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
12785 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
12786 (get_symbol_alignment): New function.
12787 (legitimate_small_data_address_p): Likewise.
12788 (legitimate_scaled_address): Update, call
12789 legitimate_small_data_address_p.
12790 (output_sdata): New static variable.
12791 (arc_print_operand): Update how we handle small data operands.
12792 (arc_print_operand_address): Likewise.
12793 (arc_legitimate_address_p): Update, use
12794 legitimate_small_data_address_p.
12795 (arc_rewrite_small_data_p): Remove.
12796 (arc_rewrite_small_data_1): Likewise.
12797 (arc_rewrite_small_data): Likewise.
12798 (small_data_pattern): Likewise.
12799 (compact_sda_memory_operand): Update to use
12800 legitimate_small_data_address_p and get_symbol_alignment.
12801 (prepare_move_operands): Don't rewite sdata pattern.
12802 (prepare_extend_operands): Remove.
12803 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
12805 (zero_extendqisi2): Likewise.
12806 (zero_extendhisi2): Likewise.
12807 (extendqihi2): Likewise.
12808 (extendqisi2): Likewise.
12809 (extendhisi2): Likewise.
12810 (addsi3): Likewise.
12811 (subsi3): Likewise.
12812 (andsi3): Likewise.
12813 * config/arc/constraints.md (Usd): Change it to memory constraint.
12815 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12817 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
12818 as source of std instructions.
12819 * config/arc/arc.md (movsi_insn): Update pattern predicate to
12820 allow 6-bit constants as source for store instructions.
12821 (movdi_insn): Update instruction pattern to allow 6-bit constants
12822 as source for store instructions.
12824 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12826 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
12828 2018-04-30 Nathan Sidwell <nathan@acm.org>
12829 Sandra Loosemore <sandra@codesourcery.com>
12831 * dumpfile.c (dump_open): Allow '-' for stdout.
12832 * doc/invoke.texi (Developer Options): Document dump filename
12833 determination early. Document stdin/stdout selection.
12835 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
12837 Microblaze Target: PIC data text relative
12839 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
12840 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
12842 * config/microblaze/microblaze.h (microblaze_constant_address_p):
12843 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
12844 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
12845 New addressing mode for data-text relative position indepenedent code.
12846 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
12847 'ADDRESS_SYMBOLIC_TXT_REL'.
12848 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
12849 (microblaze_legitimate_pic_operand): Exclude function calls from
12850 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
12851 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
12853 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12854 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12855 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
12856 for 'address + offset'.
12857 (microblaze_expand_prologue): Add new function prologue call for
12859 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
12860 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
12861 table in case of TARGET_PIC_DATA_TEXT_REL.
12862 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
12863 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
12864 Add new macros 'UNSPEC_TEXT',
12865 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
12866 + exclude function calls from 'UNSPEC_PLT' in case of data text
12868 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12869 new target hook for generating address diff vector tables in case of
12871 * doc/tm.texi : Regenerate.
12872 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
12873 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
12874 of addr diff vector generation.
12875 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12876 target hook definition.
12877 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
12878 Add default function for generate_pic_addr_diff_vec -> flag_pic.
12879 * doc/invoke.texi (Add new pic option): Add new microblaze pic
12880 option for data text relative.
12882 2018-04-30 Richard Biener <rguenther@suse.de>
12884 * tree-chrec.h (evolution_function_is_constant_p): Remove
12886 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
12888 2018-04-30 Richard Biener <rguenther@suse.de>
12891 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12893 2018-04-30 Richard Biener <rguenther@suse.de>
12895 PR tree-optimization/28364
12896 PR tree-optimization/85275
12897 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
12898 copying first exit test.
12900 2018-04-28 Mark Wielaard <mark@klomp.org>
12902 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
12903 dwarf_version >= 5.
12904 (dwarf_AT): Handle DW_AT_addr_base.
12905 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
12907 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
12910 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
12911 (*ashl<dwi>3_doubleword_mask_1): Ditto.
12912 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
12913 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
12915 2018-04-28 Richard Biener <rguenther@suse.de>
12917 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
12918 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
12919 to reflect use. Only add interesting stmts.
12921 2018-04-27 Martin Jambor <mjambor@suse.cz>
12924 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
12925 the jump function allows for passing through aggregate values.
12927 2018-04-27 David Malcolm <dmalcolm@redhat.com>
12929 * input.h (in_system_header_at): Convert from macro to inline
12931 (from_macro_expansion_at): Likewise.
12932 (from_macro_definition_at): Likewise.
12934 2018-04-27 Jeff Law <law@redhat.com>
12936 * config.gcc: Mark tile* targets as deprecated/obsolete.
12938 2018-04-27 Richard Biener <rguenther@suse.de>
12940 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
12943 2018-04-27 Richard Biener <rguenther@suse.de>
12945 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
12947 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
12949 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
12950 with Yd constraint. Set "preferred_for_speed" attribute from
12951 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
12952 with Yd constraint.
12953 (*movdi_internal): Ditto.
12954 (movti_interunit splitters): Remove
12955 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
12956 (movdi_interunit splitters): Ditto.
12957 * config/i386/constraints.md (Ye): Remove.
12958 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
12960 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12963 * config/aarch64/constraints.md (Usg): Limit to 31.
12964 (Usj): Limit to 63.
12966 2018-04-27 Jakub Jelinek <jakub@redhat.com>
12968 PR tree-optimization/85529
12969 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
12970 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
12971 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
12972 zero extension or masking of the MSB bit.
12973 (optimize_range_tests): Add FIRST_BB argument, pass it through
12974 to optimize_range_tests_var_bound.
12975 (maybe_optimize_range_tests, reassociate_bb): Adjust
12976 optimize_range_tests callers.
12978 2018-04-26 Richard Biener <rguenther@suse.de>
12979 Jakub Jelinek <jakub@redhat.com>
12981 * cgraph.h (symbol_table): Just declare debug method here.
12982 * symtab.c (symbol_table::debug): Define.
12984 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
12986 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
12988 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
12990 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
12991 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
12992 (*movdi_internal): Substitute Yi and Yj constraint with x
12993 and Ym and Yn constraint with y constraint. Update "isa"
12994 attribute and set "preferred_for_speed" attribute from
12995 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
12996 (*movsi_internal): Ditto.
12997 (*movdf_internal): Ditto.
12998 (*movsf_internal): Ditto.
12999 (*zero_extendsidi2): Ditto.
13000 * config/i386/sse.md (vec_set<mode>_0): Ditto.
13001 (sse2_loadld): Ditto.
13002 (*vec_extract<ssevecmodelower>_0): Ditto.
13003 (*vec_extractv4si_0_zext_sse4): Ditto.
13004 (vec_concatv2di): Ditto.
13005 (*vec_dup<mode>): Ditto.
13006 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
13007 * config/i386/constraints.md (Yi): Remove.
13012 2018-04-26 Nathan Sidwell <nathan@acm.org>
13014 * dumpfile.c (dump_open): New.
13015 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
13016 (dump_finish): Detect stdio/stderr by value not name.
13018 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
13020 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
13022 2018-04-26 Tom de Vries <tom@codesourcery.com>
13025 * config/nvptx/nvptx.c (verify_neutering_jumps)
13026 (verify_neutering_labels): New function
13027 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
13029 2018-04-26 Tom de Vries <tom@codesourcery.com>
13032 * config/nvptx/nvptx.c (needs_neutering_p): New function.
13033 (nvptx_single): Use needs_neutering_p to skip over insns that do not
13036 2018-04-26 Richard Biener <rguenther@suse.de>
13037 Tom de Vries <tom@codesourcery.com>
13040 * lto-streamer-out.c (output_function): Fixup loops if required to match
13041 discovery done in the reader.
13043 2018-04-26 Richard Biener <rguenther@suse.de>
13045 PR tree-optimization/85116
13046 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
13047 have a loop exit from the single latch predecessor. Remove
13048 case of header with just condition.
13049 (ch_base::copy_headers): Exclude infinite loops from any
13051 (pass_ch::execute): Record exits.
13053 2018-04-26 Richard Biener <rguenther@suse.de>
13055 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
13056 prologue cost vector and pass it to vect_get_load_cost.
13057 (vect_get_peeling_costs_all_drs): Likewise.
13058 (vect_peeling_hash_get_lowest_cost): Likewise.
13059 (vect_enhance_data_refs_alignment): Likewise.
13061 2018-04-26 Richard Biener <rguenther@suse.de>
13063 PR middle-end/85450
13064 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
13065 checking of integer<->pointer conversions.
13066 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
13067 sign-/zero-extending pointer types.
13068 (expand_omp_for_static_chunk): Likewise.
13070 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
13071 Jean Lee <xiaoyur347@gmail.com>
13073 * config/mips/mips.c (mips_asan_shadow_offset): New function.
13074 (TARGET_ASAN_SHADOW_OFFSET): Define.
13075 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
13076 true for -fsanitize=address.
13078 2018-04-25 Mark Wielaard <mark@klomp.org>
13080 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
13083 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13085 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
13086 than "alu", remove explicit "memory" and "imm_disp" attributes.
13087 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
13089 PR middle-end/85414
13090 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
13091 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
13092 gen_lowpart_no_emit.
13094 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
13097 * config/i386/i386.c (ix86_expand_builtin): Change memory
13098 operand to XI, extend p0 to Pmode.
13099 * config/i386/i386.md: Change unspec volatile and operand
13100 1 mode to XI, change operand 0 mode to P.
13102 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13104 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
13105 GET_MODE_MASK before any checking.
13106 (nds32_can_use_bset_p): Likewise.
13107 (nds32_can_use_btgl_p): Likewise.
13109 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13111 * config/nds32/nds32-doubleword.md: New define_split pattern for
13112 illegal register number.
13114 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13116 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
13118 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13120 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
13122 2018-04-25 Richard Biener <rguenther@suse.de>
13124 * lto-streamer.h (LTO_major_version): Bump to 8.
13126 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13128 * BASE-VER: Set to 9.0.0.
13130 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
13132 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
13133 in __abskf2 and __powikf2.
13135 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13138 * config/aarch64/constraints.md (Usg, Usj): New constraints.
13139 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
13140 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
13141 Use the above on operand 2. Reindent.
13142 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
13144 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13147 * common/config/i386/i386-common.c (ix86_handle_option): Don't
13149 * config/i386/i386.opt (mcet): Removed.
13150 * doc/install.texi: Remove -mcet documentation.
13151 * doc/invoke.texi: Likewise.
13153 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13156 * doc/install.texi: Remove -mcet from bootstrap-cet.
13158 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13161 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
13162 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
13166 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
13167 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
13168 containing a CONST_VECTOR.
13170 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
13172 * doc/install.texi: Update newlib dependency for nvptx.
13174 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13177 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
13178 instead of INTVAL when shifting x left.
13180 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
13182 PR tree-optimization/85478
13183 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
13184 vect_grouped_store_supported for single element vectors.
13186 2018-04-24 Richard Biener <rguenther@suse.de>
13189 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13190 load cost increase to the case of non-constant step.
13192 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13195 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13196 destination if any_malformed_asm.
13198 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13200 PR middle-end/85496
13201 * expr.c (store_field): In the bitfield case, if the value comes from
13202 a function call and is returned in registers by means of a PARALLEL,
13203 do not change the mode of the temporary unless BLKmode and VOIDmode.
13205 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13207 PR rtl-optimization/85423
13208 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13209 dependencies to debug insns when the previous insn is non-debug.
13211 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13213 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13214 enums into a single definition.
13215 (fls): Fix predicates and printing.
13218 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13220 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13221 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13222 and short u6 immediate.
13223 (check_if_valid_sleep_operand): Remove.
13224 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13226 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13228 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13229 flag_always_save_lp condition.
13230 * config/nds32/nds32.opt (malways-save-lp): New option.
13232 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13234 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13235 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13236 * config/nds32/nds32.h
13237 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13238 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13240 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13242 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13243 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13245 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13246 Chung-Ju Wu <jasonwucj@gmail.com>
13248 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13250 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13251 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13253 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13255 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13257 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13259 * config/nds32/nds32-protos.h (nds32_data_alignment,
13260 nds32_local_alignment): Declare.
13261 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13262 nds32_local_alignment): New functions.
13263 (TARGET_CONSTANT_ALIGNMENT): Define.
13264 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13266 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13268 * config/nds32/nds32.c
13269 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13270 (TARGET_MODES_TIEABLE_P): Likewise.
13272 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13274 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13275 level Ofast and Og.
13277 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13278 Chung-Ju Wu <jasonwucj@gmail.com>
13280 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13281 for unaligned access.
13282 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13284 * config/nds32/nds32-intrinsic.md: Likewise.
13285 * config/nds32/nds32_intrinsic.h: Likewise.
13286 * config/nds32/nds32.h (nds32_builtins): Likewise.
13287 * config/nds32/nds32.opt (munaligned-access): New option.
13288 * config/nds32/nds32.c (nds32_asm_file_start): Display
13289 flag_unaligned_access status.
13291 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13293 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13294 -mno-relax is present.
13295 * config/riscv/linux.h (LINK_SPEC): Ditto.
13297 2018-04-20 Martin Sebor <msebor@redhat.com>
13300 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13302 (gimple_fold_builtin_stxcpy_chk): Same.
13303 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13305 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13308 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13309 __powikf2 when long double is IEEE 128-bit.
13311 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13313 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13314 step to make sure stack always aligned.
13316 2018-04-20 Carl Love <cel@us.ibm.com>
13319 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13320 size check for arg0.
13322 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13323 Tom de Vries <tom@codesourcery.com>
13326 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13327 Emit insns for calls too.
13328 (nvptx_find_par): Always look for worker-level predecessor insn.
13329 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13331 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13332 (nvptx_process_pars): Propagate frames for calls.
13334 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13337 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13339 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13340 (ix86_handle_option): Don't handle OPT_mibt.
13341 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13343 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13345 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13346 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13347 (ix86_target_macros): Define __CET__ with flag_cf_protection
13348 for -fcf-protection.
13349 * config/i386/i386.c (isa2_opts): Remove -mibt.
13350 * config/i386/i386.h (TARGET_IBT): Removed.
13351 (TARGET_IBT_P): Likewise.
13352 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13353 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13354 * config/i386/i386.opt (mcet): Update help message.
13355 (mshstk): Likewise.
13357 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13358 -mcet as an alias for -mshstk.
13360 2018-04-20 Richard Biener <rguenther@suse.de>
13362 PR middle-end/85475
13363 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13364 complexity by forcing a single use of the multiply operand.
13366 2018-04-20 Martin Jambor <mjambor@suse.cz>
13369 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13370 recursion dependency to only apply to non-clones.
13372 2018-04-20 Martin Jambor <mjambor@suse.cz>
13375 * ipa-cp.c (create_specialized_node): Check that clones of
13376 self-recursive edges exist during IPA-CP.
13378 2018-04-19 Toon Moene <toon@moene.org>
13380 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
13383 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13385 PR tree-optimization/85467
13386 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
13387 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
13388 VECTOR_CST element to type.
13390 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13393 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
13394 * config/i386/i386.md (builtin_setjmp_setup): Removed.
13395 (builtin_longjmp): Likewise.
13396 (save_stack_nonlocal): New pattern.
13397 (restore_stack_nonlocal): Likewise.
13399 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13402 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13403 Replace ASM_OUTPUT_LABEL with fprintf.
13405 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13408 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13409 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
13410 * config/i386/i386-c.c (ix86_target_macros_internal): Also
13411 define __IBT__ and __SHSTK__ for -fcf-protection.
13412 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
13414 (ix86_trampoline_init): Likewise.
13415 (x86_output_mi_thunk): Likewise.
13416 (ix86_notrack_prefixed_insn_p): Likewise.
13417 (ix86_option_override_internal): Don't disallow -fcf-protection.
13418 * config/i386/i386.md (rdssp<mode>): Also enable for
13420 (incssp<mode>): Likewise.
13421 (nop_endbr): Likewise.
13422 * config/i386/i386.opt (mcet): Change help message to built-in
13425 (mshstk): Likewise.
13426 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
13427 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
13428 enable CET built-in functions.
13430 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
13432 * common/config/i386/i386-common.c
13433 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
13434 OPTION_MASK_ISA_MOVDIRI_UNSET,
13435 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
13436 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
13437 * config.gcc (movdirintrin.h): New header.
13438 * config/i386/cpuid.h (bit_MOVDIRI,
13439 bit_MOVDIR64B): New bits.
13440 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
13442 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
13443 (VOID, PVOID, PCVOID)): New function types.
13444 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
13445 __builtin_ia32_directstoreu_u64,
13446 __builtin_ia32_movdir64b): New builtins.
13447 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
13448 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
13450 (ix86_valid_target_attribute_inner_p): Ditto.
13451 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
13452 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
13453 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
13454 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
13455 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
13456 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
13457 (movdiri<mode>, movdir64b_<mode>): New.
13458 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
13459 * config/i386/immintrin.h: Include movdirintrin.h.
13460 * config/i386/movdirintrin.h: New file.
13461 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
13463 2018-04-19 Richard Biener <rguenther@suse.de>
13465 PR middle-end/85455
13466 * cfg.c (clear_bb_flags): When loop state says we have
13467 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
13469 2018-04-19 Richard Biener <rguenther@suse.de>
13471 PR tree-optimization/84737
13472 * tree-vect-data-refs.c (vect_copy_ref_info): New function
13473 copying restrict info.
13474 (vect_setup_realignment): Use it.
13475 * tree-vectorizer.h (vect_copy_ref_info): Declare.
13476 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
13477 the first DR to all generated stores.
13478 (vectorizable_load): Likewise for loads.
13480 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13482 PR tree-optimization/85446
13483 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
13484 the integral and pointer types to have the same precision.
13486 * doc/install.texi: Document --disable-cet being the default and
13489 2018-04-18 Martin Liska <mliska@suse.cz>
13491 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
13494 2018-04-18 Martin Liska <mliska@suse.cz>
13497 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13500 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13501 arguments if they are comparable.
13503 2018-04-18 Martin Liska <mliska@suse.cz>
13506 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13509 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13512 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
13515 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
13516 ENDBR after calling __morestack.
13518 2018-04-18 David Malcolm <dmalcolm@redhat.com>
13521 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
13522 by using gcc_base_ver to generate a gcc_driver_version, and use
13523 it when generating GCC_DRIVER_NAME.
13524 * configure: Regenerate.
13526 2018-04-18 Jakub Jelinek <jakub@redhat.com>
13529 * config.gcc: Obsolete powerpc*-*-*spe*.
13531 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13534 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
13535 (stabstr_D): Change type of unum from unsigned int to
13536 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
13539 2018-04-17 Jim Wilson <jimw@sifive.com>
13542 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
13543 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
13544 Set arg_pointer_offset after using pretend_args_size.
13546 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13548 PR rtl-optimization/85431
13549 * dse.c (record_store): Ignore zero width stores.
13552 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
13553 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
13554 __builtin_stack_restore rather than after it.
13555 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
13556 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
13557 argument instead of virtual_dynamic_stack_rtx.
13559 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
13561 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
13563 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
13564 Add note to error message to explain internal mapping of overloaded
13565 built-in function name to non-overloaded built-in function name.
13566 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
13569 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13572 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
13573 where the inputs overlap with the output.
13575 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13577 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
13578 (=v, v) alternative and explicit "memory" attribute.
13579 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
13580 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13582 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
13583 "sselog1" type instead of "sselog".
13584 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
13585 "sselog". Remove explicit "memory" attribute.
13586 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
13587 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13589 (vec_extract_hi_v32hi): Merge all alternatives into one, use
13590 "sselog1" type instead of "sselog". Remove explicit "memory"
13592 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
13593 use "sselog1" type instead of "sselog". Remove explicit "memory"
13595 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
13596 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13598 (vec_extract_hi_v64qi): Merge all alternatives into one, use
13599 "sselog1" type instead of "sselog". Remove explicit "memory"
13601 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
13602 use "sselog1" type instead of "sselog". Remove explicit "memory"
13606 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
13608 PR middle-end/85414
13609 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
13612 2018-04-17 Martin Jambor <mjambor@suse.cz>
13615 * ipa-cp.c (create_specialized_node): Call
13616 expand_all_artificial_thunks if necessary.
13618 2018-04-17 Martin Liska <mliska@suse.cz>
13621 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
13622 in message, remote space in between '_G' and '('.
13624 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13627 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
13628 avx512f_vmcmp<mode>3<round_saeonly_name>,
13629 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
13630 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
13631 avx512f_rndscale<mode><round_saeonly_name>,
13632 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
13633 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13634 Use %<iptr>2 instead of %2 for -masm=intel.
13635 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
13636 avx512f_vcvttss2usi<round_saeonly_name>,
13637 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
13639 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
13640 avx512f_vcvttsd2usi<round_saeonly_name>,
13641 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
13642 Use %q1 instead of %1 for -masm=intel.
13643 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
13644 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
13645 of %3 for -masm=intel.
13646 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
13648 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
13650 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
13652 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
13653 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
13655 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
13657 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
13658 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
13659 %g1 and one with %0 and %1.
13660 (avx512er_vmrcp28<mode><round_saeonly_name>,
13661 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
13662 %1 for -masm=intel.
13663 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
13664 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
13665 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
13666 of %0 and %{%4%} for -masm=intel.
13667 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
13668 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
13669 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
13670 order of %0 and %{%5%}%{z%} for -masm=intel.
13672 2018-04-17 Jan Hubicka <jh@suse.cz>
13675 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
13677 2018-04-17 Martin Liska <mliska@suse.cz>
13680 * multiple_target.c (create_dispatcher_calls): Set apostrophes
13681 for target_clone error message. Make default implementation
13682 clone to be a local declaration.
13683 (separate_attrs): Add new argument and check for an empty
13685 (expand_target_clones): Handle it.
13686 (ipa_target_clone): Make redirection just for target_clones
13689 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
13690 Tom de Vries <tom@codesourcery.com>
13692 PR middle-end/84955
13693 * omp-expand.c (expand_oacc_for): Add dummy false branch for
13694 tiled basic blocks without omp continue statements.
13696 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
13699 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
13700 vec_extract expression as having side effects to make sure it gets
13703 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
13706 * config/i386/i386.c (get_builtin_code_for_version): Check
13709 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
13712 * config.gcc: Support "skylake".
13713 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13715 * config/i386/i386.c (m_SKYLAKE): Define.
13716 (processor_target_table): Add "skylake".
13717 (ix86_option_override_internal): Add "skylake".
13718 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
13719 PROCESSOR_CANNONLAKE.
13720 (get_builtin_code_for_version): Fix priority for
13721 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
13722 PROCESSOR_SKYLAKE-AVX512.
13723 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
13724 (processor_type): Add PROCESSOR_SKYLAKE.
13726 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
13727 Jason Merrill <jason@redhat.com>
13730 * convert.c (convert_to_integer_1): Use direct recursion for
13731 enumeral types and types with a precision less than the number
13732 of bits in their mode.
13734 2018-04-16 Julia Koval <julia.koval@intel.com>
13737 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
13738 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
13740 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
13743 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
13744 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
13745 and -mno-direct-move.
13747 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
13750 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
13751 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
13752 Ensure negative shifts result in {0}.
13754 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
13756 PR rtl-optimization/79916
13757 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
13758 regs (if any) to define how to gnerate SD moves when LRA is in
13761 2018-04-13 Jakub Jelinek <jakub@redhat.com>
13763 PR rtl-optimization/85393
13764 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
13765 * except.c (expand_dw2_landing_pad_for_region): Make static.
13766 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
13767 a label and unconditional jump to old_bb, rather than
13768 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
13771 PR rtl-optimization/85376
13772 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
13773 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
13774 instead of a specific value.
13776 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13777 Bin Cheng <bin.cheng@arm.com>
13779 PR tree-optimization/82965
13780 PR tree-optimization/83991
13781 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
13782 by_profile_only parameter.
13783 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
13784 information if the loop was predicted to iterate too many times.
13785 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
13787 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13790 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
13793 2018-04-13 Martin Liska <mliska@suse.cz>
13794 Jakub Jelinek <jakub@redhat.com>
13796 PR middle-end/81657
13797 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
13798 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
13799 * builtins.c (expand_builtin_memory_copy_args): Use
13800 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
13801 handle dest_addr == pc_rtx.
13803 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
13806 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
13807 asked to not generate direct moves.
13808 (fix_trunc<mode>si2_stfiwx): Similar.
13809 (fix_trunc<mode>si2_internal): Similar.
13811 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13814 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
13815 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
13816 lookup if dest in some wider mode is known to be const0_rtx and
13817 if so, record permanent equivalence for it to be ZERO_EXTEND of
13818 the narrower mode destination.
13820 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13822 * lto-streamer-out.c (output_function): Revert 259346.
13823 * omp-expand.c (expand_oacc_for): Likewise.
13825 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
13827 PR rtl-optimization/85354
13828 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
13829 * sel-sched.c (sel_global_init): ... here.
13831 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
13834 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
13835 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13836 mode for PE-COFF targets.
13837 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
13838 (i386_pe_asm_lto_end): Likewise.
13839 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
13840 (TARGET_ASM_LTO_END): Likewise.
13841 * config/i386/winnt.c (saved_debug_info_level): New static variable.
13842 (i386_pe_asm_lto_start): New function.
13843 (i386_pe_asm_lto_end): Likewise.
13845 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13846 Richard Biener <rguenther@suse.de>
13848 PR middle-end/84955
13849 * lto-streamer-out.c (output_function): Fix CFG loop state before
13851 * omp-expand.c (expand_oacc_for): Handle calls to internal
13852 functions like regular functions.
13854 2018-04-12 Richard Biener <rguenther@suse.de>
13857 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
13858 for the early LTO debug to properly generate references to it
13859 during DIE emission. Do not re-use that for the skeleton for
13861 (dwarf2out_early_finish): Likewise.
13863 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13866 * config/i386/sse.md
13867 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
13868 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
13869 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
13870 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
13871 and output is a reg, avoid creating invalid lowpart subreg, but
13872 instead split into a 512-bit move. Don't split if not AVX512VL,
13873 input is xmm16+ reg and output is a mem.
13874 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
13875 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
13876 xmm16+ reg and output is a mem.
13878 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13880 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
13881 also for flag_dwarf2_cfi_asm.
13883 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13885 PR rtl-optimization/85342
13886 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
13887 a bool scalar var inside of the loop instead. Don't try to update
13888 recog_data.operand after failed apply_change_group.
13890 2018-04-12 Tom de Vries <tom@codesourcery.com>
13893 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
13894 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13895 array with flexible array member as array without given dimension.
13896 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
13897 argument for undefined param to true.
13899 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
13902 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13903 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
13904 from PowerPC section.
13905 * config/rs6000/sysv4.opt (mcall-): Improve help text.
13906 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
13907 help text that is too long.
13908 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
13909 help text that is too long.
13910 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
13911 help text that is too long.
13913 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
13915 * config/alpha/alpha.md (stack_probe_internal): Rename
13916 from "probe_stack". Update all callers.
13918 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13920 PR rtl-optimization/84566
13921 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
13922 sched_macro_fuse_insns.
13924 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13927 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
13928 (compute_block_dependences): ... from here.
13930 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13932 PR tree-optimization/85331
13933 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
13934 from int to HOST_WIDE_INT.
13936 2018-04-11 Martin Jambor <mjambor@suse.cz>
13939 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
13940 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
13941 not the same as the source val.
13942 (cgraph_edge_brings_value_p): New parameter.
13943 (gather_edges_for_value): Pass destination value to
13944 cgraph_edge_brings_value_p.
13945 (perhaps_add_new_callers): Likewise.
13946 (get_info_about_necessary_edges): Likewise and exclude values brought
13947 only by self-recursive edges.
13948 (create_specialized_node): Redirect only clones of self-calling edges.
13949 (+self_recursive_pass_through_p): New function.
13950 (find_more_scalar_values_for_callers_subset): Use it.
13951 (find_aggregate_values_for_callers_subset): Likewise.
13952 (known_aggs_to_agg_replacement_list): Removed.
13953 (decide_whether_version_node): Re-calculate known constants for all
13954 remaining context clones.
13956 2018-04-11 Richard Biener <rguenther@suse.de>
13959 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
13960 from early DWARF output.
13961 (dwarf2out_early_finish): Output line info unconditionally into
13962 early DWARF and add reference to it.
13964 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13967 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
13968 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
13969 other than V2DFmode using iptr mode attribute.
13970 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
13972 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13974 PR rtl-optimization/84659
13975 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
13977 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13980 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
13982 (output_loc_list): Pass address of a dummy size variable even in the
13983 locview handling loop.
13984 (index_location_lists): Add comment on why skip_loc_list_entry can't
13987 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
13990 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
13993 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
13996 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13997 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
13998 and -mstring-compare-inline-limit.
14000 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14003 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
14004 for stack clash protection in a register whenever we need it to be in
14007 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14009 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
14010 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
14012 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14015 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
14017 (mlong-double-): Ditto.
14018 * config/rs6000/sysv4.opt (msdata=): Ditto.
14019 (mtls-size=): Ditto.
14021 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14023 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14024 erroneous entries for
14025 "vector int vec_ldl (int, long int *)", and
14026 "vector unsigned int vec_ldl (int, unsigned long int *)".
14027 Add comments and entries for
14028 "vector bool char vec_ldl (int, bool char *)",
14029 "vector bool short vec_ldl (int, bool short *)",
14030 "vector bool int vec_ldl (int, bool int *)",
14031 "vector bool long long vec_ldl (int, bool long long *)",
14032 "vector pixel vec_ldl (int, pixel *)",
14033 "vector long long vec_ldl (int, long long *)",
14034 "vector unsigned long long vec_ldl (int, unsigned long long *)".
14035 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
14036 type tree bool_long_long_type_node and correct definition of
14037 bool_V2DI_type_node to make reference to this new type tree.
14038 (rs6000_mangle_type): Replace erroneous reference to
14039 bool_long_type_node with bool_long_long_type_node.
14040 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
14041 comments to emphasize sign distinctions for char and int types and
14042 replace RS6000_BTI_bool_long constant with
14043 RS6000_BTI_bool_long_long constant. Also add comment to restrict
14044 use of RS6000_BTI_pixel.
14045 (bool_long_type_node): Remove this macro definition.
14046 (bool_long_long_type_node): New macro definition
14048 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14050 PR rtl-optimization/85300
14051 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
14052 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
14053 simplify_unary_operation fails.
14055 2018-04-10 Martin Liska <mliska@suse.cz>
14057 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
14058 cgraph_edge and ipa_ref.
14060 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14064 * config/i386/sse.md
14065 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
14066 computation of the VEC_MERGE selector from mask.
14067 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14068 Fix decoding of the VEC_MERGE selector into mask.
14070 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14072 PR tree-optimization/85286
14073 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14075 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14077 * final.c (final_1): Set insn_last_address as well as
14078 insn_current_address.
14080 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14083 * explow.c (emit_stack_probe): Call validize_mem on memory location
14084 before passing it to gen_probe_stack. Create address operand and
14085 legitimize it for the probe_stack_address case.
14087 2018-04-09 Jan Hubicka <jh@suse.cz>
14090 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
14091 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
14092 * tree.c (free_lang_data_in_type): Fix handling of binfos;
14094 (free_lang_data): Rebuild type inheritance graph.
14096 2018-04-09 Martin Sebor <msebor@redhat.com>
14098 * invoke.texi (-finline-small-functions): Mention other optimization
14100 (-findirect-inlining, -fpartial-inlining): Same.
14101 (-finline-functions-called-once): Same.
14102 (-freorder-blocks-and-partition): Same.
14104 2018-04-09 Jan Hubicka <jh@suse.cz>
14107 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
14108 jumps; choose last target that matches the criteria (i.e.
14109 no partition changes for non-crossing jumps).
14110 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
14111 support for redirecting crossing jumps to non-crossing.
14113 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
14115 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
14116 also for naked functions.
14118 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
14120 * config/arc/arc.md (add_shift): New pattern.
14121 (add_shift2): Likewise.
14122 (sub_shift): Likewise.
14123 (sub_shift_cmp0_noout): Likewise.
14124 (compare_si_ashiftsi): Likewise.
14125 (xbfu_cmp0_noout): New combine pattern.
14126 (xbfu_cmp0"): Likewise.
14127 (movsi_set_cc_insn): Place the predicable variant first.
14128 (commutative_binary_cmp0_noout): Remove clobber.
14129 (commutative_binary_cmp0): New pattern.
14130 (noncommutative_binary_cmp0): Likewise.
14131 (noncommutative_binary_cmp0_noout): Likewise.
14132 (noncommutative_binary_comparison_result_used): Removed.
14133 (rsub_cmp0): New pattern.
14134 (rsub_cmp0_noout): Likewise.
14135 (extzvsi): Changed, keep only meaningful variants.
14136 (SQH, SEZ): New iterators.
14137 (SQH_postfix): New mode attribute.
14138 (SEZ_prefix): New code attribute.
14139 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
14140 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
14141 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
14142 of numerical value.
14143 (noncommutative_operator): Check the availability of barrel
14146 2018-04-09 Richard Biener <rguenther@suse.de>
14148 PR tree-optimization/85284
14149 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
14150 Only use the niter constraining form of simple_iv when the exit
14151 is always executed.
14153 2018-04-09 Tom de Vries <tom@codesourcery.com>
14156 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14157 (define_expand "*memory_barrier"): New define_expand.
14158 (define_insn "memory_barrier"): New insn.
14160 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14162 PR rtl-optimization/80463
14163 PR rtl-optimization/83972
14164 PR rtl-optimization/83480
14166 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
14167 correct producer for the insn.
14168 (tidy_control_flow): Fixup seqnos in case of debug insns.
14170 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14172 PR rtl-optimization/83913
14174 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
14175 different sched-times when merging exprs.
14177 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14179 PR rtl-optimization/83962
14181 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
14182 tidy_fallthru_edge and tidy_control_flow.
14184 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14186 PR rtl-optimization/83530
14188 * sel-sched.c (force_next_insn): New global variable.
14189 (remove_insn_for_debug): When force_next_insn is true, also leave only
14190 next insn in the ready list.
14191 (sel_sched_region): When the region wasn't scheduled, make another pass
14192 over it with force_next_insn set to 1.
14194 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14196 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14198 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14199 for interrupt control.
14200 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14201 functions for interrupt control.
14202 * config/nds32/nds32-intrinsic.md: Likewise.
14203 * config/nds32/nds32_intrinsic.h: Likewise.
14204 * config/nds32/nds32.h (nds32_builtins): Likewise.
14206 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14208 * config/nds32/nds32.c (nds32_init_machine_status,
14209 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14210 strict_aligned_p field.
14211 (nds32_expand_to_rtl_hook): New function.
14212 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14213 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14215 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14216 Chung-Ju Wu <jasonwucj@gmail.com>
14218 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14219 * config/nds32/nds32-n7.md: New file.
14220 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14221 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14223 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14224 * config/nds32/nds32.md (pipeline_model): Add n7.
14225 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14226 * config/nds32/pipelines.md: Include n7 settings.
14228 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14229 Chung-Ju Wu <jasonwucj@gmail.com>
14231 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14232 * config/nds32/nds32-e8.md: New file.
14233 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14234 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14236 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14237 * config/nds32/nds32.md (pipeline_model): Add e8.
14238 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14239 * config/nds32/pipelines.md: Include e8 settings.
14241 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14242 Chung-Ju Wu <jasonwucj@gmail.com>
14244 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14245 * config/nds32/nds32-n8.md: New file.
14246 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14247 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14249 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14250 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14251 * config/nds32/nds32.md (pipeline_model): Add n8.
14252 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14253 * config/nds32/pipelines.md: Include n8 settings.
14255 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14256 Chung-Ju Wu <jasonwucj@gmail.com>
14258 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14259 * config/nds32/nds32-n9-2r1w.md: New file.
14260 * config/nds32/nds32-n9-3r2w.md: New file.
14261 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14262 nds32_register_ports): New or modify for cpu n9.
14263 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14265 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14266 * config/nds32/nds32-utils.c: New file.
14267 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14268 TARGET_MUL_SLOW): Define.
14269 * config/nds32/nds32.md (pipeline_model): New attribute.
14270 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14271 New options that support cpu n9.
14272 * config/nds32/pipelines.md: Include n9 settings.
14273 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14275 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14277 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14278 information if necessary.
14279 (output_cond_branch_compare_zero): Likewise.
14280 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14281 (nds32_target_alignment): Refine for alignment.
14282 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14283 (FUNCTION_BOUNDARY): Modify.
14284 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14286 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14288 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14290 * config/nds32/constants.md (unspec_volatile_element): Add values for
14291 TLB operation and data prefetch.
14292 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14293 functions for TLB operation and data prefetch.
14294 * config/nds32/nds32-intrinsic.md: Likewise.
14295 * config/nds32/nds32_intrinsic.h: Likewise.
14296 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14297 (nds32_print_operand): Likewise.
14298 * config/nds32/nds32.h (nds32_builtins): Likewise.
14300 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14301 Andrew Pinski <pinsika@gcc.gnu.org>
14303 PR middle-end/82976
14304 * match.pd: Use constant_boolean_node of correct type instead of
14305 boolean_true_node or boolean_false_node for simplifying
14306 pointer comparisons to zero.
14308 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14310 PR tree-optimization/80021
14311 * tree.c (verify_type_variant): Make error call in verify_variant_match
14312 translatable and remove final full stop.
14314 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14316 * config/nds32/constants.md (unspec_volatile_element): Add
14317 UNSPEC_VOLATILE_EH_RETURN.
14318 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14319 nds32_output_stack_pop): Support dwarf exception handling process.
14320 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14321 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14322 exception handling process.
14323 (nds32_compute_stack_frame): Likewise.
14324 (nds32_return_addr_rtx): Likewise.
14325 (nds32_initial_elimination_offset): Likewise.
14326 (nds32_expand_prologue): Likewise.
14327 (nds32_expand_epilogue): Likewise.
14328 (nds32_dynamic_chain_address): New function.
14329 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14330 exception handling.
14331 (DYNAMIC_CHAIN_ADDRESS): Define.
14332 (EH_RETURN_DATA_REGNO): Define.
14333 (EH_RETURN_STACKADJ_RTX): Define.
14334 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14335 patterns for dwarf exception handling.
14337 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14339 * config/nds32/nds32.h: Clean up obsolete macros.
14341 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14343 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14344 Add enum values for particular instructions.
14345 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14346 particular intrinsic functions.
14347 * config/nds32/nds32-intrinsic.md: Likewise.
14348 * config/nds32/nds32_intrinsic.h: Likewise.
14349 * config/nds32/nds32.h (nds32_builtins): Likewise.
14350 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14351 (btst, ave): New patterns for particular instructions.
14353 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14355 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14356 Add enum values for atomic load/store and memory sync.
14357 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14359 * config/nds32/nds32-intrinsic.md: Likewise.
14360 * config/nds32/nds32_intrinsic.h: Likewise.
14361 * config/nds32/nds32.h (nds32_builtins): Likewise.
14363 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14365 PR tree-optimization/85257
14366 * fold-const.c (native_encode_vector): If not all elts could fit
14367 and off is -1, return 0 rather than offset.
14368 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14369 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14370 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14371 adjust buffer in native_interpret_expr call.
14373 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14375 * config/nds32/constants.md (unspec_volatile_element): Add cache
14376 control enum values.
14377 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
14378 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
14379 * config/nds32/nds32.c (nds32_cctl_names): New.
14380 (nds32_print_operand): Handle cache control register names.
14381 * config/nds32/nds32.h (nds32_builtins): New enum values.
14382 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
14384 * config/nds32/nds32.md (type): Add mmu.
14385 * config/nds32/pipelines.md (simple_insn): Add mmu.
14387 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14389 * config/nds32/nds32.md (type): Remove call.
14390 * config/nds32/pipelines.md (simple_insn): Likewise.
14392 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14394 * config/nds32/constants.md (unspec_volatile_element): Add
14395 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
14396 UNSPEC_VOLATILE_FMFCFG.
14397 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
14398 description for fmfcfg and fmfcsr.
14399 (bdesc_1arg): Add fmtcsr.
14400 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
14401 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
14402 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
14403 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
14404 unspec_fmfcfg): New patterns.
14405 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
14406 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
14407 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
14408 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
14409 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
14410 __nds32__fmfcfg): Define.
14412 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14414 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
14415 intrinsic register names.
14416 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
14417 intrinsic register enum values and macros.
14419 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14421 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
14422 for load/store addressing form.
14423 (nds32_print_operand_address): Likewise.
14425 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
14428 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
14429 based on LABEL_REF. Remove useless assertion.
14430 (pic_address_needs_scratch): Fix formatting.
14431 (sparc_legitimize_pic_address): Minor tweaks.
14432 (sparc_delegitimize_address): Adjust assertion accordingly.
14433 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
14434 into symbolic_operand.
14435 (movsi_high_pic_label_ref): Likewise.
14436 (movsi_lo_sum_pic_label_ref): Likewise.
14437 (movdi_pic_label_ref): Likewise.
14438 (movdi_high_pic_label_ref): Likewise.
14439 (movdi_lo_sum_pic_label_ref): Likewise.
14441 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
14443 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
14444 custom LIB_SPEC setup.
14446 2018-04-06 Ruslan Bukin <br@bsdpad.com>
14447 Kito Cheng <kito.cheng@gmail.com>
14449 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
14450 * config/riscv/freebsd.h: New.
14452 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14454 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
14455 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
14458 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14459 Kito Cheng <kito.cheng@gmail.com>
14461 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
14462 nds32_output_call, nds32_symbol_binds_local_p): New functions.
14463 * config/nds32/nds32-protos.h (nds32_output_call,
14464 nds32_output_return): Declare.
14465 * config/nds32/nds32.md: Refine all the call and return patterns.
14467 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14470 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
14471 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
14473 PR rtl-optimization/84872
14474 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
14475 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
14476 EDGE_CROSSING edge.
14478 2018-04-06 Tamar Christina <tamar.christina@arm.com>
14480 * expr.c (copy_blkmode_to_reg): Revert 254862.
14481 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
14483 2018-04-06 Richard Biener <rguenther@suse.de>
14485 PR middle-end/85244
14486 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
14487 after seeing a component reference with an adjacent field. Treat
14488 refs to arrays at struct end of external decls similar to
14489 refs to unconstrained commons.
14491 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14494 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
14495 look through SAVE_EXPRs with non-side-effects argument. Adjust
14497 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
14500 2018-04-06 Richard Biener <rguenther@suse.de>
14502 PR middle-end/85180
14503 * alias.c (find_base_term): New wrapper around find_base_term
14504 unwinding CSELIB_VAL_PTR changes.
14505 (find_base_term): Do not restore CSELIB_VAL_PTR during the
14508 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14510 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
14512 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
14513 constant definitions.
14514 ("nop"): lr 0,0 -> nopr r0
14515 ("nop_lr0", "nop_lr1"): New insn definitions.
14517 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14519 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
14520 NDS32_V3PUSH_AVAILABLE_P macro.
14522 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
14523 Chung-Ju Wu <jasonwucj@gmail.com>
14525 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
14526 (nds32*-*-*): Add float and fpu_config into supported_defaults.
14527 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
14528 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
14529 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
14530 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
14531 * config/nds32/constraints.md: New constraints and checking for hard
14532 float configuration.
14533 * config/nds32/iterators.md: New mode iterator and attribute for hard
14534 float configuration.
14535 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
14537 * config/nds32/nds32-fpu.md: New file.
14538 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
14539 deal with hard float code generation.
14540 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
14542 (abi_type, float_reg_number): New enum type.
14543 * config/nds32/nds32-predicates.c: New predicates for hard float.
14544 * config/nds32/nds32-protos.h: Declare functions for hard float.
14545 * config/nds32/nds32.c: Implementation for hard float configuration.
14546 * config/nds32/nds32.h: Definitions for hard float configuration.
14547 * config/nds32/nds32.md: Include hard float machine description and
14548 modify patterns for hard float configuration.
14549 * config/nds32/nds32.opt: New options for hard float configuration.
14550 * config/nds32/predicates.md: New predicates for hard float
14553 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14555 * common/config/nds32/nds32-common.c
14556 (nds32_option_optimization_table): Enable -mreleax-hint by default.
14558 2018-04-05 Jakub Jelinek <jakub@redhat.com>
14560 PR middle-end/85195
14561 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
14562 CONSTRUCTOR_ELT (ctor, ...)->value.
14564 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
14567 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
14569 2018-04-05 Tom de Vries <tom@codesourcery.com>
14572 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
14575 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14576 Kito Cheng <kito.cheng@gmail.com>
14578 * config/nds32/constraints.md (U33): Fine-tune checking condition.
14579 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
14580 * config/nds32/nds32.h (nds32_16bit_address_type): Add
14581 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
14583 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14584 Kito Cheng <kito.cheng@gmail.com>
14586 * config/nds32/constraints.md (Ufe): New memory constraint.
14587 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
14588 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
14589 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
14591 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
14592 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
14594 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14596 * config/nds32/nds32.md: Use optimize_size in the condition for
14597 alu-shift instructions.
14599 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14601 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
14603 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14605 * config/nds32/nds32.md (negsi2): Refine pattern.
14607 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14608 Chung-Ju Wu <jasonwucj@gmail.com>
14610 * config/nds32/iterators.md (shift_rotate): New code iterator.
14611 (shift): New code attribute.
14612 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
14613 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
14614 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
14615 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
14616 bit-wise operations.
14617 (andsi3, *andsi3): Ditto.
14618 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
14619 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
14620 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
14621 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
14622 nds32_ior_operand, nds32_xor_operand): New predicates.
14624 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14626 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
14627 (addsi3, subsi3): ... this.
14629 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14631 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
14633 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14635 * config/nds32/nds32.md: Adjust indention.
14637 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14639 * config/nds32/nds32.md (feature): New attribute.
14641 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14643 * config/nds32/nds32.md (subtype): New attribute.
14645 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
14648 * config/arm/arm-builtins.c (arm_expand_builtin): Change
14649 expansion to perform a bitwise AND of the argument followed by a
14650 boolean negation of the result.
14652 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
14654 PR rtl-optimization/84878
14655 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
14656 the basic block. Assert the use reference is not artificial and that
14657 it has an associated insn.
14659 2018-04-04 Michael Matz <matz@suse.de>
14661 * builtins.c (compute_objsize): Pass correct operand
14662 to array_at_struct_end_p.
14664 2018-04-04 Richard Biener <rguenther@suse.de>
14667 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
14668 from contexts for DINFO_LEVEL_TERSE and below.
14670 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14672 * config/nds32/nds32-doubleword.md (move_<mode>): Require
14673 resiter_operand condition.
14674 * config/nds32/nds32.md (*move<mode>): Ditto.
14676 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14677 Monk Chiang <sh.chiang04@gmail.com>
14679 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
14681 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14683 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
14685 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14686 Kito Cheng <kito.cheng@gmail.com>
14688 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
14689 nds32_cond_code_str, output_cond_branch,
14690 output_cond_branch_compare_zero, nds32_expand_cbranch,
14691 nds32_expand_cstore, nds32_expand_movcc,
14692 nds32_output_cbranchsi4_equality_zero,
14693 nds32_output_cbranchsi4_equality_reg,
14694 nds32_output_cbranchsi4_equality_reg_or_const_int,
14695 nds32_output_cbranchsi4_greater_less_zero: New functions.
14696 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
14697 nds32_expand_cstore, nds32_expand_movcc,
14698 nds32_output_cbranchsi4_equality_zero,
14699 nds32_output_cbranchsi4_equality_reg,
14700 nds32_output_cbranchsi4_equality_reg_or_const_int,
14701 nds32_output_cbranchsi4_greater_less_zero): Declare.
14702 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
14703 nds32_rimm11s_operand): New predicates.
14704 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
14705 * config/nds32/nds32.md: Rewrite all the branch and conditional move
14708 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14710 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
14711 * config/nds32/nds32.md: Ditto.
14712 * config/nds32/pipelines.md: Ditto.
14714 2018-04-04 Richard Biener <rguenther@suse.de>
14716 PR tree-optimization/85168
14717 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
14718 propagating abnormals.
14720 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14722 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
14724 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14725 Kito Cheng <kito.cheng@gmail.com>
14727 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
14728 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
14729 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
14730 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
14731 * config/nds32/nds32.md (sibcall_internal): New.
14732 (sibcall_register): Remove.
14733 (sibcall_immediate): Remove.
14734 (sibcall_value_internal): New.
14735 (sibcall_value_register): Remove.
14736 (sibcall_value_immediate): Remove.
14737 * config/nds32/predicates.md (nds32_general_register_operand): New.
14738 (nds32_call_address_operand): New.
14740 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14742 PR rtl-optimization/85167
14743 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
14744 bb_defs if *split_p, instead preinitialize it to NULL.
14746 PR tree-optimization/85156
14747 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
14748 evaluating the argument multiple times.
14750 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
14752 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
14754 (_mm_cvtpd_ps): Likewise.
14755 (_mm_cvttpd_epi32): Likewise.
14756 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
14757 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
14758 vector, pixel, and bool following altivec.h include.
14760 2018-04-03 Martin Sebor <msebor@redhat.com>
14762 * doc/extend.texi (Common Function Attributes): Clarify.
14763 (const attribute): Likewise.
14764 (pure attribute): Likewise.
14766 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14769 * config/i386/i386.c (ix86_expand_vector_set): Use
14770 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
14772 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
14774 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
14775 instructions when changing rounding bits to preserve precision bits
14776 in the x87 control word.
14778 2018-04-03 Martin Liska <mliska@suse.cz>
14780 PR tree-optimization/82491
14781 * rtl.h (strip_offset_and_add): Replace += suboffset with
14782 poly_uint64 () + suboffset.
14784 2018-03-29 Martin Liska <mliska@suse.cz>
14785 Martin Jambor <mjambor@suse.cz>
14788 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
14789 param_type is not an integral or pointer type.
14791 2018-04-03 Richard Biener <rguenther@suse.de>
14793 * sese.h (recompute_all_dominators): Remove.
14795 2018-04-02 Martin Sebor <msebor@redhat.com>
14797 * doc/invoke.texi (-Wrestrict): Fix typos.
14799 2018-04-02 Jim Wilson <jimw@sifive.com>
14801 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
14802 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
14803 (<optab>di3, <optab>si3_extend): Likewise.
14804 (<optab>si3_mask, <optab>si3_mask_1): New.
14805 (<optab>di3_mask, <optab>di3_mask_1): New.
14806 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
14807 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
14808 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
14810 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
14812 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
14815 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
14817 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
14818 (nds32_canonicalize_comparison): New function.
14820 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14821 Kito Cheng <kito.cheng@gmail.com>
14822 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14824 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
14825 * config/nds32/constants.md (unspec_volatile_element): Add
14826 UNSPEC_VOLATILE_RELAX_GROUP.
14827 * config/nds32/nds32-relax-opt.c: New file.
14828 * config/nds32/nds32-predicates.c
14829 (nds32_symbol_load_store_p): New function.
14830 * config/nds32/nds32-protos.h
14831 (nds32_symbol_load_store_p): Declare function.
14832 (make_pass_nds32_relax_opt): Declare new rtl pass function.
14833 * config/nds32/nds32.c
14834 (nds32_register_pass): New function to register pass.
14835 (nds32_register_passes): New function to register passes.
14836 * config/nds32/nds32.md (relax_group): New pattern.
14837 * config/nds32/nds32.opt (mrelax-hint): New option.
14838 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
14840 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
14842 * config/nds32/t-nds32: Modify files dependency.
14844 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14846 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
14847 (PROFILE_HOOK): Define its implementation.
14849 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14851 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
14852 type and 32-bit size.
14854 2018-04-01 Jakub Jelinek <jakub@redhat.com>
14856 PR middle-end/85090
14857 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
14858 (V_128_256): New mode iterator.
14859 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
14860 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
14861 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
14863 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
14864 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
14866 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
14869 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
14870 NaN inputs correctly.
14872 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
14875 * config/rs6000/vsx.md (??r): New mode attribute.
14876 (*vsx_mov<mode>_64bit): Use it.
14877 (*vsx_mov<mode>_32bit): Likewise.
14879 2018-03-30 Martin Sebor <msebor@redhat.com>
14881 PR tree-optimization/84818
14882 * builtins.c (check_access): Use warning_n.
14884 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14887 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
14889 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
14892 2018-03-30 Julia Koval <julia.koval@intel.com>
14895 * x86-tune.def (movx, partial_reg_dependency): Enable for
14898 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
14900 PR inline-asm/84985
14901 * lra-constraints.c (process_alt_operands): Move setting
14902 this_alternative_matches below.
14904 2018-03-29 Martin Liska <mliska@suse.cz>
14907 * doc/invoke.texi: Document how LTO works with debug info.
14908 Describe auto-load support of binutils. Mention 'x86-64'
14909 as valid option value of -march option.
14911 2018-03-29 Jakub Jelinek <jakub@redhat.com>
14913 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
14916 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
14917 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
14918 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
14921 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14924 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
14925 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
14926 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
14927 for RS6000_BTM_POWERPC64.
14928 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
14929 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
14930 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
14933 (DIVDEU): Likewise.
14935 2018-03-28 Carl Love <cel@us.ibm.com>
14938 2017-09-27 Carl Love <cel@us.ibm.com>
14940 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14941 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14942 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14945 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14947 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14948 instead of __vector bool.
14949 (_mm_max_pu8): Likewise.
14950 (_mm_min_pi16): Likewise.
14952 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14955 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
14956 (DIVWEUO): Likewise.
14957 (DIVDEO): Likewise.
14958 (DIVDEUO): Likewise.
14959 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
14960 DIVWEUO and DIVDEUO.
14961 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
14962 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
14963 (div_extend): Likewise.
14964 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
14966 (__builtin_divweuo): Likewise.
14967 (__builtin_divdeo): Likewise.
14968 (__builtin_divdeuo): Likewise.
14970 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14973 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
14974 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
14976 PR tree-optimization/82004
14977 * gimple-match-head.c (optimize_pow_to_exp): New function.
14978 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
14979 Don't fold to exp if optimize_pow_to_exp is false.
14981 2018-03-28 Martin Liska <mliska@suse.cz>
14984 * calls.c (initialize_argument_information): Fix trailing space.
14985 * common.opt: Fix typo and provide better explanation for
14986 -fsanitize-coverage option.
14987 * config/i386/i386.opt: Fix typo.
14989 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14990 Martin Liska <mliska@suse.cz>
14993 * gimplify.c (asan_poison_variable): Don't do the check for
14994 gimplify_omp_ctxp here.
14995 (gimplify_decl_expr): Do it here.
14996 (gimplify_target_expr): Likewise.
14998 2018-03-28 Martin Liska <mliska@suse.cz>
15001 * config/i386/i386.c (ix86_function_arg_advance): Do not call
15002 chkp_type_bounds_count if MPX is not enabled.
15004 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
15006 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
15008 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
15011 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15012 function to create the function decl for complex long double
15013 multiply and divide for -mabi=ieeelongdouble.
15014 (init_float128_ieee): Call it.
15016 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
15019 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
15020 -fcf-protection=branch -mibt.
15021 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
15023 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15026 * config/arm/arm.c (arm_valid_symbolic_address): Handle
15027 arm_word_relocations.
15029 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
15032 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
15033 extern array declarations.
15035 2018-03-27 Richard Biener <rguenther@suse.de>
15037 PR middle-end/84067
15038 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
15039 explicit single_use checks.
15041 2018-03-27 Richard Biener <rguenther@suse.de>
15043 PR tree-optimization/85082
15044 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
15047 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15049 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
15050 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
15051 Turn on fasynchronous-unwind-tables and funwind-tables.
15053 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
15056 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
15057 (*bmi_blsr_<mode>_ccz): Ditto.
15059 2018-03-26 Tom de Vries <tom@codesourcery.com>
15061 PR tree-optimization/85063
15062 * omp-general.c (offloading_function_p): New function. Factor out
15064 * omp-offload.c (pass_omp_target_link::gate): ... here.
15065 * omp-general.h (offloading_function_p): Declare.
15066 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
15067 with attribute omp declare target for offloading functions.
15069 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
15071 PR tree-optimization/84005
15072 * tree-data-ref.h (get_base_for_alignment): Declare.
15073 * tree-data-ref.c (get_base_for_alignment_1): New function.
15074 (get_base_for_alignment): Likewise.
15075 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
15076 get_base_for_alignment to find a suitable base object, instead
15077 of always using drb->base_address.
15079 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15081 PR inline-asm/85022
15082 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
15083 known size by default.
15085 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
15087 PR inline-asm/85030
15088 * lra-constraints.c (process_alt_operands): Don't match BLKmode
15089 and non BLKmode operands.
15091 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15094 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
15095 Clean up attributes.
15097 2018-03-23 Richard Biener <rguenther@suse.de>
15100 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
15101 we are going to emit early debug for LTO.
15103 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15105 PR inline-asm/85034
15106 * function.c (match_asm_constraints_1): Don't optimize if input
15107 doesn't satisfy general_operand predicate for output's mode.
15109 PR inline-asm/85022
15110 * alias.c (write_dependence_p): Don't require for x_canonicalized
15111 non-VOIDmode if x has VOIDmode.
15114 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
15115 just don't try to optimize it rather than assert it never happens.
15117 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15119 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
15120 macro expansions for definition of ST_INTERNAL_<mode> and
15121 LD_INTERNAL_<mode> builtins.
15122 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
15124 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
15126 (altivec_expand_st_builtin): Likewise.
15127 (altivec_expand_builtin): Remove calls to deleted functions.
15128 (rs6000_address_for_altivec): Delete this function.
15129 * config/rs6000/vector.md: Remove expands for
15130 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
15132 2018-03-22 Sudakshina Das <sudi.das@arm.com>
15135 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
15136 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
15137 re-computing once computed.
15138 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
15139 (arm_init_machine_status): Initialize
15140 machine->static_chain_stack_bytes.
15142 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15145 * doc/extend.texi: Add four new prototypes for vec_ld.
15146 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
15147 definitions for more logical presentation.
15148 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
15149 entries for V1TI variants of __builtin_altivec_ld builtin.
15150 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
15151 handling of V1TI variant of LVX icode pattern.
15152 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
15153 (rs6000_gimple_fold_builtin): Likewise.
15154 (altivec_init_builtins): Add code to define
15155 __builtin_altivec_lvx_v1ti function.
15157 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15159 PR inline-asm/84941
15160 * function.c (match_asm_constraints_1): Don't do the optimization
15161 if input isn't a REG, SUBREG, MEM or constant.
15163 2018-03-22 Tom de Vries <tom@codesourcery.com>
15165 PR tree-optimization/84956
15166 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
15167 bb_has_abnormal_pred.
15169 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15172 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
15173 DECL_INITIAL (decl) to decl at the end.
15174 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
15175 adjust the comment.
15177 2018-03-21 Joseph Myers <joseph@codesourcery.com>
15179 * doc/extend.texi (__builtin_tgmath): Document when complex
15180 integer types are treated as _Complex _Float64.
15182 2018-03-21 Tom de Vries <tom@codesourcery.com>
15184 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
15186 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15188 PR tree-optimization/84960
15189 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15190 if it is ENTRY block, move them into single succ of ENTRY in that case.
15192 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15194 PR tree-optimization/84811
15195 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15196 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15197 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15198 (known_subrange_p): Update accordingly. Cast each value involved
15199 in the size comparison, rather than casting the result of the
15202 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15204 PR tree-optimization/84982
15205 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15206 by flipping the least significant bit rather than all bits from
15207 bitpos to bitpos + bitsize - 1.
15209 2018-03-21 Nathan Sidwell <nathan@acm.org>
15211 * doc/extend.texi (Deprecated Features): Remove mention of
15212 long-deleted deprecations.
15214 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15217 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15218 * configure: Regenerate.
15220 2018-03-21 Tom de Vries <tom@codesourcery.com>
15222 PR tree-optimization/83126
15223 * tree-parloops.c (num_phis): New function.
15224 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15226 2018-03-21 Nathan Sidwell <nathan@acm.org>
15228 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15229 mention anon-struct/union members and trailing attributes.
15231 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15233 PR tree-optimization/84969
15234 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15235 builtin memset partitions if they set different rhs values.
15237 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15239 PR rtl-optimization/84989
15240 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15241 VEC_DUPLICATE with scalar result mode.
15243 2018-03-21 Martin Liska <mliska@suse.cz>
15246 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15247 not intended return statement.
15249 2018-03-21 Martin Liska <mliska@suse.cz>
15252 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15253 (chkp_find_bound_slots_1): Limit number of iterations.
15255 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15258 * Minor grammar fixes for x86 options.
15260 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15263 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15264 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15266 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15269 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15270 (altivec_lvx_<mode>_1op): Likewise.
15271 (altivec_stvx_<mode>_2op): Likewise.
15272 (altivec_stvx_<mode>_1op): Likewise.
15273 (altivec_lvx_<VM2:mode>): New define_expand.
15274 (altivec_stvx_<VM2:mode>): Likewise.
15275 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15276 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15277 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15278 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15279 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15280 (rs6000_gen_lvx): Likewise.
15281 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15282 (altivec_expand_stv_builtin): Likewise.
15283 (altivec_expand_builtin): Likewise.
15284 * config/rs6000/vector.md: Likewise.
15286 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15289 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15292 2018-03-20 Richard Biener <rguenther@suse.de>
15295 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15296 sign-conversions as zero, fall back to standard scalar_stmt
15299 2018-03-20 Martin Liska <mliska@suse.cz>
15302 * predict.c (rebuild_frequencies): Handle case when we have
15303 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15305 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15308 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15309 flag_section_anchors.
15310 * varasm.c (use_blocks_for_decl_p): Remove hack for
15311 dw2_force_const_mem.
15314 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15316 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15317 be created, use lowpart_subreg of operands[0] rather than operands[0]
15319 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15320 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15321 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15322 and n constraint instead of aarch64_shift_imm_di and Usd.
15323 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15324 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15326 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15329 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15330 to favor GPR over NEON registers.
15331 (<shift>di3_neon): Likewise.
15333 2018-03-20 Tom de Vries <tom@codesourcery.com>
15336 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15337 (nvptx_process_pars): Emit bar.sync asap and alap.
15339 2018-03-20 Tom de Vries <tom@codesourcery.com>
15342 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15343 seen_label if seen_label is already set.
15345 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15348 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15349 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15350 Use 1U instead of 1. Formatting fixes.
15353 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15354 instead of TREE_TYPE (s1) for the return value.
15356 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15358 PR tree-optimization/84946
15359 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15360 bitsize + bitsize in poly_uint64 rather than poly_int64.
15363 * dwarf2asm.c: Include fold-const.c.
15364 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15365 of decl rather than decl itself.
15367 PR rtl-optimization/84643
15368 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15370 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15373 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15374 calling assemble_variable.
15376 2018-03-19 Sudakshina Das <sudi.das@arm.com>
15379 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
15380 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
15382 2018-03-19 Jim Wilson <jimw@sifive.com>
15385 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
15386 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
15387 (riscv_first_stack_step): Likewise.
15388 (riscv_option_override): Use STACK_BOUNDARY instead of
15389 MIN_STACK_BOUNDARY.
15390 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
15391 MIN_STACK_BOUNDARY.
15392 (BIGGEST_ALIGNMENT): Set to 128.
15393 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
15394 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
15397 2018-03-19 Richard Biener <rguenther@suse.de>
15399 PR tree-optimization/84933
15400 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
15401 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
15403 2018-03-19 Richard Biener <rguenther@suse.de>
15405 PR tree-optimization/84859
15406 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
15407 (cond_if_else_store_replacement): Perform sinking operation on
15408 single-store BBs regardless of MAX_STORES_TO_SINK setting.
15409 Generalize what a BB with a single eligible store is.
15411 2018-03-19 Richard Biener <rguenther@suse.de>
15413 PR tree-optimization/84929
15414 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
15415 chrec_is_positive against non-chrec arg.
15417 2018-03-19 Tamar Christina <tamar.christina@arm.com>
15420 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
15422 2018-03-18 Martin Liska <mliska@suse.cz>
15424 PR rtl-optimization/84635
15425 * regrename.c (build_def_use): Use matches_mode only when
15428 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
15430 PR tree-optimization/84913
15431 * tree-vect-loop.c (vectorizable_reduction): Don't try to
15432 vectorize chains of COND_EXPRs.
15434 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15436 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
15438 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15440 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
15442 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15444 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
15446 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15447 Kito Cheng <kito.cheng@gmail.com>
15449 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
15450 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
15451 (nds32_adjust_reg_alloc_order): New function.
15452 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
15454 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15456 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
15457 nds32_print_operand, nds32_print_operand_address): Use
15458 HOST_WIDE_INT_PRINT_DEC instead.
15460 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15462 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
15464 2018-03-17 Jakub Jelinek <jakub@redhat.com>
15467 * config/i386/i386.c (initial_ix86_tune_features,
15468 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
15469 unsigned long long.
15470 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
15471 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
15472 rather than 1u << ix86_tune. Formatting fix.
15473 (ix86_option_override_internal): Change ix86_arch_mask from
15474 unsigned int to unsigned HOST_WIDE_INT, initialize to
15475 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
15476 (ix86_function_specific_restore): Likewise.
15478 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15481 * postreload.c (reload_combine_recognize_pattern): Perform
15482 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
15483 truncate_int_for_mode the result for the destination's mode.
15486 * hsa-gen.c (mem_type_for_type): Fix comment typo.
15487 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
15489 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15492 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
15495 * lra-assigns.c (lra_split_hard_reg_for): Don't use
15496 regno_allocno_class_array and sorted_pseudos.
15497 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
15498 insns where regno is used.
15500 2018-03-16 Martin Liska <mliska@suse.cz>
15503 * multiple_target.c (create_dispatcher_calls): Redirect
15504 reference in the symbol table.
15506 2018-03-16 Martin Liska <mliska@suse.cz>
15509 * multiple_target.c (create_dispatcher_calls): Redirect also
15512 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15516 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
15517 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
15518 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
15520 2018-03-16 Julia Koval <julia.koval@intel.com>
15522 * doc/invoke.texi (Skylake Server): Add CLWB.
15523 Cannonlake): Remove CLWB.
15525 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15527 PR tree-optimization/84841
15528 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
15530 (FLOAT_ONE_CONST_TYPE): Define.
15531 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
15532 (sort_by_operand_rank): Put entries with higher constant_type last
15533 rather than first to match comments.
15535 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
15537 * config/nios2/nios2.md (movsi_internal): Fix thinko in
15540 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15543 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
15544 check and use address of target always.
15546 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
15549 * config/i386/i386.c (indirect_thunk_needed): Update comments.
15550 (indirect_thunk_bnd_needed): Likewise.
15551 (indirect_thunks_used): Likewise.
15552 (indirect_thunks_bnd_used): Likewise.
15553 (indirect_return_needed): New.
15554 (indirect_return_bnd_needed): Likewise.
15555 (output_indirect_thunk_function): Add a bool argument for
15557 (output_indirect_thunk_function): Don't generate alias for
15558 function return thunk.
15559 (ix86_code_end): Call output_indirect_thunk_function to generate
15560 function return thunks.
15561 (ix86_output_function_return): Set indirect_return_bnd_needed
15562 and indirect_return_needed instead of indirect_thunk_bnd_needed
15563 and indirect_thunk_needed.
15565 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
15567 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
15568 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
15569 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
15571 2018-03-15 David Malcolm <dmalcolm@redhat.com>
15572 Paul Hua <paul.hua.gm@gmail.com>
15575 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
15577 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
15579 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
15580 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
15581 resp. SFmode cases.
15583 2018-03-15 Tamar Christina <tamar.christina@arm.com>
15586 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
15587 instead of GET_MODE_SIZE when comparing Units.
15589 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
15592 * varasm.c (hash_section): Return an unchangeble hash value
15593 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
15594 Return !aarch64_can_use_per_function_literal_pools_p ().
15596 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15599 * optabs.c (emit_conditional_move): Pass address of cmode's copy
15600 rather than address of cmode as last argument to prepare_cmp_insn.
15602 2018-03-15 Julia Koval <julia.koval@intel.com>
15604 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
15605 F_AVX512VNNI, F_AVX512BITALG): New.
15607 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
15610 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
15611 insn for floating-point loads and stores.
15613 2018-03-14 Carl Love <cel@us.ibm.com>
15615 * config/rs6000/rs6000-c.c: Add macro definitions for
15616 ALTIVEC_BUILTIN_VEC_PERMXOR.
15617 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
15618 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
15619 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
15620 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
15622 * config/doc/extend.texi: Add prototypes for vec_permxor.
15624 2018-03-14 David Malcolm <dmalcolm@redhat.com>
15627 * diagnostic-show-locus.c (class layout_point): Convert m_line
15628 from int to linenum_type.
15629 (line_span::comparator): Use linenum "compare" function when
15630 comparing line numbers.
15631 (test_line_span): New function.
15632 (layout_range::contains_point): Convert param "row" from int to
15634 (layout_range::intersects_line_p): Likewise.
15635 (layout::will_show_line_p): Likewise.
15636 (layout::print_source_line): Likewise.
15637 (layout::should_print_annotation_line_p): Likewise.
15638 (layout::print_annotation_line): Likewise.
15639 (layout::print_leading_fixits): Likewise.
15640 (layout::annotation_line_showed_range_p): Likewise.
15641 (struct line_corrections): Likewise for field m_row.
15642 (line_corrections::line_corrections): Likewise for param "row".
15643 (layout::print_trailing_fixits): Likewise.
15644 (layout::get_state_at_point): Likewise.
15645 (layout::get_x_bound_for_row): Likewise.
15646 (layout::print_line): Likewise.
15647 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
15648 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
15649 * input.c (selftest::test_linenum_comparisons): New function.
15650 (selftest::input_c_tests): Call it.
15651 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
15652 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
15653 * selftest.h (ASSERT_GT): New macro.
15654 (ASSERT_GT_AT): New macro.
15655 (ASSERT_LT): New macro.
15656 (ASSERT_LT_AT): New macro.
15658 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
15660 PR rtl-optimization/84780
15661 * combine.c (distribute_links): Don't make a link based on pc_rtx.
15663 2018-03-14 Martin Liska <mliska@suse.cz>
15665 * tree.c (record_node_allocation_statistics): Use
15666 get_stats_node_kind.
15667 (get_stats_node_kind): New function extracted from
15668 record_node_allocation_statistics.
15669 (free_node): Use get_stats_node_kind.
15671 2018-03-14 Richard Biener <rguenther@suse.de>
15673 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
15674 that the value-set of ANTIC_IN doesn't grow.
15677 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15679 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15680 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15682 2018-03-14 Julia Koval <julia.koval@intel.com>
15684 * config.gcc (icelake-client, icelake-server): New.
15686 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
15687 (initial_ix86_arch_features): Ditto.
15688 (PTA_SKYLAKE): Add SGX.
15689 (PTA_ICELAKE): Remove.
15690 (PTA_ICELAKE_CLIENT): New.
15691 (PTA_ICELAKE_SERVER): New.
15692 (ix86_option_override_internal): Split up icelake on icelake client and
15694 (get_builtin_code_for_version): Ditto.
15695 (fold_builtin_cpu): Ditto.
15696 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
15697 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
15698 * config/i386/i386.h (processor_type): Ditto.
15699 * doc/invoke.texi: Ditto.
15701 2018-03-14 Jakub Jelinek <jakub@redhat.com>
15704 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
15705 INTEGER_CST offset, add it together with bitpos / 8 and
15706 sign extend based on POINTER_SIZE.
15710 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
15713 * config/i386/constraints.md (Yc): New register constraint.
15714 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
15715 Use Yc constraint for alternative 2 of operand 0. Remove
15716 preferred_for_speed attribute.
15718 2018-03-14 Richard Biener <rguenther@suse.de>
15720 PR tree-optimization/84830
15721 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
15722 with the old one to avoid oscillations.
15724 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
15727 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
15729 (assign_by_spills): Return a flag of reload assignment failure.
15730 Do not process the reload assignment failures. Do not spill other
15731 reload pseudos if they has the same reg class. Update n if
15733 (lra_assign): Add a return arg. Set up from the result of
15734 assign_by_spills call.
15735 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15736 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15737 usage_insns if it is not NULL.
15738 (spill_hard_reg_in_range): New function.
15739 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15740 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15741 function prototypes.
15742 (lra_assign): Change prototype.
15743 * lra.c (lra): Add code to deal with fails by splitting hard reg
15746 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
15748 * config/riscv/riscv.opt (mrelax): New option.
15749 * config/riscv/riscv.c (riscv_file_start): Emit ".option
15750 "norelax" when riscv_mrelax is disabled.
15751 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
15753 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15756 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
15757 reassociation for int modes.
15759 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15761 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
15762 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
15764 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
15765 * config/aarch64/aarch64-sve.md
15766 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
15767 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
15768 (*extend<mode><Vwide>2): Rename to...
15769 (aarch64_sve_extend<mode><Vwide>2): ...this.
15770 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
15771 renaming the old pattern to...
15772 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
15774 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
15775 define_expand, renaming the old pattern to...
15776 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
15777 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
15778 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
15779 account when deciding which SVE instruction the optab should use.
15780 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
15782 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15784 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
15785 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
15786 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
15787 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
15788 (tlsdesc_small_<mode>): Turn a define_expand and use
15789 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
15790 (tlsdesc_small_advsimd_<mode>): ...this.
15791 (tlsdesc_small_sve_<mode>): New pattern.
15793 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15795 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
15796 (UNSPEC_UMUL_HIGHPART): New constants.
15797 (MUL_HIGHPART): New int iteraor.
15798 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
15799 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
15801 (*<su>mul<mode>3_highpart): New define_insn.
15803 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15806 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15809 2018-03-13 Martin Liska <mliska@suse.cz>
15812 * (sem_item_optimizer::sem_item_optimizer): Initialize new
15814 (sem_item_optimizer::~sem_item_optimizer): Release it.
15815 (sem_item_optimizer::merge_classes): Register variable aliases.
15816 (sem_item_optimizer::fixup_pt_set): New function.
15817 (sem_item_optimizer::fixup_points_to_sets): Likewise.
15818 * ipa-icf.h: Declare new variables and functions.
15820 2018-03-13 Jakub Jelinek <jakub@redhat.com>
15822 PR middle-end/84834
15823 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
15824 integer_pow2p@2 and test integer_pow2p in condition.
15825 (A < 0 ? C : 0): Similarly for @1.
15827 PR middle-end/84831
15828 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
15829 characters starting at p contain '\0' character, don't look beyond
15833 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
15834 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
15837 * reg-stack.c (change_stack): Change update_end var from int to
15838 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
15839 also call set_block_for_insn on the newly added insns and rescan.
15842 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
15843 on the last operand.
15846 * tree.c (stabilize_reference_1): Return save_expr (e) for
15847 STATEMENT_LIST even if it doesn't have side-effects.
15849 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
15851 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
15853 2018-03-12 Renlin Li <renlin.li@arm.com>
15855 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
15856 aarch64_output_scalar_simd_mov_immediate.
15858 2018-03-12 Martin Sebor <msebor@redhat.com>
15860 PR tree-optimization/83456
15861 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
15862 for perfectly overlapping calls to memcpy.
15863 (gimple_fold_builtin_memory_chk): Same.
15864 (gimple_fold_builtin_strcpy): Handle no-warning.
15865 (gimple_fold_builtin_stxcpy_chk): Same.
15866 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
15868 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15870 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
15871 parameter. Use it for SFmode.
15872 (rs6000_function_arg_advance_1): Adjust.
15873 (rs6000_function_arg): Adjust.
15874 (rs6000_gimplify_va_arg): Pass false for that new parameter.
15876 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15878 PR rtl-optimization/84169
15879 PR rtl-optimization/84780
15880 * combine.c (can_combine_p): Check for a 2-insn combination whether
15881 the destination register is used between the two insns, too.
15883 2018-03-12 Richard Biener <rguenther@suse.de>
15885 PR tree-optimization/84803
15886 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
15887 for refs DR analysis didn't process.
15889 2018-03-12 Richard Biener <rguenther@suse.de>
15891 PR tree-optimization/84777
15892 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
15893 force-vectorize loops ignore whether we are optimizing for size.
15895 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
15897 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
15898 (TARGET_MD_ASM_ADJUST): Define.
15900 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
15901 Kito Cheng <kito.cheng@gmail.com>
15902 Chung-Ju Wu <jasonwucj@gmail.com>
15904 * config/nds32/nds32.c (nds32_compute_stack_frame,
15905 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
15906 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
15907 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
15908 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
15909 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
15910 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
15911 * config/nds32/nds32.md (prologue, epilogue): Use macro
15912 NDS32_V3PUSH_AVAILABLE_P to do checking.
15914 2018-03-11 Jakub Jelinek <jakub@redhat.com>
15917 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
15918 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
15919 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
15920 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
15921 addition of most attributes on !orig_type_die or the attribute not
15922 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
15924 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15925 Chung-Ju Wu <jasonwucj@gmail.com>
15927 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
15928 __NDS32_VH__ macro.
15929 * config/nds32/nds32.opt (mvh): New option.
15931 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15932 Chung-Ju Wu <jasonwucj@gmail.com>
15934 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
15936 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
15937 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
15940 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15941 Chung-Ju Wu <jasonwucj@gmail.com>
15943 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
15945 * config/nds32/nds32-multiple.md (strlensi): New pattern.
15946 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
15948 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
15949 Kito Cheng <kito.cheng@gmail.com>
15950 Chung-Ju Wu <jasonwucj@gmail.com>
15952 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
15953 UNSPEC_FFMISM and UNSPEC_FLMISM.
15954 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
15955 for ffb, ffmism and flmism.
15956 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
15957 (unspec_ffmism): Ditto.
15958 (unspec_flmism): Ditto.
15959 (nds32_expand_builtin_impl): Check if string extension is available.
15960 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
15961 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
15963 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
15966 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15969 * lra-assigns.c (assign_by_spills): Return a flag of reload
15970 assignment failure. Do not process the reload assignment
15971 failures. Do not spill other reload pseudos if they has the same
15973 (lra_assign): Add a return arg. Set up from the result of
15974 assign_by_spills call.
15975 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15976 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15977 usage_insns if it is not NULL.
15978 (spill_hard_reg_in_range): New function.
15979 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15980 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15981 function prototypes.
15982 (lra_assign): Change prototype.
15983 * lra.c (lra): Add code to deal with fails by splitting hard reg
15986 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
15989 * config/i386/i386.opt: Replace Enforcment with Enforcement.
15991 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
15994 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
15995 (dw_val_node): Add val_symbolic_view.
15996 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
15997 (symview_upper_bound): New.
15998 (new_line_info_table): Initialize symviews_since_reset.
15999 (dwarf2out_source_line): Count symviews_since_reset and set
16000 symview_upper_bound.
16001 (dw_val_equal_p): Handle symview.
16002 (add_AT_symview): New.
16003 (print_dw_val): Handle symview.
16004 (attr_checksum, attr_checksum_ordered): Likewise.
16005 (same_dw_val_p, size_of_die): Likewise.
16006 (value_format, output_die): Likewise.
16007 (add_high_low_attributes): Use add_AT_symview for entry_view.
16008 (dwarf2out_finish): Reset symview_upper_bound, clear
16011 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
16014 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
16015 Add strict argument and use it.
16016 (rs6000_split_multireg_move): Update for new strict argument.
16017 (mem_operand_gpr): Disallow all non-offsettable addresses.
16018 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
16020 2018-03-09 Jakub Jelinek <jakub@redhat.com>
16023 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
16024 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
16025 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
16028 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
16029 decl, use remap_type if we want to use the type.
16031 2018-03-09 Martin Sebor <msebor@redhat.com>
16033 PR tree-optimization/84526
16034 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16036 (builtin_access::generic_overlap): Be prepared to handle non-array
16039 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
16041 PR rtl-optimization/84682
16042 * lra-constraints.c (process_address_1): Check is_address flag
16043 for address constraints.
16044 (process_alt_operands): Likewise.
16045 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
16046 preprocess_constraints.
16047 * recog.h (preprocess_constraints): Add oploc parameter.
16049 * recog.c (preprocess_constraints): Test address_operand for
16050 CT_ADDRESS constraints.
16052 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16055 * lra-assigns.c (assign_by_spills): Return a flag of reload
16056 assignment failure. Do not process the reload assignment
16057 failures. Do not spill other reload pseudos if they has the same
16059 (lra_assign): Add a return arg. Set up from the result of
16060 assign_by_spills call.
16061 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16062 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16063 usage_insns if it is not NULL.
16064 (spill_hard_reg_in_range): New function.
16065 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16066 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16067 function prototypes.
16068 (lra_assign): Change prototype.
16069 * lra.c (lra): Add code to deal with fails by splitting hard reg
16072 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16075 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
16076 Accept complain bool parameter. Only emit errors if it is true.
16077 (arm_parse_cpu_option_name): Likewise.
16078 (arm_target_thumb_only): Adjust callers of the above.
16079 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
16080 prototype to take a default true bool parameter.
16081 (arm_parse_arch_option_name): Likewise.
16083 2018-03-09 David Malcolm <dmalcolm@redhat.com>
16084 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
16088 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
16089 * configure: Regenerate.
16090 * configure.ac ("linker --version-script option"): New.
16091 ("linker soname option"): New.
16093 2018-03-09 Richard Biener <rguenther@suse.de>
16095 PR tree-optimization/84775
16096 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
16097 immediate uses of predicate stmts and mark them modified.
16100 PR tree-optimization/84178
16101 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16103 (version_loop_for_if_conversion): Delay update_ssa call.
16104 (tree_if_conversion): Delay update_ssa until after predicate
16107 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
16110 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
16111 when the function accesses prior frames.
16113 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16116 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
16117 gen_llsym, otherwise call maybe_gen_llsym.
16119 PR inline-asm/84742
16120 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
16121 has ',' character inside of it.
16123 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16126 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
16127 as clobbering CC_REGNUM.
16129 2018-03-08 Richard Biener <rguenther@suse.de>
16131 PR middle-end/84552
16132 * tree-scalar-evolution.c: Include tree-into-ssa.h.
16133 (follow_copies_to_constant): Do not follow SSA names registered
16136 2018-03-08 Richard Biener <rguenther@suse.de>
16138 PR tree-optimization/84178
16139 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16141 (version_loop_for_if_conversion): Delay update_ssa call.
16142 (tree_if_conversion): Delay update_ssa until after predicate
16145 2018-03-08 David Malcolm <dmalcolm@redhat.com>
16147 PR tree-optimization/84178
16148 * tree-if-conv.c (release_bb_predicate): Remove the
16149 the assertion that the stmts have NULL use_ops.
16150 Discard the statements, asserting that they haven't
16151 yet been added to a BB.
16153 2018-03-08 Richard Biener <rguenther@suse.de>
16155 PR tree-optimization/84746
16156 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
16157 (phi_translate): Pass in destination ANTIC_OUT set.
16158 (phi_translate_1): Likewise. For a simplified result lookup
16159 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
16160 (phi_translate_set): Adjust.
16161 (do_pre_regular_insertion): Likewise.
16162 (do_pre_partial_partial_insertion): Likewise.
16164 2018-03-08 Martin Liska <mliska@suse.cz>
16166 PR gcov-profile/84735
16167 * doc/gcov.texi: Document usage of profile files.
16168 * gcov-io.h: Document changes in the format.
16170 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
16174 * dwarf2out.c (struct dw_line_info_table): Update comments for
16176 (FORCE_RESET_NEXT_VIEW): New.
16177 (FORCE_RESETTING_VIEW_P): New.
16178 (RESETTING_VIEW_P): Check for -1 too.
16179 (ZERO_VIEW_P): Likewise.
16180 (new_line_info_table): Force-reset next view.
16181 (dwarf2out_begin_function): Likewise.
16182 (dwarf2out_source_line): Simplify zero_view_p initialization.
16183 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
16184 view directly. Omit view when omitting .loc at line 0.
16186 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16188 PR tree-optimization/84740
16189 * tree-switch-conversion.c (process_switch): Call build_constructors
16190 only if info.phi_count is non-zero.
16192 PR tree-optimization/84739
16193 * tree-tailcall.c (find_tail_calls): Check call arguments against
16194 DECL_ARGUMENTS (current_function_decl) rather than
16195 DECL_ARGUMENTS (func) when checking for tail recursion.
16197 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16199 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16200 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16201 Volker Reichelt's entry and add entries for people that perform
16202 GCC fuzzy testing and report numerous bugs.
16204 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16207 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16208 readonly data in sdata, if that is disabled.
16209 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16210 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16211 -mreadonly-in-sdata option.
16213 2018-03-07 Martin Sebor <msebor@redhat.com>
16215 PR tree-optimization/84468
16216 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16217 basic block when looking for nul assignment.
16219 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16222 * except.h (output_function_exception_table): Adjust prototype.
16223 * except.c (output_function_exception_table): Remove FNNAME parameter
16224 and add SECTION parameter. Ouput one part of the table at a time.
16225 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16226 the first part of the exception table and emit unwind directives.
16227 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16228 (i386_pe_seh_cold_init): Likewise.
16229 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16230 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16231 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16232 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16233 * config/i386/winnt.c: Include except.h.
16234 (struct seh_frame_state): Add reg_offset, after_prologue and
16235 in_cold_section fields.
16236 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16237 (i386_pe_seh_cold_init): New function.
16238 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16239 to seh->in_cold_section.
16240 (seh_emit_push): Record the offset of the push.
16241 (seh_emit_save): Record the offet of the save.
16242 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16243 Test seh->after_prologue to disregard the epilogue.
16244 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16245 (i386_pe_end_cold_function): New function.
16247 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16250 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16251 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16254 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16256 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16257 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16259 PR middle-end/84723
16260 * multiple_target.c: Include tree-inline.h and intl.h.
16261 (expand_target_clones): Diagnose and fail if node->definition and
16262 !tree_versionable_function_p (node->decl).
16264 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16266 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16268 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16269 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16270 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16272 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16275 * combine.c (try_combine): Use reg_or_subregno instead of handling
16276 just paradoxical SUBREGs and REGs.
16278 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16280 * config/arc/arc.c (arc_finalize_pic): Remove function.
16281 (arc_must_save_register): We use single base PIC register, remove
16282 checks to save/restore the PIC register.
16283 (arc_expand_prologue): Likewise.
16284 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16286 (arc_verify_short): Likewise.
16287 (arc_attr_type): Likewise.
16288 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16289 (walk_stores): Likewise.
16290 (arc_address_cost): Make it static.
16291 (arc_verify_short): Likewise.
16292 (branch_dest): Likewise.
16293 (arc_attr_type): Likewise.
16294 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16295 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16296 (arc_final_prescan_insn): Remove inserting the nops due to
16297 hardware hazards. It is done in reorg step.
16298 (insn_length_variant_t): Remove.
16299 (insn_length_parameters_t): Likewise.
16300 (arc_insn_length_parameters): Likewise.
16301 (arc_get_insn_variants): Likewise.
16302 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16304 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16306 PR inline-asm/84683
16307 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16310 PR tree-optimization/84687
16311 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16313 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16315 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16317 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16318 Rename to ppc_speculation_barrier.
16319 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16320 __builtin_ppc_speculation_barrier.
16322 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16325 * combine.c (combine_simplify_rtx): Don't try to simplify if
16326 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16329 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16331 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16332 to 32 bytes when compiling for POWER9.
16334 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16337 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16338 regparm >= 3 with no arg reg available also for calls with
16339 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16342 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16344 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16346 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16349 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16351 2018-03-05 Richard Biener <rguenther@suse.de>
16353 PR tree-optimization/84486
16354 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16355 When inserting a __builtin_assume_aligned call set the LHS
16356 SSA name alignment info accordingly.
16358 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16360 PR tree-optimization/84114
16361 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16362 Avoid reassociation of FLOAT_MODE addition.
16364 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16366 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16367 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16368 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16369 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16370 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16371 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16372 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16374 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16375 __builtin_ia32_wbinvd): New builtins.
16376 (SPECIAL_ARGS2): New.
16377 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
16378 (SPECIAL_ARGS2): New.
16379 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
16380 (ix86_valid_target_attribute_inner_p): Ditto.
16381 (ix86_init_mmx_sse_builtins): Add special_args2.
16382 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
16383 TARGET_WBNOINVD_P): New.
16384 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
16385 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
16386 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
16387 * config/i386/immintrin.h (_wbinvd): New intrinsic.
16388 * config/i386/pconfigintrin.h: New file.
16389 * config/i386/wbnoinvdintrin.h: Ditto.
16390 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
16392 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
16394 2018-03-05 Richard Biener <rguenther@suse.de>
16396 PR tree-optimization/84670
16397 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16399 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16400 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16401 (compute_antic_aux): Only assert the number of values in ANTIC_IN
16402 doesn't grow if all successors (recursively) were visited at least
16405 2018-03-05 Richard Biener <rguenther@suse.de>
16407 PR tree-optimization/84650
16408 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
16409 if executed in the loop pipeline.
16411 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
16413 * doc/configfiles.texi (Configuration Files): Move info about
16414 conditionalizing $target-protos.h to...
16415 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
16416 differs from $target-protos.h.
16418 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
16419 Chung-Ju Wu <jasonwucj@gmail.com>
16421 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
16422 * config/nds32/nds32-multiple.md (setmemsi): Define.
16423 * config/nds32/nds32-memory-manipulation.c
16424 (nds32_gen_dup_4_byte_to_word_value): New.
16425 (emit_setmem_word_loop): New.
16426 (emit_setmem_byte_loop): New.
16427 (nds32_expand_setmem_loop): New.
16428 (nds32_expand_setmem_loop_v3m): New.
16429 (nds32_expand_setmem_unroll): New.
16430 (nds32_expand_setmem): New.
16432 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16433 Chung-Ju Wu <jasonwucj@gmail.com>
16435 * config/nds32/nds32-memory-manipulation.c
16436 (nds32_emit_load_store): New.
16437 (nds32_emit_post_inc_load_store): New.
16438 (nds32_emit_mem_move): New.
16439 (nds32_emit_mem_move_block): New.
16440 (nds32_expand_movmemsi_loop_unknown_size): New.
16441 (nds32_expand_movmemsi_loop_known_size): New.
16442 (nds32_expand_movmemsi_loop): New.
16443 (nds32_expand_movmemsi_unroll): New.
16444 (nds32_expand_movmemqi): Rename ...
16445 (nds32_expand_movmemsi): ... to this.
16446 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
16447 (movmemsi): ... to this.
16448 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
16449 (nds32_expand_movmemsi): ... to this.
16451 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16452 Monk Chiang <sh.chiang04@gmail.com>
16453 Chung-Ju Wu <jasonwucj@gmail.com>
16455 * config/nds32/nds32-protos.h
16456 (nds32_expand_load_multiple): New arguments.
16457 (nds32_expand_store_multiple): Ditto.
16458 (nds32_valid_multiple_load_store): Rename ...
16459 (nds32_valid_multiple_load_store_p): ... to this.
16460 * config/nds32/nds32-memory-manipulation.c
16461 (nds32_expand_load_multiple): Refine implementation.
16462 (nds32_expand_store_multiple): Ditto.
16463 * config/nds32/nds32-multiple.md
16464 (load_multiple): Update nds32_expand_load_multiple interface.
16465 (store_multiple): Update nds32_expand_store_multiple interface.
16466 * config/nds32/nds32-predicates.c
16467 (nds32_valid_multiple_load_store): Rename ...
16468 (nds32_valid_multiple_load_store_p): ... to this and refine
16470 * config/nds32/predicates.md
16471 (nds32_load_multiple_and_update_address_operation): New predicate.
16472 (nds32_store_multiple_and_update_address_operation): New predicate.
16474 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16475 Chung-Ju Wu <jasonwucj@gmail.com>
16477 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
16478 (combo): New attribute.
16479 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
16481 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
16483 * config/nds32/nds32.opt: Change -mcmodel= default value.
16485 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
16486 Monk Chiang <sh.chiang04@gmail.com>
16487 Chung-Ju Wu <jasonwucj@gmail.com>
16489 * config/nds32/constants.md (unspec_element): New enum.
16490 * config/nds32/constraints.md (Umw): New constraint.
16491 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
16492 * config/nds32/nds32-intrinsic.md: Likewise.
16493 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
16494 (nds32_valid_smw_lwm_base_p): New.
16495 (nds32_output_smw_single_word): New.
16496 (nds32_output_lmw_single_word): New.
16497 (nds32_expand_unaligned_load): New.
16498 (nds32_expand_unaligned_store): New.
16499 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
16500 (nds32_output_smw_single_word): Declare.
16501 (nds32_output_lmw_single_word): Declare.
16502 (nds32_expand_unaligned_load): Declare.
16503 (nds32_expand_unaligned_store): Declare.
16504 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
16505 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
16506 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
16507 NDS32_BUILTIN_UASTORE_DW.
16508 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
16511 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
16512 Kito Cheng <kito.cheng@gmail.com>
16513 Chung-Ju Wu <jasonwucj@gmail.com>
16515 * config/nds32/nds32-intrinsic.c
16516 (nds32_expand_builtin_null_ftype_reg): Delete.
16517 (nds32_expand_builtin_reg_ftype_imm): Ditto.
16518 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
16519 (nds32_read_argument): New.
16520 (nds32_legitimize_target): Ditto.
16521 (nds32_legitimize_argument): Ditto.
16522 (nds32_check_constant_argument): Ditto.
16523 (nds32_expand_unop_builtin): Ditto.
16524 (nds32_expand_unopimm_builtin): Ditto.
16525 (nds32_expand_binop_builtin): Ditto.
16526 (nds32_builtin_decl_impl): Ditto.
16527 (builtin_description): Ditto.
16528 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
16529 (nds32_init_builtins_impl): Ditto.
16530 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
16531 (nds32_builtin_decl): New.
16532 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
16533 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
16535 2018-03-02 Jeff Law <law@redhat.com>
16537 * reorg.c (stop_search_p): Handle DEBUG_INSN.
16538 (redundant_insn, fill_simple_delay_slots): Likewise.
16539 (fill_slots_from_thread): Likewise.
16540 * resource.c (mark_referenced_resources): Likewise.
16541 (mark_set_resources, find_dead_or_set_registers): Likewise.
16543 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16545 * substring-locations.h (format_warning_va): Formatting fix for
16546 ATTRIBUTE_GCC_DIAG.
16547 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
16549 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
16550 * substring-locations.c: Include intl.h.
16551 (format_warning_va): Turned into small wrapper around
16552 format_warning_n_va, renamed to ...
16553 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
16554 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
16556 (format_warning_at_substring_n): New function.
16557 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
16558 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
16559 format_warning_at_substring with just a shorter name instead of
16560 const function pointer.
16561 (fmtwarn_n): New function.
16562 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
16563 appropriate, get rid of all the fmtstr temporaries, move conditionals
16564 with G_() wrapped string literals directly into fmtwarn arguments,
16565 cast dir.len to (int), formatting fixes.
16567 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
16569 * doc/invoke.texi: Remove "Cilk Plus" references.
16571 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16572 Richard Biener <rguenther@suse.de>
16575 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
16576 for error or warning attributes if CALL_FROM_THUNK_P is set.
16579 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16582 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
16583 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
16586 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
16587 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
16589 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
16590 instead of -1U in last predictors element's probability member.
16592 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16595 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16596 arguments if they are comparable.
16598 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16600 PR tree-optimization/84634
16601 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
16602 masks and masked_loop_p with a single loop_masks, making sure it's
16603 null for bb vectorization.
16605 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16607 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
16608 (vect_analyze_data_ref_access): Use loop->safe_len rather than
16609 loop->force_vectorize to check whether there is no alias.
16611 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16614 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
16616 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
16618 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
16619 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
16620 instead of a loop around prev_real_insn.
16621 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
16624 PR inline-asm/84625
16625 * config/i386/i386.c (ix86_print_operand): Use conditional
16626 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
16629 2018-03-02 Richard Biener <rguenther@suse.de>
16631 PR tree-optimization/84427
16632 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
16633 (bitmap_set_subtract_values): Rewrite to handle multiple
16636 (prune_clobbered_mems): Likewise.
16637 (phi_translate): Take edge instead of pred/phiblock.
16638 (phi_translate_1): Likewise.
16639 (phi_translate_set): Likewise. Insert all translated
16640 exprs for a value into the set, keeping possibly multiple
16641 expressions per value.
16642 (compute_antic_aux): Adjust for phi_translate changes.
16643 When intersecting union the expressions and prune those
16644 not in the final value set, keeping possibly multiple
16645 expressions per value. Do not use value-insertion
16646 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
16647 all expressions. Add verification that the value-sets
16648 only shrink during iteration.
16649 (compute_partial_antic_aux): Adjust for the phi_translate changes.
16650 (do_pre_regular_insertion): Likewise.
16651 (do_pre_partial_partial_insertion): Likewise.
16653 2018-03-02 Richard Biener <rguenther@suse.de>
16656 * config/darwin.c (saved_debug_info_level): New static global.
16657 (darwin_asm_lto_start): Disable debug info generation for LTO out.
16658 (darwin_asm_lto_end): Restore debug info generation settings.
16660 2018-03-01 Martin Liska <mliska@suse.cz>
16663 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
16664 volatile arguments.
16666 2018-03-01 Richard Biener <rguenther@suse.de>
16669 * dwarf2out.c (gen_variable_die): Properly handle late VLA
16670 type annotation with LTO when debug was disabled at compile-time.
16672 2018-03-01 Matthew Fortune <mfortune@gmail.com>
16674 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
16676 (mips_final_postscan_insn): Likewise.
16678 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
16680 PR rtl-optimization/84528
16681 * alias.c (init_alias_target): Add commentary.
16682 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
16683 a unique base value if the frame pointer is not eliminated
16684 to the stack pointer.
16686 2018-03-01 Tom de Vries <tom@codesourcery.com>
16688 PR rtl-optimization/83327
16689 * lra-int.h (hard_regs_spilled_into): Declare.
16690 * lra.c (hard_regs_spilled_into): Define.
16691 (init_reg_info): Init hard_regs_spilled_into.
16692 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
16693 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
16694 (process_bb_lives): Handle hard_regs_spilled_into.
16695 (lra_create_live_ranges_1): Before doing liveness propagation, clear
16696 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
16698 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
16700 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
16701 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
16702 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
16703 * config/rs6000/aix72.h: New file.
16705 2018-02-28 Jakub Jelinek <jakub@redhat.com>
16707 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
16708 instead of warning_at with conditional singular and plural messages
16712 * stor-layout.c (update_alignment_for_field): For
16713 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
16714 && !DECL_PACKED (field), do the alignment update, just use
16715 only desired_align instead of MAX (type_align, desired_align)
16717 (place_field): Don't do known_align < desired_align handling
16718 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
16719 is non-NULL, instead do it after rli->prev_field handling and
16720 only if not within a bitfield word. For DECL_PACKED (field)
16721 use type_align of BITS_PER_UNIT.
16723 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
16725 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
16726 superfluous parentheses and trailing spaces.
16728 2018-02-28 Richard Biener <rguenther@suse.de>
16730 PR tree-optimization/84584
16731 * graphite-scop-detection.c (scop_detection::add_scop): Discard
16732 SCoPs with fake exit edge.
16734 2018-02-28 Martin Liska <mliska@suse.cz>
16737 * timevar.c (timer::print): Fix format to properly print 100%
16740 2018-02-28 Richard Biener <rguenther@suse.de>
16742 PR middle-end/84607
16743 * genmatch.c (capture_info::walk_match): Do not mark
16744 captured expressions without operands as expr_p given
16745 they act more like predicates and should be subject to
16746 "lost tail" side-effect preserving.
16748 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
16750 PR rtl-optimization/81611
16751 * auto-inc-dec.c (attempt_change): Move dead note from
16752 mem_insn if it's the next use of regno
16753 (find_address): Take address use of reg holding
16754 non-incremented value. Add parm to limit search to the named
16756 (merge_in_block): Attempt to use a mem insn that is the next
16757 use of the original regno.
16759 2018-02-27 Martin Sebor <msebor@redhat.com>
16762 * doc/invoke.texi (-Wmissing-attributes): New option.
16763 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
16765 2018-02-27 Martin Sebor <msebor@redhat.com>
16767 PR translation/84207
16768 * diagnostic-core.h (warning_n, error_n, inform_n): Change
16769 n argument to unsigned HOST_WIDE_INT.
16770 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
16771 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
16772 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
16773 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
16775 2018-02-27 Richard Biener <rguenther@suse.de>
16777 PR tree-optimization/84512
16778 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16779 Do not use the estimate returned from record_stmt_cost for
16780 the scalar iteration cost but sum properly using add_stmt_cost.
16782 2018-02-27 Richard Biener <rguenther@suse.de>
16784 PR tree-optimization/84466
16785 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16786 Adjust last change to less strictly validate use operands.
16788 2018-02-27 Martin Liska <mliska@suse.cz>
16790 PR gcov-profile/84548
16791 * gcov.c (process_file): Allow partial overlap and consider it
16792 also as group functions.
16793 (output_lines): Properly calculate range of lines for a group.
16795 2018-02-27 Martin Liska <mliska@suse.cz>
16797 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
16798 'ggc' suffixes. Change first column width.
16799 (timer::print): Fix formatting of the column.
16801 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
16803 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
16804 preserve inline entry blocks for the sake of debug inline
16805 entry point markers alone.
16806 (remove_unused_locals): Suggest in comments a better place to
16807 force the preservation of inline entry blocks that are
16808 otherwise unused, but do not preserve them.
16810 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16812 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
16814 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16817 * config/i386/constraints.md (Bs): Replace
16818 ix86_indirect_branch_register with
16819 TARGET_INDIRECT_BRANCH_REGISTER.
16821 * config/i386/i386.md (indirect_jump): Likewise.
16822 (tablejump): Likewise.
16823 (*sibcall_memory): Likewise.
16824 (*sibcall_value_memory): Likewise.
16825 Peepholes of indirect call and jump via memory: Likewise.
16826 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
16827 (*sibcall_value_GOT_32): Likewise.
16828 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16829 (GOT_memory_operand): Likewise.
16830 (call_insn_operand): Likewise.
16831 (sibcall_insn_operand): Likewise.
16832 (GOT32_symbol_operand): Likewise.
16833 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
16835 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16837 PR rtl-optimization/83496
16838 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
16839 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
16840 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
16841 redundant insn, if any.
16842 (relax_delay_slots): Likewise.
16843 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
16845 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
16847 PR tree-optimization/83965
16848 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
16849 that grouped statements are part of a reduction chain. Return
16850 true if the statement is not marked as a reduction itself but
16851 is part of a group.
16852 (vect_recog_dot_prod_pattern): Don't check whether the statement
16853 is part of a group here.
16854 (vect_recog_sad_pattern): Likewise.
16855 (vect_recog_widen_sum_pattern): Likewise.
16857 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16860 * final.c (rest_of_clean_state): Also look for calls inside sequences.
16862 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16865 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
16867 (ix86_output_indirect_function_return): New prototype.
16868 (ix86_split_simple_return_pop_internal): Likewise.
16869 * config/i386/i386.c (indirect_return_via_cx): New.
16870 (indirect_return_via_cx_bnd): Likewise.
16871 (indirect_thunk_name): Handle return va CX_REG.
16872 (output_indirect_thunk_function): Create alias for
16873 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
16874 (ix86_output_indirect_jmp): Remove the bool argument.
16875 (ix86_output_indirect_function_return): New function.
16876 (ix86_split_simple_return_pop_internal): Likewise.
16877 * config/i386/i386.md (*indirect_jump): Don't pass false
16878 to ix86_output_indirect_jmp.
16879 (*tablejump_1): Likewise.
16880 (simple_return_pop_internal): Change it to define_insn_and_split.
16881 Call ix86_split_simple_return_pop_internal to split it for
16882 -mfunction-return=.
16883 (simple_return_indirect_internal): Call
16884 ix86_output_indirect_function_return instead of
16885 ix86_output_indirect_jmp.
16887 2018-02-26 Jakub Jelinek <jakub@redhat.com>
16890 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
16891 memset and value initialization afterwards.
16893 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
16895 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
16897 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16900 * common/config/aarch64/aarch64-common.c
16901 (aarch_option_optimization_table[]): Switch
16902 off fomit-frame-pointer
16904 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16905 Chung-Ju Wu <jasonwucj@gmail.com>
16907 * config/nds32/nds32-multiple.md (load_multiple): Disallow
16909 (store_multiple): Ditto.
16911 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16913 * config.gcc: Add --with-cpu support for nds32 target.
16914 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
16915 * config/nds32/nds32.opt: Add -mcpu= option.
16917 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
16919 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
16920 isel=yes): Warn for these deprecated options.
16922 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
16924 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
16925 ISA_2_5_MASKS_EMBEDDED.
16927 2018-02-23 Jakub Jelinek <jakub@redhat.com>
16929 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
16930 p->max as pointers rather than using iterative_hash_expr.
16932 2018-02-23 Carl Love <cel@us.ibm.com>
16934 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
16935 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
16937 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
16938 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
16939 P8V_BUILTIN_VEC_VUNSIGNED2.
16941 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
16944 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
16945 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
16947 * lra-constraints.c (curr_insn_transform): Set up
16948 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
16950 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
16951 * lra-eliminations.c (spill_pseudos): Ditto.
16952 (process_insn_for_elimination): Ditto.
16953 * lra-lives.c (reg_early_clobber_p): Use the new macros.
16954 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
16955 LRA_NON_CLOBBERED_ALT.
16957 2018-02-22 Martin Sebor <msebor@redhat.com>
16959 PR tree-optimization/84480
16960 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
16961 to maybe_diag_stxncpy_trunc. Call it.
16962 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
16963 from gimple_fold_builtin_strcpy. Print inlining stack.
16964 (handle_builtin_stxncpy): Print inlining stack.
16965 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
16967 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
16970 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
16971 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
16972 and -fcheck-pointer-bounds are used together.
16973 (indirect_thunk_prefix): New enum.
16974 (indirect_thunk_need_prefix): New function.
16975 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
16976 "_nt" instead of "_bnd" for NOTRACK prefix.
16977 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
16978 (output_indirect_thunk_function): Likewise.
16980 (ix86_code_end): Update output_indirect_thunk_function calls.
16981 (ix86_output_indirect_branch_via_reg): Replace
16982 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
16983 (ix86_output_indirect_branch_via_push): Likewise.
16984 (ix86_output_function_return): Likewise.
16985 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
16986 incompatible with -fcf-protection=branch and
16987 -fcheck-pointer-bounds.
16989 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16992 * config/aarch64/aarch64.c (aarch64_print_address_internal):
16993 Change gcc_assert call to output_operand_lossage.
16995 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16997 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
16999 2018-02-22 DJ Delorie <dj@redhat.com>
17000 Sebastian Perta <sebastian.perta@renesas.com>
17001 Oleg Endo <olegendo@gcc.gnu.org>
17003 * config/rx/rx.c (rx_rtx_costs): New function.
17004 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
17006 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
17008 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
17010 2018-02-22 Martin Liska <mliska@suse.cz>
17013 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
17014 Add "native" as a possible value.
17016 2018-02-22 Martin Liska <mliska@suse.cz>
17019 * config/i386/i386.c (ix86_option_override_internal):
17020 Add "native" as a possible value for -march and -mtune.
17022 2018-02-22 Jakub Jelinek <jakub@redhat.com>
17025 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
17026 to all type variants.
17028 PR tree-optimization/84503
17029 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
17030 width as info->bitpos + info->bitsize - start.
17031 (merged_store_group::merge_overlapping): Simplify width computation.
17032 (check_no_overlap): New function.
17033 (imm_store_chain_info::try_coalesce_bswap): Compute expected
17034 start + width and last_order of the group, fail if check_no_overlap
17036 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
17037 to group if check_no_overlap fails.
17039 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17041 * config/rs6000/altivec.md: Delete contraint arguments to
17042 define_expand, define_split, and define_peephole2, and in
17043 define_insn_and_split if always unused.
17044 * config/rs6000/darwin.md: Ditto.
17045 * config/rs6000/dfp.md: Ditto.
17046 * config/rs6000/rs6000.md: Ditto.
17047 * config/rs6000/sync.md: Ditto.
17048 * config/rs6000/vector.md: Ditto.
17049 * config/rs6000/vsx.md: Ditto.
17051 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17053 * config/rs6000/altivec.md: Write output control strings as braced
17054 blocks instead of double-quoted strings.
17055 * config/rs6000/darwin.md: Ditto.
17056 * config/rs6000/rs6000.md: Ditto.
17057 * config/rs6000/vector.md: Ditto.
17058 * config/rs6000/vsx.md: Ditto.
17060 2018-02-21 Jason Merrill <jason@redhat.com>
17062 PR c++/84314 - ICE with templates and fastcall attribute.
17063 * attribs.c (build_type_attribute_qual_variant): Remove assert.
17065 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17067 * ipa-cp.c (determine_versionability): Fix comment typos.
17069 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17072 * ipa-cp.c (determine_versionability): Do not version functions caling
17075 2018-02-21 Martin Liska <mliska@suse.cz>
17078 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
17079 Add "native" as a possible value.
17080 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
17081 the macro when native cpu detection is available.
17083 2018-02-21 Martin Liska <mliska@suse.cz>
17086 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
17087 Add "native" as a possible value.
17088 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
17089 when native cpu detection is available.
17091 2018-02-21 Jakub Jelinek <jakub@redhat.com>
17092 Martin Sebor <msebor@redhat.com>
17094 PR tree-optimization/84478
17095 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
17097 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
17098 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
17099 support which is conservatively correct, for 2 only stay conservative
17100 for maxlen. Formatting and comment capitalization fixes. Add STRICT
17101 argument to the 2 argument get_range_strlen, adjust 6 arg
17102 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
17104 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
17105 (gimple_fold_builtin_strlen): Pass true as last argument to
17108 2018-02-20 Martin Sebor <msebor@redhat.com>
17110 PR middle-end/84095
17111 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
17112 (builtin_memref::set_base_and_offset): Same. Handle inner references.
17113 (builtin_memref::builtin_memref): Factor out parts into
17114 set_base_and_offset and call it.
17116 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17118 PR middle-end/84406
17119 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
17120 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
17121 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
17122 search at the associated MODE_INT.
17124 2018-02-20 Jeff Law <law@redhat.com>
17126 PR middle-end/82123
17127 PR tree-optimization/81592
17128 PR middle-end/79257
17129 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
17130 for range data rather than using global data.
17131 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
17132 range data rather than using global data.
17133 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
17134 pass it to children as needed.
17135 (struct directive::fmtresult): Similarly.
17136 (struct directive::set_width): Similarly.
17137 (struct directive::set_precision): Similarly.
17138 (format_integer, format_directive, parse_directive): Similarly.
17139 (format_none): Accept unnamed vr_values parameter.
17140 (format_percent, format_floating, format_character): Similarly.
17141 (format_string, format_plain): Similarly.
17142 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
17143 the EVRP range analyzer for range data rather than using global data.
17144 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
17145 gimple-ssa-evrp-analyze.h
17146 (class sprintf_dom_walker): Add after_dom_children member function.
17147 Add evrp_range_analyzer member.
17148 (sprintf_dom_walker::before_dom_children): Call into the EVRP
17149 range analyzer as needed.
17150 (sprintf_dom_walker::after_dom_children): New member function.
17151 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
17153 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
17154 (evrp_range_analyzer::pop_to_marker): Likewise.
17156 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17158 PR tree-optimization/84419
17159 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
17160 with the required type if its current type is compatible but
17163 2018-02-20 Jakub Jelinek <jakub@redhat.com>
17165 PR middle-end/82004
17166 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
17167 after vectorization.
17169 2018-02-20 Martin Liska <mliska@suse.cz>
17172 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
17173 possible values if we don't have a hint.
17175 2018-02-20 Martin Liska <mliska@suse.cz>
17179 * final.c (shorten_branches): Build align_tab array with one
17181 * opts.c (finish_options): Add alignment option limit check.
17182 (MAX_CODE_ALIGN): Likewise.
17183 (MAX_CODE_ALIGN_VALUE): Likewise.
17184 * doc/invoke.texi: Document maximum allowed option value for
17185 all -falign-* options.
17187 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17190 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17191 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17192 * var-tracking.c (emit_note_insn_var_location): Remove all references
17193 to NOTE_INSN_CALL_ARG_LOCATION.
17194 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17195 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17196 Use copy_rtx_if_shared.
17197 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17198 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17199 (dwarf2out_var_location): Remove handling of
17200 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17202 * final.c (final_scan_insn): Remove all references to
17203 NOTE_INSN_CALL_ARG_LOCATION.
17204 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17205 before dumping final insns.
17206 * except.c (emit_note_eh_region_end): Remove all references to
17207 NOTE_INSN_CALL_ARG_LOCATION.
17208 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17209 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17210 * config/arc/arc.c (hwloop_optimize): Likewise.
17211 * config/arm/arm.c (create_fix_barrier): Likewise.
17212 * config/s390/s390.c (s390_chunkify_start): Likewise.
17213 * config/sh/sh.c (find_barrier): Likewise.
17214 * config/i386/i386.c (rest_of_insert_endbranch,
17215 ix86_seh_fixup_eh_fallthru): Likewise.
17216 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17217 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17218 * config/frv/frv.c (frv_function_prologue): Likewise.
17219 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17221 (note_outside_basic_block_p): Remove all references to
17222 NOTE_INSN_CALL_ARG_LOCATION.
17223 * gengtype.c (adjust_field_rtx_def): Likewise.
17224 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17226 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17227 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17230 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17233 PR tree-optimization/84452
17234 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17235 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17238 2018-02-19 Martin Liska <mliska@suse.cz>
17241 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17243 2018-02-19 Martin Liska <mliska@suse.cz>
17244 Richard Sandiford <richard.sandiford@linaro.org>
17246 PR tree-optimization/82491
17247 * gimple-fold.c (get_base_constructor): Make earlier bail out
17250 2018-02-19 Carl Love <cel@us.ibm.com>
17252 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17253 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17255 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17256 P8V_BUILTIN_VEC_NEG.
17258 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17260 * config/rl78/rl78.md (movdf): New define expand.
17262 2018-02-19 Martin Liska <mliska@suse.cz>
17265 * doc/invoke.texi: Fix typo.
17266 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17268 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17270 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17271 handle rs6000_single_float and rs6000_double_float specially for
17274 2018-02-16 Jeff Law <law@redhat.com>
17276 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17277 (pop_regs): Use it.
17279 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17282 * ipa-inline.c (inline_small_functions): Fix a typo.
17284 2018-02-16 Nathan Sidwell <nathan@acm.org>
17286 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17288 2018-02-16 Carl Love <cel@us.ibm.com>
17290 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17291 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17292 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17293 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17294 expansion to P8V_BUILTIN_VEC_FLOAT2.
17296 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17298 PR rtl-optimization/70023
17299 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17300 src_regno into account.
17302 2018-02-16 Carl Love <cel@us.ibm.com>
17304 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17305 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17306 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17307 * config/rs6000/rs6000.c: Remove case statements for
17308 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17309 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17310 and P9V_BUILTIN_VEC_VINSERT4B.
17311 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17312 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17313 * config/rs6000/vsx.md:
17314 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17317 2018-02-16 Carl Love <cel@us.ibm.com>
17319 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17321 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17323 * config/rs6000/rs6000-c.c: Add the definitions for
17324 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17325 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17326 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17327 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17328 definition for insert4b and define insn *insert3b_internal.
17329 * doc/extend.texi: Add documentation for vec_extract4b.
17331 2018-02-16 Nathan Sidwell <nathan@acm.org>
17333 * doc/extend.texi (Backwards Compatibility): Mention friend
17334 injection. Note for-scope is deprecated.
17335 * doc/invoke.texi (-ffriend-injection): Deprecate.
17337 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17339 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17340 that moved to I2, also allow destinations that are a paradoxical
17341 subreg (instead of a normal reg).
17343 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17346 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17349 2018-02-16 Richard Biener <rguenther@suse.de>
17351 PR tree-optimization/84037
17352 PR tree-optimization/84016
17354 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17355 Adjust vec_construct for the fact we need additional higher latency
17356 128bit inserts for AVX256 and AVX512 vector builds.
17357 (ix86_add_stmt_cost): Scale vector construction cost for
17360 2018-02-16 Richard Biener <rguenther@suse.de>
17362 PR tree-optimization/84417
17363 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17364 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17365 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17367 2018-02-16 Martin Liska <mliska@suse.cz>
17370 * internal-fn.def (ASAN_CHECK): Set proper flags.
17371 (ASAN_MARK): Likewise.
17373 2018-02-16 Julia Koval <julia.koval@intel.com>
17375 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17376 from PTA_CANNONLAKE.
17378 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17381 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
17382 Use ++iter rather than iter++ for std::list iterators.
17383 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
17384 defer deleting them until all nodes in the forest are processed. Do
17385 free even leaf nodes. Change to_process into auto_vec.
17388 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
17389 * vec.h (vec_default_construct): Use memset instead of placement new
17390 if BROKEN_VALUE_INITIALIZATION is defined.
17391 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
17392 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
17395 PR rtl-optimization/83723
17396 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
17397 * lra.c (lra_substitute_pseudo): Likewise. If true, use
17398 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
17400 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
17402 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
17404 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
17406 PR rtl-optimization/81443
17407 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
17408 from inner REGs to paradoxical SUBREGs.
17410 2018-02-16 Richard Biener <rguenther@suse.de>
17412 PR tree-optimization/84399
17413 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17414 For operands we can analyze at their definition make sure we can
17415 analyze them at each use as well.
17417 2018-02-16 Richard Biener <rguenther@suse.de>
17419 PR tree-optimization/84190
17420 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
17421 volatile accesses if the decl isn't volatile.
17423 2018-02-15 Jason Merrill <jason@redhat.com>
17425 PR c++/84314 - ICE with templates and fastcall attribute.
17426 * attribs.c (build_type_attribute_qual_variant): Don't clobber
17427 TYPE_CANONICAL on an existing type.
17429 2018-02-15 Jakub Jelinek <jakub@redhat.com>
17431 PR tree-optimization/84383
17432 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
17433 dstoff nor call operand_equal_p if dstbase is NULL.
17435 PR tree-optimization/84334
17436 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
17437 also a CONSTANT_CLASS_P, punt.
17439 2018-02-14 Jim Wilson <jimw@sifive.com>
17441 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
17442 first SMALL_OPERAND check. New local min_second_step. Move assert
17443 to where locals are set. Add TARGET_RVC support.
17444 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
17446 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
17448 * doc/invoke.texi: Correct -Wformat-overflow code sample.
17450 2018-02-14 Martin Sebor <msebor@redhat.com>
17452 PR tree-optimization/83698
17453 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
17454 arrays constrain the offset range to their bounds.
17455 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
17456 (builtin_access::overlap): Avoid setting the size of overlap if it's
17458 (maybe_diag_overlap): Also consider arrays when deciding what values
17459 of offsets to include in diagnostics.
17461 2018-02-14 Martin Sebor <msebor@redhat.com>
17464 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
17465 that correspond to the kind of a declaration.
17467 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
17470 * config/pa/pa.md: Load address of PIC label using the linkage table
17471 if the label is nonlocal.
17473 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
17475 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
17476 warning message if user requests -maltivec=be.
17477 * doc/invoke.texi: Document deprecation of -maltivec=be.
17479 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
17482 * config/rs6000/rs6000-c.c: Update definitions for
17483 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
17484 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
17486 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17489 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
17490 add _get_ssp intrinsics. Remove argument from
17491 __builtin_ia32_rdssp[d|q].
17492 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
17493 * config/i386/i386-builtin.def: Remove argument from
17494 __builtin_ia32_rdssp[d|q].
17495 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
17496 ix86_expand_special_args_builtin for _rdssp[d|q].
17497 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
17498 Clear register before usage.
17499 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
17500 Add documentation for new _get_ssp and _inc_ssp intrinsics.
17502 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
17504 PR tree-optimization/84357
17505 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
17506 operand 1 of an ARRAY_REF too.
17508 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
17511 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
17512 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17514 (set_of_reg): New struct.
17515 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
17516 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
17517 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17519 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
17520 Split into bitclr, bitset, bitinvert patterns if appropriate.
17521 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
17522 use rx_fuse_in_memory_bitop.
17523 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
17524 to named insn, correct maximum insn length.
17526 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
17529 * machmode.def: Define a complex mode for PARTIAL_INT.
17530 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
17532 * doc/rtl.texi: Document CSPImode.
17533 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
17535 (msp430_hard_regno_nregs_with_padding): Likewise.
17537 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
17540 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
17542 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
17544 PR rtl-optimization/84169
17545 * combine.c (try_combine): New variable split_i2i3. Set it to true if
17546 we generated a parallel as new i3 and we split that to new i2 and i3
17547 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
17548 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
17549 those to i2, not i1. Partially rewrite this scan code.
17551 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17554 * stor-layout.c (place_field): For variable length fields, adjust
17555 offset_align afterwards not just based on the field's alignment,
17556 but also on the size.
17558 PR middle-end/84309
17559 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
17560 of exps and logs in the use_exp2 case.
17562 2018-02-13 Jeff Law <law@redhat.com>
17564 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
17565 entry for "vector".
17567 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
17570 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
17574 * common.opt (gas-loc-support, gas-locview-support): New.
17575 (ginline-points, ginternal-reset-location-views): New.
17576 * doc/invoke.texi: Document them. Use @itemx where intended.
17577 (gvariable-location-views): Adjust.
17578 * target.def (reset_location_view): New.
17579 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
17580 (TARGET_RESET_LOCATION_VIEW): New.
17581 * doc/tm.texi: Rebuilt.
17582 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
17583 (dwarf2out_default_as_locview_support): New.
17584 (output_asm_line_debug_info): Use option variables.
17585 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
17586 (output_loc_list): Likewise.
17587 (add_high_low_attributes): Check option variables.
17588 Don't output entry view attribute in strict mode.
17589 (gen_inlined_subroutine_die): Check option variables.
17590 (dwarf2out_inline_entry): Likewise.
17591 (init_sections_and_labels): Likewise.
17592 (dwarf2out_early_finish): Likewise.
17593 (maybe_reset_location_view): New, from...
17594 (dwarf2out_var_location): ... here. Call it.
17595 * debug.h (dwarf2out_default_as_loc_support): Declare.
17596 (dwarf2out_default_as_locview_support): Declare.
17597 * hooks.c (hook_int_rtx_insn_0): New.
17598 * hooks.h (hook_int_rtx_insn_0): Declare.
17599 * toplev.c (process_options): Take -gas-loc-support and
17600 -gas-locview-support from dwarf2out. Enable
17601 -gvariable-location-views by default only with locview
17602 assembler support. Enable -ginternal-reset-location-views by
17603 default only if the target defines the corresponding hook.
17604 Enable -ginline-points by default if location views are
17605 enabled; force it disabled if statement frontiers are
17607 * tree-inline.c (expand_call_inline): Check option variables.
17608 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
17610 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
17612 PR tree-optimization/84321
17613 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
17614 handling. Also check whether the anti-range contains any values
17615 that satisfy the mask; switch to a VR_RANGE if not.
17617 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
17620 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
17622 2018-02-13 Martin Jambor <mjambor@suse.cz>
17625 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
17626 of call statements, also set location of a load to a temporary.
17628 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
17630 * config/rl78/rl78.c (add_vector_labels): New function.
17631 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
17632 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
17633 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
17634 which checks that no arguments are passed.
17635 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
17636 * doc/extend.texi: Documentation for the new attribute.
17638 2018-02-13 Andreas Schwab <schwab@suse.de>
17640 * config/riscv/linux.h (CPP_SPEC): Define.
17642 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17645 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
17646 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
17647 OPTION_MASK_ISA_AES as first argument to def_builtin_const
17648 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
17649 instead of OPTION_MASK_ISA_PCLMUL as first argument to
17650 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
17651 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
17652 temporarily for AES and PCLMUL builtins.
17654 PR tree-optimization/84339
17655 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
17656 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
17659 PR middle-end/84309
17660 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
17661 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
17662 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
17664 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
17666 * omp-simd-clone.h: New file.
17667 * omp-simd-clone.c: Include omp-simd-clone.h.
17668 (expand_simd_clones): No longer static.
17669 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
17670 cgraph.h and omp-simd-clone.h.
17671 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
17672 (vect_recog_widen_shift_pattern): Formatting fix.
17673 (vect_pattern_recog_1): Don't check optab for calls.
17676 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
17677 operands[2] into a REG before using gen_lowpart on it.
17679 2018-02-12 Jeff Law <law@redhat.com>
17682 * config/sh/sh.c (find_barrier): Consider a sibling call
17685 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
17686 successfully back substituting a reg.
17688 2018-02-12 Richard Biener <rguenther@suse.de>
17690 PR tree-optimization/84037
17691 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
17692 parameter, move visited init to caller.
17693 (vect_slp_analyze_operations): Separate cost from validity
17694 check, initialize visited once for all instances.
17695 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
17697 * tree-vect-stmts.c (vect_model_simple_cost): Make early
17699 (vect_model_promotion_demotion_cost): Likewise.
17700 (vectorizable_bswap): Guard cost modeling with !slp_node
17701 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
17703 (vectorizable_call): Likewise.
17704 (vectorizable_conversion): Likewise.
17705 (vectorizable_assignment): Likewise.
17706 (vectorizable_shift): Likewise.
17707 (vectorizable_operation): Likewise.
17708 (vectorizable_store): Likewise.
17709 (vectorizable_load): Likewise.
17710 (vectorizable_condition): Likewise.
17711 (vectorizable_comparison): Likewise.
17713 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
17716 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
17717 (ASAN_MARK): Fix fnspec to account for return value, change pointer
17718 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
17720 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17722 PR middle-end/83665
17723 * params.def (inline-min-speedup): Increase from 8 to 15.
17724 (max-inline-insns-auto): Decrease from 40 to 30.
17725 * ipa-split.c (consider_split): Add some buffer for function to
17726 be considered inlining candidate.
17727 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
17730 2018-02-12 Richard Biener <rguenther@suse.de>
17732 PR tree-optimization/84037
17733 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
17734 matched stmts if we cannot swap the non-matched ones.
17736 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17738 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
17739 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
17740 _mm_maskz_scalef_round_ss): New intrinsics.
17741 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
17742 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
17743 __builtin_ia32_scalefss_round): Remove.
17744 (__builtin_ia32_scalefsd_mask_round,
17745 __builtin_ia32_scalefss_mask_round): New intrinsics.
17746 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
17747 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
17748 ((match_operand:VF_128 2 "<round_nimm_predicate>"
17749 "<round_constraint>")): Changed to ...
17750 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
17751 "<round_scalar_constraint>")): ... this.
17752 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
17753 %0, %1, %2<round_op3>}"): Changed to ...
17754 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
17755 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
17756 %2<round_scalar_mask_op3>}"): ... this.
17757 * config/i386/subst.md (round_scalar_nimm_predicate): New.
17759 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17761 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
17762 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
17763 (_mm_maskz_sqrt_round_ss): New intrinsics.
17764 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
17765 (__builtin_ia32_sqrtsd_mask_round)
17766 (__builtin_ia32_sqrtss_mask_round): New builtins.
17767 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
17768 (__builtin_ia32_sqrtss_round): Remove.
17769 (__builtin_ia32_sqrtsd_mask_round)
17770 (__builtin_ia32_sqrtss_mask_round): New builtins.
17771 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
17772 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
17773 ((match_operand:VF_128 1 "vector_operand"
17774 "xBm,<round_constraint>")): Changed to ...
17775 ((match_operand:VF_128 1 "vector_operand"
17776 "xBm,<round_scalar_constraint>")): ... this.
17777 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
17778 %0, %2, %<iptr>1<round_op3>}): Changed to ...
17779 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
17780 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
17781 %<iptr>1<round_scalar_mask_op3>}): ... this.
17782 ((set_attr "prefix" "<round_prefix>")): Changed to ...
17783 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
17785 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
17788 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
17789 Cast vec_cmpeq result to correct type.
17790 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
17791 Cast vec_cmpgt result to correct type.
17793 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
17795 * final.c (final_scan_insn_1): Renamed from...
17796 (final_scan_insn): ... this. New wrapper, to recover
17797 seen from the outermost call in recursive ones.
17798 * config/sparc/sparc.c (output_return): Drop seen from call.
17799 (output_sibcall): Likewise.
17800 * config/visium/visium.c (output_branch): Likewise.
17802 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
17804 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
17807 2018-02-10 Alan Modra <amodra@gmail.com>
17810 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
17811 Specify LR as an input.
17813 2018-02-10 Jakub Jelinek <jakub@redhat.com>
17816 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
17817 remove_member_access_dummy_vars): New functions.
17818 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
17819 lower_omp_1, execute_lower_omp): Use them.
17821 PR rtl-optimization/84308
17822 * shrink-wrap.c (spread_components): Release todo vector.
17824 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
17826 PR rtl-optimization/57193
17827 * ira-color.c (struct allocno_color_data): Add member
17828 conflict_allocno_hard_prefs.
17829 (update_conflict_allocno_hard_prefs): New.
17830 (bucket_allocno_compare_func): Add a preference based on
17831 conflict_allocno_hard_prefs.
17832 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
17833 (color_allocnos): Remove a dead code. Initiate
17834 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
17836 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17839 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
17840 constraint from =wa to wa. Avoid a subreg on the output operand,
17841 instead use a pseudo and subreg it in a move.
17842 (p9_xxbrd_<mode>): Changed to ...
17843 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
17844 (p9_xxbrd_v2df): New expander.
17845 (p9_xxbrw_<mode>): Changed to ...
17846 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
17847 (p9_xxbrw_v4sf): New expander.
17849 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17851 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
17853 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
17856 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
17857 multiply in 32-bit mode.
17858 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
17859 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
17862 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17864 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
17865 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
17866 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
17867 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
17869 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
17872 * dwarf2out.c (is_trivial_indirect_ref): New function.
17873 (dwarf2out_late_global_decl): Do not generate a location
17874 attribute for variables that have a non-trivial DECL_VALUE_EXPR
17875 and that are not defined in the current unit.
17877 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17879 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
17880 instead of a libcall for UNORDERED.
17882 2018-02-09 Tamar Christina <tamar.christina@arm.com>
17885 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
17886 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
17888 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17891 * config/s390/s390.c (s390_set_current_function): Invoke
17892 s390_indirect_branch_settings also if fndecl didn't change.
17894 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17896 * config/rs6000/rs6000.md (blockage): Set length to zero.
17898 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17900 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
17902 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17905 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
17906 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
17910 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
17911 PARALLEL incoming that failed vt_get_decl_and_offset check.
17913 PR middle-end/84237
17914 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
17915 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
17917 (get_variable_section): For decls in named .bss* sections pass true as
17918 second argument to bss_initializer_p.
17920 2018-02-09 Marek Polacek <polacek@redhat.com>
17921 Jakub Jelinek <jakub@redhat.com>
17924 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
17925 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
17926 Sync some changes from cxx_fold_indirect_ref.
17928 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17930 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
17932 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
17933 (BLOCK_INLINE_ENTRY_LABEL): New.
17934 (dwarf2out_var_location): Disregard inline entry markers.
17935 (inline_entry_data): New struct.
17936 (inline_entry_data_hasher): New hashtable type.
17937 (inline_entry_data_hasher::hash): New.
17938 (inline_entry_data_hasher::equal): New.
17939 (inline_entry_data_table): New variable.
17940 (add_high_low_attributes): Add DW_AT_entry_pc and
17941 DW_AT_GNU_entry_view attributes if a pending entry is found
17942 in inline_entry_data_table. Add old entry_pc attribute only
17943 if debug nonbinding markers are disabled.
17944 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
17945 markers are enabled.
17946 (block_within_block_p, dwarf2out_inline_entry): New.
17947 (dwarf2out_finish): Check that no entries remained in
17948 inline_entry_data_table.
17949 * final.c (reemit_insn_block_notes): Handle inline entry notes.
17950 (final_scan_insn, notice_source_line): Likewise.
17951 (rest_of_clean_state): Skip inline entry markers.
17952 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
17954 * gimple.c (gimple_build_debug_inline_entry): New.
17955 * gimple.h (enum gimple_debug_subcode): Add
17956 GIMPLE_DEBUG_INLINE_ENTRY.
17957 (gimple_build_debug_inline_entry): Declare.
17958 (gimple_debug_inline_entry_p): New.
17959 (gimple_debug_nonbind_marker_p): Adjust.
17960 * insn-notes.def (INLINE_ENTRY): New.
17961 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
17962 inline entry marker notes.
17963 (print_insn): Likewise.
17964 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
17965 (INSN_DEBUG_MARKER_KIND): Likewise.
17966 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
17967 * tree-inline.c (expand_call_inline): Build and insert
17968 debug_inline_entry stmt.
17969 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
17970 inline entry blocks early, if nonbind markers are enabled.
17971 (dump_scope_block): Dump fragment info.
17972 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
17973 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
17974 (gimple_build_debug_inline_entry): New.
17975 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
17976 Enable/disable inline entry points too.
17977 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
17978 (DEBUG_INSN): Describe inline entry markers.
17980 * common.opt (gvariable-location-views): New.
17981 (gvariable-location-views=incompat5): New.
17982 * config.in: Rebuilt.
17983 * configure: Rebuilt.
17984 * configure.ac: Test assembler for view support.
17985 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
17986 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
17987 * dwarf2out.c (var_loc_view): New typedef.
17988 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
17989 (dwarf2out_locviews_in_attribute): New.
17990 (dwarf2out_locviews_in_loclist): New.
17991 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
17992 (enum dw_line_info_opcode): Add LI_adv_address.
17993 (struct dw_line_info_table): Add view.
17994 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
17995 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
17996 (zero_view_p): New variable.
17997 (ZERO_VIEW_P): New macro.
17998 (output_asm_line_debug_info): New.
17999 (struct var_loc_node): Add view.
18000 (add_AT_view_list, AT_loc_list): New.
18001 (add_var_loc_to_decl): Add view param. Test it against last.
18002 (new_loc_list): Add view params. Record them.
18003 (AT_loc_list_ptr): Handle loc and view lists.
18004 (view_list_to_loc_list_val_node): New.
18005 (print_dw_val): Handle dw_val_class_view_list.
18006 (size_of_die): Likewise.
18007 (value_format): Likewise.
18008 (loc_list_has_views): New.
18009 (gen_llsym): Set vl_symbol too.
18010 (maybe_gen_llsym, skip_loc_list_entry): New.
18011 (dwarf2out_maybe_output_loclist_view_pair): New.
18012 (output_loc_list): Output view list or entries too.
18013 (output_view_list_offset): New.
18014 (output_die): Handle dw_val_class_view_list.
18015 (output_dwarf_version): New.
18016 (output_compilation_unit_header): Use it.
18017 (output_skeleton_debug_sections): Likewise.
18018 (output_rnglists, output_line_info): Likewise.
18019 (output_pubnames, output_aranges): Update version comments.
18020 (output_one_line_info_table): Output view numbers in asm comments.
18021 (dw_loc_list): Determine current endview, pass it to new_loc_list.
18022 Call maybe_gen_llsym.
18023 (loc_list_from_tree_1): Adjust.
18024 (add_AT_location_description): Create view list attribute if
18025 needed, check it's absent otherwise.
18026 (convert_cfa_to_fb_loc_list): Adjust.
18027 (maybe_emit_file): Call output_asm_line_debug_info for test.
18028 (dwarf2out_var_location): Reset views as needed. Precompute
18029 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
18030 attribute. Set view.
18031 (new_line_info_table): Reset next view.
18032 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
18033 (dwarf2out_source_line): Likewise. Output view resets and labels to
18034 the assembler, or select appropriate line info opcodes.
18035 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
18036 (optimize_string_length): Catch it. Adjust.
18037 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
18038 dw_val_class_view_list, and remove it if no longer needed.
18039 (hash_loc_list): Hash view numbers.
18040 (loc_list_hasher::equal): Compare them.
18041 (optimize_location_lists): Check whether a view list symbol is
18042 needed, and whether the locview attribute is present, and
18043 whether they match. Remove the locview attribute if no longer
18045 (index_location_lists): Call skip_loc_list_entry for test.
18046 (dwarf2out_finish): Call output_asm_line_debug_info for test.
18047 Use output_dwarf_version.
18048 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
18049 (struct dw_val_node): Add val_view_list.
18050 * final.c (SEEN_NEXT_VIEW): New.
18051 (set_next_view_needed): New.
18052 (clear_next_view_needed): New.
18053 (maybe_output_next_view): New.
18054 (final_start_function): Rename to...
18055 (final_start_function_1): ... this. Take pointer to FIRST,
18056 add SEEN parameter. Emit param bindings in the initial view.
18057 (final_start_function): Reintroduce SEEN-less interface.
18058 (final): Rename to...
18059 (final_1): ... this. Take SEEN parameter. Output final pending
18060 next view at the end.
18061 (final): Reintroduce seen-less interface.
18062 (final_scan_insn): Output pending next view before switching
18063 sections or ending a block. Mark the next view as needed when
18064 outputting variable locations. Notify debug backend of section
18065 changes, and of location view changes.
18066 (rest_of_handle_final): Adjust.
18067 * toplev.c (process_options): Autodetect value for debug variable
18068 location views option. Warn on incompat5 without -gdwarf-5.
18069 * doc/invoke.texi (gvariable-location-views): New.
18070 (gvariable-location-views=incompat5): New.
18071 (gno-variable-location-views): New.
18073 2018-02-08 David Malcolm <dmalcolm@redhat.com>
18075 PR tree-optimization/84136
18076 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
18077 that the result of find_edge is non-NULL.
18079 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18082 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
18083 storing integer register in SImode. Fix cost of 256 and 512
18084 byte aligned SSE register store.
18086 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18088 * config/i386/i386.c (ix86_multiplication_cost): Fix
18089 multiplication cost for TARGET_AVX512DQ.
18091 2018-02-08 Marek Polacek <polacek@redhat.com>
18093 PR tree-optimization/84238
18094 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
18097 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18099 PR tree-optimization/84265
18100 * tree-vect-stmts.c (vectorizable_store): Don't treat
18101 VMAT_CONTIGUOUS accesses as grouped.
18102 (vectorizable_load): Likewise.
18104 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18106 PR tree-optimization/81635
18107 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
18108 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
18109 (test_round_for_mask): New functions.
18110 (wide_int_cc_tests): Call test_round_for_mask.
18111 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
18112 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
18113 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
18114 range returned by get_range_info.
18116 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18119 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
18120 * symtab.c: Include builtins.h
18121 (symtab_node::output_to_lto_symbol_table_p): Move here
18122 from lto-streamer-out.c:output_symbol_p.
18123 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
18124 (output_symbol_p): Move all logic to symtab.c
18125 (produce_symtab): Update.
18127 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18129 * config/s390/s390-opts.h (enum indirect_branch): Define.
18130 * config/s390/s390-protos.h (s390_return_addr_from_memory)
18131 (s390_indirect_branch_via_thunk)
18132 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
18133 (enum s390_indirect_branch_type): Define.
18134 * config/s390/s390.c (struct s390_frame_layout, struct
18135 machine_function): Remove.
18136 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
18137 (indirect_branch_table_label_no, indirect_branch_table_name):
18139 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
18140 (enum s390_indirect_branch_option): Define.
18141 (s390_return_addr_from_memory): New function.
18142 (s390_handle_string_attribute): New function.
18143 (s390_attribute_table): Add new attribute handler.
18144 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
18145 (s390_indirect_branch_via_thunk): New function.
18146 (s390_indirect_branch_via_inline_thunk): New function.
18147 (s390_function_ok_for_sibcall): When jumping via thunk disallow
18148 sibling call optimization for non z10 compiles.
18149 (s390_emit_call): Force indirect branch target to be a single
18150 register. Add r1 clobber for non-z10 compiles.
18151 (s390_emit_epilogue): Emit return jump via return_use expander.
18152 (s390_reorg): Handle JUMP_INSNs as execute targets.
18153 (s390_option_override_internal): Perform validity checks for the
18154 new command line options.
18155 (s390_indirect_branch_attrvalue): New function.
18156 (s390_indirect_branch_settings): New function.
18157 (s390_set_current_function): Invoke s390_indirect_branch_settings.
18158 (s390_output_indirect_thunk_function): New function.
18159 (s390_code_end): Implement target hook.
18160 (s390_case_values_threshold): Implement target hook.
18161 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
18163 * config/s390/s390.h (struct s390_frame_layout)
18164 (struct machine_function): Move here from s390.c.
18165 (TARGET_INDIRECT_BRANCH_NOBP_RET)
18166 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
18167 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
18168 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
18169 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
18170 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
18171 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
18172 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
18173 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
18174 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
18175 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
18176 (mnemonic attribute): Add values which aren't recognized
18178 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
18179 pattern for branch conversion. Fix mnemonic attribute.
18180 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
18181 indirect branch via thunk if requested.
18182 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
18183 ("*indirect_jump"): Disable for branch conversion using out of
18185 ("indirect_jump_via_thunk<mode>_z10")
18186 ("indirect_jump_via_thunk<mode>")
18187 ("indirect_jump_via_inlinethunk<mode>_z10")
18188 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18189 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18190 ("casesi_jump_via_inlinethunk<mode>_z10")
18191 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18192 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18193 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18194 ("*indirect2_jump"): Disable for branch conversion.
18195 ("casesi_jump"): Turn into expander and expand patterns for branch
18197 ("return_use"): New expander.
18198 ("*return"): Emit return via thunk and rename it to ...
18199 ("*return<mode>"): ... this one.
18200 * config/s390/s390.opt: Add new options and and enum for the
18203 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18205 * lra-constraints.c (match_reload): Unconditionally use
18206 gen_lowpart_SUBREG, rather than selecting between that
18207 and equivalent gen_rtx_SUBREG code.
18209 2018-02-08 Richard Biener <rguenther@suse.de>
18211 PR tree-optimization/84233
18212 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18213 changed flag instead of boguously re-using phi_inserted.
18215 2018-02-08 Martin Jambor <mjambor@suse.cz>
18217 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18218 static local variables.
18220 2018-02-08 Richard Biener <rguenther@suse.de>
18222 PR tree-optimization/84278
18223 * tree-vect-stmts.c (vectorizable_store): When looking for
18224 smaller vector types to perform grouped strided loads/stores
18225 make sure the mode is supported by the target.
18226 (vectorizable_load): Likewise.
18228 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18230 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18231 Increase LDP/STP opportunities by adding adjacent callee-saves.
18233 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18235 PR rtl-optimization/84068
18236 PR rtl-optimization/83459
18237 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18239 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18241 PR tree-optimization/84224
18242 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18243 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18244 non-zero arguments.
18246 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18249 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18250 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18251 count, remove one USE.
18253 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18255 * doc/install.texi (Configuration): Document the
18256 --with-long-double-format={ibm,ieee} PowerPC configuration
18260 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18261 Convert from define_expand to be define_insn_and_split. Rework
18262 float/double/_Float128 conversions to QI/HI/SImode to work with
18263 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18264 conversions to QI/HImode types did a store and then a load to
18265 truncate the value. For conversions to VSX registers, don't split
18266 the insn, instead emit the code directly. Use the code iterator
18267 any_fix to combine signed and unsigned conversions.
18268 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18269 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18270 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18271 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18272 (fix_<mode>di2_hw): Likewise.
18273 (fixuns_<mode>di2_hw): Likewise.
18274 (fix_<mode>si2_hw): Likewise.
18275 (fixuns_<mode>si2_hw): Likewise.
18276 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18277 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18278 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18279 fix<uns>_trunc<SFDF:mode>si2_p8.
18280 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18281 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18282 (fix<uns>_<mode>_mem): Likewise.
18283 (fctiw<u>z_<mode>_mem): Likewise.
18284 (fix<uns>_<mode>_mem): Likewise.
18285 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18286 the register allocator from doing a direct move to the GPRs to do
18287 a store, and instead use the ISA 3.0 store byte/half-word from
18288 vector register instruction. For IEEE 128-bit floating point,
18289 also optimize stores of 32-bit ints.
18290 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18292 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18294 * genextract.c (push_pathstr_operand): New function to support
18296 (walk_rtx): Call push_pathstr_operand.
18297 (print_path): Support [a-zA-Z].
18299 2018-02-07 Richard Biener <rguenther@suse.de>
18301 PR tree-optimization/84037
18302 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18303 (cse_and_gimplify_to_preheader): Declare.
18304 (vect_get_place_in_interleaving_chain): Likewise.
18305 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18307 (_loop_vec_info::~_loop_vec_info): Delete it.
18308 (cse_and_gimplify_to_preheader): New function.
18309 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18310 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18311 (vectorizable_load): Likewise. For grouped stores always base
18312 the IV on the first element.
18313 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18314 condition before gimplifying.
18316 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18318 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18319 *DIV_EXPR and *MOD_EXPR.
18321 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18324 * config/i386/i386.c (ix86_option_override_internal): Mask out
18325 the CF_SET bit when checking -fcf-protection.
18327 2018-02-07 Tom de Vries <tom@codesourcery.com>
18330 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18333 2018-02-07 Richard Biener <rguenther@suse.de>
18335 PR tree-optimization/84204
18336 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18339 PR tree-optimization/84205
18340 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18341 special-case isl_ast_op_zdiv_r.
18343 PR tree-optimization/84223
18344 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18345 Only add conditions from within the region.
18346 (gather_bbs::after_dom_children): Adjust.
18348 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18351 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18352 * config/avr/avr.md: Only post-reload split REG-REG moves if
18353 either register is GENERAL_REG_P.
18355 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18357 PR tree-optimization/84235
18358 * tree-ssa-scopedtables.c
18359 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18360 if the subtraction is performed in floating point type where NaNs are
18361 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18362 build 1. Formatting fix.
18364 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18367 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18368 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18369 and skip it regardless of bb boundaries. Use CALL_P macro,
18370 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18371 unnecessarily, formatting fix.
18373 2018-02-06 Michael Collison <michael.collison@arm.com>
18375 * config/arm/thumb2.md:
18376 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
18377 (*thumb_mov_notscc): Ditto.
18379 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
18382 * config/rs6000/rs6000.md (su code attribute): Use "u" for
18383 unsigned_fix, not "s".
18385 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18387 * configure.ac (gcc_fn_eh_frame_ro): New function.
18388 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
18389 correct .eh_frame permissions.
18390 * configure: Regenerate.
18392 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
18394 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
18395 irrelevant options.
18397 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18399 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18400 Display warning message for -mno-speculate-indirect-jumps.
18402 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
18404 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
18406 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
18408 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
18410 PR tree-optimization/84225
18411 * tree-eh.c (find_trapping_overflow): Only call
18412 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
18414 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18417 * config/i386/i386.c: Reimplement the check of possible options
18418 -mibt/-mshstk conbination. Change error messages.
18419 * doc/invoke.texi: Fix a typo: remove extra '='.
18421 2018-02-06 Marek Polacek <polacek@redhat.com>
18423 PR tree-optimization/84228
18424 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
18426 2018-02-06 Tamar Christina <tamar.christina@arm.com>
18429 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
18430 emitted arch directives.
18431 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
18432 __ARM_FEATURE_COPROC before changing architectures.
18434 2018-02-06 Richard Biener <rguenther@suse.de>
18436 * config/i386/i386.c (print_reg): Fix typo.
18437 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
18439 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
18441 * configure: Regenerate.
18443 2018-02-05 Martin Sebor <msebor@redhat.com>
18445 PR tree-optimization/83369
18446 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
18449 2018-02-05 Martin Liska <mliska@suse.cz>
18451 * doc/invoke.texi: Cherry-pick upstream r323995.
18453 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
18455 * ira.c (ira_init_register_move_cost): Adjust comment.
18457 2018-02-05 Martin Liska <mliska@suse.cz>
18459 PR gcov-profile/84137
18460 * doc/gcov.texi: Fix typo in documentation.
18462 2018-02-05 Martin Liska <mliska@suse.cz>
18464 PR gcov-profile/83879
18465 * doc/gcov.texi: Document necessity of --dynamic-list-data when
18466 using dlopen functionality.
18468 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
18470 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
18471 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
18472 _mm_maskz_range_ss, _mm_mask_range_round_ss,
18473 _mm_maskz_range_round_ss): New intrinsics.
18474 (__builtin_ia32_rangesd128_round)
18475 (__builtin_ia32_rangess128_round): Remove.
18476 (__builtin_ia32_rangesd128_mask_round,
18477 __builtin_ia32_rangess128_mask_round): New builtins.
18478 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
18479 __builtin_ia32_rangess128_round): Remove.
18480 (__builtin_ia32_rangesd128_mask_round,
18481 __builtin_ia32_rangess128_mask_round): New builtins.
18482 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
18483 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
18484 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
18485 "<round_saeonly_constraint>")): Changed to ...
18486 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
18487 "<round_saeonly_scalar_constraint>")): ... this.
18488 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
18489 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
18490 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
18491 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
18492 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
18494 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
18496 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
18498 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
18499 Remove all values except native, 8540 and 8548.
18501 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
18503 * config/i386/i386.c (ix86_output_function_return): Pass
18504 INVALID_REGNUM, instead of -1, as invalid register number to
18505 indirect_thunk_name and output_indirect_thunk.
18507 2018-02-02 Julia Koval <julia.koval@intel.com>
18509 * config.gcc: Add -march=icelake.
18510 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
18511 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
18512 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
18513 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
18514 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
18515 (processor_target_table): Add icelake.
18516 (ix86_option_override_internal): Handle new PTAs.
18517 (get_builtin_code_for_version): Handle icelake.
18518 (M_INTEL_COREI7_ICELAKE): New.
18519 (fold_builtin_cpu): Handle icelake.
18520 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
18521 * doc/invoke.texi: Add -march=icelake.
18523 2018-02-02 Julia Koval <julia.koval@intel.com>
18525 * config/i386/i386.c (ix86_option_override_internal): Change flags type
18526 to wide_int_bitmask.
18527 * wide-int-bitmask.h: New.
18529 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18532 * config/i386/i386.md: Replace Pmode with word_mode in
18533 builtin_setjmp_setup and builtin_longjmp to support x32.
18535 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
18539 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
18541 (rs6000_supported_cpu_names): New static variable.
18542 (linux_cpu_translation_table): Likewise.
18543 (elf_platform) <cpu>: Define new static variable and use it.
18544 Translate kernel AT_PLATFORM name to canonical name if needed.
18545 Error if platform name is unknown.
18547 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
18550 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
18552 2018-02-01 Jeff Law <law@redhat.com>
18555 * config/i386/i386.c (release_scratch_register_on_entry): Add new
18556 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
18557 the scratch if RELEASE_VIA_POP is false.
18558 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
18559 If we have to save a temporary register, decrement SIZE appropriately.
18560 Pass new arguments to release_scratch_register_on_entry.
18561 (ix86_adjust_stack_and_probe): Likewise.
18562 (ix86_emit_probe_stack_range): Pass new arguments to
18563 release_scratch_register_on_entry.
18565 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
18567 PR rtl-optimization/84157
18568 * combine.c (change_zero_ext): Use REG_P predicate in
18569 front of HARD_REGISTER_P predicate.
18571 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
18573 * config/avr/avr.c (avr_option_override): Move disabling of
18574 -fdelete-null-pointer-checks to...
18575 * common/config/avr/avr-common.c (avr_option_optimization_table):
18578 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18580 PR tree-optimization/81635
18581 * tree-data-ref.c (split_constant_offset_1): For types that
18582 wrap on overflow, try to use range info to prove that wrapping
18585 2018-02-01 Renlin Li <renlin.li@arm.com>
18588 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
18589 TAILCALL_ADDR_REGS.
18590 (aarch64_register_move_cost): Likewise.
18591 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
18592 TAILCALL_ADDR_REGS.
18593 (REG_CLASS_NAMES): Likewise.
18594 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
18595 TAILCALL_ADDR_REGS. Remove IP registers.
18596 * config/aarch64/aarch64.md (Ucs): Update register constraint.
18598 2018-02-01 Richard Biener <rguenther@suse.de>
18600 * domwalk.h (dom_walker::dom_walker): Add additional constructor
18601 for specifying RPO order and allow NULL for that.
18602 * domwalk.c (dom_walker::dom_walker): Likewise.
18603 (dom_walker::walk): Handle NULL RPO order.
18604 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
18606 (rewrite_update_dom_walker): Likewise.
18607 (mark_def_dom_walker): Likewise.
18609 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18611 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
18612 (aarch64_maybe_expand_sve_subreg_move): Declare.
18613 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
18614 * config/aarch64/predicates.md (aarch64_any_register_operand): New
18616 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
18617 that are semantically a reverse operation.
18618 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
18619 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
18620 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
18622 (aarch64_can_change_mode_class): For big-endian, forbid changes
18623 between two SVE modes if they have different element sizes.
18625 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18627 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
18628 the TImode handling for big-endian targets.
18630 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18632 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
18633 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
18635 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
18636 Remove BSWAP handing for big-endian targets and use the form of
18637 LD1RQ appropariate for the mode.
18639 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18641 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
18642 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
18643 duplicated element.
18645 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18648 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
18649 check for operands that need to go through aarch64_sve_reload_be.
18651 2018-02-01 Jakub Jelinek <jakub@redhat.com>
18653 PR tree-optimization/81661
18654 PR tree-optimization/84117
18655 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
18656 * tree-eh.c: Include gimplify.h.
18657 (find_trapping_overflow, replace_trapping_overflow,
18658 rewrite_to_non_trapping_overflow): New functions.
18659 * tree-vect-loop.c: Include tree-eh.h.
18660 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
18661 * tree-data-ref.c: Include tree-eh.h.
18662 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
18664 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
18666 PR rtl-optimization/84123
18667 * combine.c (change_zero_ext): Check if hard register satisfies
18668 can_change_dest_mode before calling gen_lowpart_SUBREG.
18670 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
18673 * ira.c (ira_init_register_move_cost): Remove assert.
18675 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18677 PR rtl-optimization/84071
18678 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
18679 * doc/tm.texi: Regenerate.
18681 2018-01-31 Richard Biener <rguenther@suse.de>
18683 PR tree-optimization/84132
18684 * tree-data-ref.c (analyze_miv_subscript): Properly
18685 check whether evolution_function_is_affine_multivariate_p
18686 before calling gcd_of_steps_may_divide_p.
18688 2018-01-31 Julia Koval <julia.koval@intel.com>
18691 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
18692 * config/i386/i386.md (rdpid_rex64) New.
18693 (rdpid): Make 32bit only.
18695 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
18698 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
18699 an IDENTIFIER_NODE for FUNCTION_TYPE's.
18701 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18704 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18706 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18708 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18710 PR rtl-optimization/84071
18711 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
18712 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
18714 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
18716 * config/arc/arc.c (arc_handle_aux_attribute): New function.
18717 (arc_attribute_table): Add 'aux' attribute.
18718 (arc_in_small_data_p): Consider aux like variables.
18719 (arc_is_aux_reg_p): New function.
18720 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
18721 (arc_get_aux_arg): New function.
18722 (prepare_move_operands): Handle aux-register access.
18723 (arc_handle_aux_attribute): New function.
18724 * doc/extend.texi (ARC Variable attributes): Add subsection.
18726 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
18728 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
18729 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
18730 (arc_attribute_table): Add 'uncached' attribute.
18731 (arc_print_operand): Print '.di' flag for uncached memory
18733 (arc_in_small_data_p): Do not consider for small data the uncached
18735 (arc_is_uncached_mem_p): New function.
18736 * config/arc/predicates.md (compact_store_memory_operand): Check
18737 for uncached memory accesses.
18738 (nonvol_nonimm_operand): Likewise.
18739 * doc/extend.texi (ARC Type Attribute): New subsection.
18741 2018-01-31 Jakub Jelinek <jakub@redhat.com>
18744 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
18745 falign-loops=): Add Optimization flag.
18747 2018-01-30 Jeff Law <law@redhat.com>
18750 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
18751 INT_REGISTERS_SAVED. Check it prior to calling
18752 get_scratch_register_on_entry.
18753 (ix86_adjust_stack_and_probe): Similarly.
18754 (ix86_emit_probe_stack_range): Similarly.
18755 (ix86_expand_prologue): Corresponding changes.
18757 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18760 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
18761 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
18763 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
18766 * lra-constraints.c (curr_insn_transform): Process AND in the
18769 2018-01-30 Jakub Jelinek <jakub@redhat.com>
18771 PR rtl-optimization/83986
18772 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
18773 dependence against last_pending_memory_flush in addition to
18774 pending_jump_insns.
18776 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
18778 PR tree-optimization/81611
18779 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
18782 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18785 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
18788 2018-01-30 Richard Biener <rguenther@suse.de>
18789 Jakub Jelinek <jakub@redhat.com>
18791 PR tree-optimization/84111
18792 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
18793 inner loops added during recursion, as they don't have up-to-date
18796 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18799 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
18800 (can_inline_edge_by_limits_p): ... here.
18801 (can_early_inline_edge_p, check_callers,
18802 update_caller_keys, update_callee_keys, recursive_inlining,
18803 add_new_edges_to_heap, speculation_useful_p,
18804 inline_small_functions,
18805 inline_small_functions, flatten_function,
18806 inline_to_all_callers_1): Update.
18808 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18810 * profile-count.c (profile_count::combine_with_ipa_count): Handle
18813 2018-01-30 Richard Biener <rguenther@suse.de>
18815 PR tree-optimization/83008
18816 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
18817 invariant and constant vector uses in stmts when they need
18818 more than one stmt.
18820 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18823 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
18824 * configure: Regenerate.
18826 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18828 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
18830 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
18831 Use gen_rtx_REG rather than gen_lowpart.
18833 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18835 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
18836 rather than 0 when creating partial subregs.
18838 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18840 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
18843 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
18846 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
18847 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
18848 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
18849 flags. This restores the settings used before the 2017-07-24.
18850 Turning off pre increment/decrement/modify allows IVOPTS to
18851 optimize DF/SF loops where the index is an int.
18853 2018-01-29 Richard Biener <rguenther@suse.de>
18854 Kelvin Nilsen <kelvin@gcc.gnu.org>
18857 * tree-vect-stmts.c (vectorizable_call): Don't call
18858 targetm.vectorize_builtin_md_vectorized_function if callee is
18861 2018-01-22 Carl Love <cel@us.ibm.com>
18863 * doc/extend.tex: Fix typo in second arg in
18864 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
18866 2018-01-29 Richard Biener <rguenther@suse.de>
18868 PR tree-optimization/84086
18869 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
18870 (flush_ssaname_freelist): When SSA names were released reset
18871 the SCEV hash table.
18873 2018-01-29 Richard Biener <rguenther@suse.de>
18875 PR tree-optimization/84057
18876 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
18877 removed paths when removing edges.
18879 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
18881 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
18882 -mfunction-return=@var{choice}.
18884 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18886 PR diagnostic/84034
18887 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
18888 Handle CR like TAB.
18889 (layout::print_source_line): Likewise.
18890 (test_get_line_width_without_trailing_whitespace): Add test cases.
18892 2018-01-27 Jakub Jelinek <jakub@redhat.com>
18894 PR middle-end/84040
18895 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
18898 2018-01-26 Jim Wilson <jimw@sifive.com>
18900 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
18902 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
18905 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18907 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
18908 and CMP + SUB-immediate -> SUBS.
18910 2018-01-26 Martin Sebor <msebor@redhat.com>
18912 PR tree-optimization/83896
18913 * tree-ssa-strlen.c (get_string_len): Rename...
18914 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
18915 Avoid assuming length is constant.
18916 (handle_char_store): Use HOST_WIDE_INT for string length.
18918 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
18921 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
18922 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
18924 2018-01-26 Richard Biener <rguenther@suse.de>
18926 PR rtl-optimization/84003
18927 * dse.c (record_store): Only record redundant stores when
18928 the earlier store aliases at least all accesses the later one does.
18930 2018-01-26 Jakub Jelinek <jakub@redhat.com>
18932 PR rtl-optimization/83985
18933 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
18934 REG_CFA_RESTORE insns.
18935 (delete_unmarked_insns): Don't ignore separate shrink wrapping
18936 REG_CFA_RESTORE insns here.
18939 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
18940 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
18942 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18944 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
18945 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
18946 (arc_init): Likewise.
18947 (arc_override_options): Likewise.
18948 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
18950 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
18952 * config/arc/arc.h (TARGET_DBNZ): Define.
18953 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
18954 properly set the tune attribute.
18955 (dbnz): Use TARGET_DBNZ guard.
18956 * config/arc/arc.opt (mtune): Add core3 option.
18958 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18960 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
18961 recognize new pic like addresses.
18962 (arc_delegitimize_address): Clean up.
18964 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18966 * config/arc/arc-arches.def: Option mrf16 valid for all
18968 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
18969 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
18970 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
18971 * config/arc/arc-tables.opt: Regenerate.
18972 * config/arc/arc.c (arc_conditional_register_usage): Handle
18973 reduced register file case.
18974 (arc_file_start): Set must have build attributes.
18975 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
18976 mrf16 option value.
18977 * config/arc/arc.opt (mrf16): Add new option.
18978 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
18979 * config/arc/genmultilib.awk: Handle new mrf16 option.
18980 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
18981 * config/arc/t-multilib: Regenerate.
18982 * doc/invoke.texi (ARC Options): Document mrf16 option.
18984 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18986 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
18987 * config/arc/arc.c (arc_handle_secure_attribute): New function.
18988 (arc_attribute_table): Add 'secure_call' attribute.
18989 (arc_print_operand): Print secure call operand.
18990 (arc_function_ok_for_sibcall): Don't optimize tail calls when
18992 (arc_is_secure_call_p): New function. * config/arc/arc.md
18993 (call_i): Add support for sjli instruction.
18994 (call_value_i): Likewise.
18995 * config/arc/constraints.md (Csc): New constraint.
18997 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18998 John Eric Martin <John.Martin@emmicro-us.com>
19000 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
19001 * config/arc/arc.c (_arc_jli_section): New struct.
19002 (arc_jli_section): New type.
19003 (rc_jli_sections): New static variable.
19004 (arc_handle_jli_attribute): New function.
19005 (arc_attribute_table): Add jli_always and jli_fixed attribute.
19006 (arc_file_end): New function.
19007 (TARGET_ASM_FILE_END): Define.
19008 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
19009 (arc_add_jli_section): New function.
19010 (jli_call_scan): Likewise.
19011 (arc_reorg): Call jli_call_scan.
19012 (arc_output_addsi): Remove 'S' from printing asm operand.
19013 (arc_is_jli_call_p): New function.
19014 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
19016 (movhi_insn): Likewise.
19017 (movsi_insn): Likewise.
19018 (movsi_set_cc_insn): Likewise.
19019 (loadqi_update): Likewise.
19020 (load_zeroextendqisi_update): Likewise.
19021 (load_signextendqisi_update): Likewise.
19022 (loadhi_update): Likewise.
19023 (load_zeroextendhisi_update): Likewise.
19024 (load_signextendhisi_update): Likewise.
19025 (loadsi_update): Likewise.
19026 (loadsf_update): Likewise.
19027 (movsicc_insn): Likewise.
19028 (bset_insn): Likewise.
19029 (bxor_insn): Likewise.
19030 (bclr_insn): Likewise.
19031 (bmsk_insn): Likewise.
19032 (bicsi3_insn): Likewise.
19033 (cmpsi_cc_c_insn): Likewise.
19034 (movsi_ne): Likewise.
19035 (movsi_cond_exec): Likewise.
19036 (clrsbsi2): Likewise.
19037 (norm_f): Likewise.
19047 (call_i): Remove 'S' asm letter, add jli instruction.
19048 (call_value_i): Likewise.
19049 * config/arc/arc.op (mjli-always): New option.
19050 * config/arc/constraints.md (Cji): New constraint.
19051 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
19053 (subsf3_fpx): Likewise.
19054 (mulsf3_fpx): Likewise.
19055 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
19057 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
19058 function attrbutes.
19059 * doc/invoke.texi (ARC): Document mjli-always option.
19061 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
19063 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
19064 avoid addition with 0 and use incw and decw where possible.
19066 2018-01-26 Richard Biener <rguenther@suse.de>
19068 PR tree-optimization/81082
19069 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
19070 association if it requires casting to unsigned.
19071 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
19072 from fold_plusminus_mult_expr to catch important cases late when
19073 range info is available.
19075 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19077 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
19078 * configure.ac (hidden_linkonce): New test.
19079 * configure: Regenerate.
19080 * config.in: Regenerate.
19082 2018-01-26 Julia Koval <julia.koval@intel.com>
19084 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
19085 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
19086 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
19087 _mm_mask_bitshuffle_epi64_mask): Fix type.
19088 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
19089 USI_FTYPE_V4DI_V4DI_USI): Remove.
19090 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
19091 __builtin_ia32_vpshufbitqmb256_mask,
19092 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
19093 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
19094 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
19096 2018-01-26 Alan Modra <amodra@gmail.com>
19099 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
19100 UNSPEC_VBPERMQ. Sort other unspecs.
19102 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
19104 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
19106 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
19108 PR middle-end/83055
19109 * predict.c (drop_profile): Do not push/pop cfun; update also
19111 (handle_missing_profiles): Fix logic looking for zero profiles.
19113 2018-01-25 Jakub Jelinek <jakub@redhat.com>
19115 PR middle-end/83977
19116 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
19117 on functions with #pragma omp declare simd or functions with simd
19119 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
19120 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
19121 Remove trailing \n from warning_at calls.
19123 2018-01-25 Tom de Vries <tom@codesourcery.com>
19126 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19127 for neutered workers.
19129 2018-01-24 Joseph Myers <joseph@codesourcery.com>
19132 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
19133 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
19135 2018-01-24 Jeff Law <law@redhat.com>
19138 * i386.c (get_probe_interval): Move to earlier point.
19139 (ix86_compute_frame_layout): If -fstack-clash-protection and
19140 the frame is larger than the probe interval, then use pushes
19141 to save registers rather than reg->mem moves.
19142 (ix86_expand_prologue): Remove conditional for int_registers_saved
19145 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
19148 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
19149 min/max for never referenced object.
19151 2018-01-24 Jakub Jelinek <jakub@redhat.com>
19153 PR middle-end/83977
19154 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
19156 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
19157 attributes from DECL_ATTRIBUTES (decl) without affecting
19158 DECL_ATTRIBUTES (current_function_decl).
19159 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
19160 functions with non-NULL DECL_ABSTRACT_ORIGIN.
19162 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
19164 PR tree-optimization/83979
19165 * fold-const.c (fold_comparison): Use constant_boolean_node
19166 instead of boolean_{true,false}_node.
19168 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
19170 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
19173 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19175 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19176 Simplify the clause that sets the length attribute.
19177 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19178 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
19179 clause that sets the length attribute.
19180 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19182 2018-01-24 Tom de Vries <tom@codesourcery.com>
19185 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
19186 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
19187 Add strict parameter.
19188 (prevent_branch_around_nothing): Insert dummy insn between branch to
19189 label and label with no ptx insn inbetween.
19190 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19192 2018-01-24 Tom de Vries <tom@codesourcery.com>
19195 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19196 for neutered threads in warp.
19197 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19199 2018-01-24 Richard Biener <rguenther@suse.de>
19201 PR tree-optimization/83176
19202 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19205 2018-01-24 Richard Biener <rguenther@suse.de>
19207 PR tree-optimization/82819
19208 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19209 code generating pluses that are no-ops in the target precision.
19211 2018-01-24 Richard Biener <rguenther@suse.de>
19213 PR middle-end/84000
19214 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19216 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19218 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19219 to merge probabilities.
19220 * predict.c (probably_never_executed): Also mark as cold functions
19221 with global 0 profile and guessed local profile.
19222 * profile-count.c (profile_probability::combine_with_count): New
19224 * profile-count.h (profile_probability::operator*,
19225 profile_probability::operator*=, profile_probability::operator/,
19226 profile_probability::operator/=): Reduce precision to adjusted
19227 and set value to guessed on contradictory divisions.
19228 (profile_probability::combine_with_freq): Remove.
19229 (profile_probability::combine_wiht_count): Declare.
19230 (profile_count::force_nonzero):: Set to adjusted.
19231 (profile_count::probability_in):: Set quality to adjusted.
19232 * tree-ssa-tail-merge.c (replace_block_by): Use
19233 combine_with_count.
19235 2018-01-23 Andrew Waterman <andrew@sifive.com>
19236 Jim Wilson <jimw@sifive.com>
19238 * config/riscv/riscv.c (riscv_stack_boundary): New.
19239 (riscv_option_override): Set riscv_stack_boundary. Handle
19240 riscv_preferred_stack_boundary_arg.
19241 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19242 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19243 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19244 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19245 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19246 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19248 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19251 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19252 of struct ix86_frame.
19253 (ix86_expand_epilogue): Likewise. Add a local variable for
19254 the reg_save_offset field in struct ix86_frame.
19256 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19258 PR tree-optimization/82604
19259 * tree-loop-distribution.c (enum partition_kind): New enum item
19260 PKIND_PARTIAL_MEMSET.
19261 (partition_builtin_p): Support above new enum item.
19262 (generate_code_for_partition): Ditto.
19263 (compute_access_range): Differentiate cases that equality can be
19264 proven at all loops, the innermost loops or no loops.
19265 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19266 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19267 (finalize_partitions, distribute_loop): Don't fuse partition of
19268 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19269 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19270 parloop is enabled.
19272 2018-01-23 Martin Liska <mliska@suse.cz>
19274 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19275 order to ignore the predictor.
19276 (PRED_POLYMORPHIC_CALL): Likewise.
19277 (PRED_RECURSIVE_CALL): Likewise.
19279 2018-01-23 Martin Liska <mliska@suse.cz>
19281 * tree-profile.c (tree_profiling): Print function header to
19282 aware reader which function we are working on.
19283 * value-prof.c (gimple_find_values_to_profile): Do not print
19284 not interesting value histograms.
19286 2018-01-23 Martin Liska <mliska@suse.cz>
19288 * profile-count.h (enum profile_quality): Add
19289 profile_uninitialized as the first value. Do not number values
19290 as they are zero based.
19291 (profile_count::verify): Update sanity check.
19292 (profile_probability::verify): Likewise.
19294 2018-01-23 Nathan Sidwell <nathan@acm.org>
19296 * doc/invoke.texi (ffor-scope): Deprecate.
19298 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19300 PR tree-optimization/83510
19301 * domwalk.c (set_all_edges_as_executable): New function.
19302 (dom_walker::dom_walker): Convert bool param
19303 "skip_unreachable_blocks" to enum reachability. Move setup of
19304 edge flags to set_all_edges_as_executable and only do it when
19305 reachability is REACHABLE_BLOCKS.
19306 * domwalk.h (enum dom_walker::reachability): New enum.
19307 (dom_walker::dom_walker): Convert bool param
19308 "skip_unreachable_blocks" to enum reachability.
19309 (set_all_edges_as_executable): New decl.
19310 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19311 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19313 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19314 but converting true to REACHABLE_BLOCKS.
19315 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19317 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19318 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19319 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19321 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19322 if check_all_array_refs will be called.
19324 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19326 * tree.c (selftest::test_location_wrappers): Add more test
19329 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19331 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19332 (selftest::test_bit_in_range): Likewise.
19334 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19337 * doc/sourcebuild.texi (vect_float): Say that the selector
19338 only describes the situation when -funsafe-math-optimizations is on.
19339 (vect_float_strict): Document.
19341 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19343 PR tree-optimization/83965
19344 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19345 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19346 instead of checking only for a reduction.
19347 (vect_recog_widen_sum_pattern): Likewise.
19349 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19351 * predict.c (probably_never_executed): Only use precise profile info.
19352 (compute_function_frequency): Skip after inlining hack since we now
19353 have quality checking.
19355 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19357 * profile-count.h (profile_probability::very_unlikely,
19358 profile_probability::unlikely, profile_probability::even): Set
19359 precision to guessed.
19361 2018-01-23 Richard Biener <rguenther@suse.de>
19363 PR tree-optimization/83963
19364 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19365 Properly terminate dominator walk when crossing the exit edge not
19366 when visiting its source block.
19368 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19371 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19372 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19374 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19376 PR tree-optimization/83957
19377 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
19378 semicolon after for body surrounded by braces.
19380 PR tree-optimization/83081
19381 * profile-count.h (profile_probability::split): New method.
19382 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
19383 Use profile_probability::split.
19384 (do_compare_rtx_and_jump): Fix adjustment of probabilities
19385 when splitting a single conditional jump into 2.
19387 2018-01-22 David Malcolm <dmalcolm@redhat.com>
19389 PR tree-optimization/69452
19390 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
19393 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19395 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
19396 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
19397 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
19399 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19401 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
19403 * config/rl78/rl78.md (movdi): New define_expand.
19404 * config/rl78/rl78.c (rl78_split_movdi): New function.
19406 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
19409 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
19411 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
19412 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
19413 128-bit to produce an UNSPEC move to get the double word with the
19414 signbit and then a shift directly to do signbit.
19415 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
19416 implementation with a new version that just does either a direct
19417 move or a regular move. Move memory interface to separate insns.
19418 Move insns so they are next to the expander.
19419 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
19420 with signbit move. Split big and little endian case.
19421 (signbit<mode>2_dm_mem_le): Likewise.
19422 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
19423 (signbit<mode>2_dm2): Likewise.
19425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19427 * config/rl78/rl78.md (anddi3): New define_expand.
19429 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19431 * config/rl78/rl78.md (umindi3): New define_expand.
19433 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19435 * config/rl78/rl78.md (smindi3): New define_expand.
19437 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19439 * config/rl78/rl78.md (smaxdi3): New define_expand.
19441 2018-01-22 Carl Love <cel@us.ibm.com>
19443 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
19444 LVX_V1TI): Add macro expansion.
19445 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
19446 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
19447 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
19448 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
19449 Change check to determine if the instruction is a byte reversing
19450 entry. Fix typo in comment.
19451 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
19452 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
19453 Add def_builtin calls for new builtins.
19454 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
19455 Add define_insn expansion.
19457 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19459 * config/rl78/rl78.md (umaxdi3): New define_expand.
19461 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19463 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
19464 for non-QImode registers.
19466 2018-01-22 Richard Biener <rguenther@suse.de>
19468 PR tree-optimization/83963
19469 * graphite-scop-detection.c (scop_detection::get_sese): Delay
19470 including the loop exit block.
19471 (scop_detection::merge_sese): Likewise.
19472 (scop_detection::add_scop): Do it here instead.
19474 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19476 * doc/sourcebuild.texi (arm_softfloat): Document.
19478 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
19481 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
19482 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
19483 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
19485 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19486 David Edelsohn <dje.gcc@gmail.com>
19489 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19490 Change "crset eq" to "crset 2".
19491 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19492 (*call_indirect_aix<mode>_nospec): Likewise.
19493 (*call_value_indirect_aix<mode>_nospec): Likewise.
19494 (*call_indirect_elfv2<mode>_nospec): Likewise.
19495 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
19496 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
19497 change assembly output from . to $.
19498 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19499 (indirect_jump<mode>_nospec): Change assembly output from . to $.
19500 (*tablejump<mode>_internal1_nospec): Likewise.
19502 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
19505 * config/sh/sh_optimize_sett_clrt.cc:
19506 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
19508 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
19510 PR tree-optimization/83940
19511 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
19512 offset_dt to vect_constant_def rather than vect_unknown_def_type.
19513 (vect_check_load_store_mask): Add a mask_dt_out parameter and
19514 use it to pass back the definition type.
19515 (vect_check_store_rhs): Likewise rhs_dt_out.
19516 (vect_build_gather_load_calls): Add a mask_dt argument and use
19517 it instead of a call to vect_is_simple_use.
19518 (vectorizable_store): Update calls to vect_check_load_store_mask
19519 and vect_check_store_rhs. Use the dt returned by the latter instead
19520 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
19521 instead of calls to vect_is_simple_use. Pass the scalar rather
19522 than the vector operand to vect_is_simple_use when handling
19523 second and subsequent copies of an rhs value.
19524 (vectorizable_load): Update calls to vect_check_load_store_mask
19525 and vect_build_gather_load_calls. Use the cached mask_dt and
19526 gs_info.offset_dt instead of calls to vect_is_simple_use.
19528 2018-01-20 Jakub Jelinek <jakub@redhat.com>
19530 PR middle-end/83945
19531 * tree-emutls.c: Include gimplify.h.
19532 (lower_emutls_2): New function.
19533 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
19534 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
19535 it before further processing.
19538 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
19539 UINTVAL (trueop1) instead of INTVAL (op1).
19541 2018-01-19 Jakub Jelinek <jakub@redhat.com>
19545 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
19546 INCOMING_FRAME_SP_OFFSET if not defined.
19547 (scan_trace): Add ENTRY argument. If true and
19548 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
19549 emit a note to adjust the CFA offset.
19550 (create_cfi_notes): Adjust scan_trace callers.
19551 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
19552 INCOMING_FRAME_SP_OFFSET in the CIE.
19553 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
19554 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
19556 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
19557 * doc/tm.texi: Regenerated.
19559 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19561 PR rtl-optimization/83147
19562 * lra-constraints.c (remove_inheritance_pseudos): Use
19563 lra_substitute_pseudo_within_insn.
19565 2018-01-19 Tom de Vries <tom@codesourcery.com>
19566 Cesar Philippidis <cesar@codesourcery.com>
19569 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
19571 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
19574 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
19575 spaces for function labels.
19577 2018-01-19 Martin Liska <mliska@suse.cz>
19579 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
19580 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
19581 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
19582 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
19583 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
19584 (PRED_CONST_RETURN): Change from 69 to 65.
19585 (PRED_NULL_RETURN): Change from 91 to 71.
19586 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
19587 (PRED_LOOP_GUARD): Change from 66 to 73.
19589 2018-01-19 Martin Liska <mliska@suse.cz>
19591 * predict.c (predict_insn_def): Add new assert.
19592 (struct branch_predictor): Change type to signed integer.
19593 (test_prediction_value_range): Amend test to cover
19594 PROB_UNINITIALIZED.
19595 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
19596 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
19597 (PRED_LOOP_ITERATIONS_MAX): Likewise.
19598 (PRED_LOOP_IV_COMPARE): Likewise.
19599 * predict.h (PROB_UNINITIALIZED): Define new constant.
19601 2018-01-19 Martin Liska <mliska@suse.cz>
19603 * predict.c (dump_prediction): Add new format for
19604 analyze_brprob.py script which is enabled with -details
19606 * profile-count.h (precise_p): New function.
19608 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19610 PR tree-optimization/83922
19611 * tree-vect-loop.c (vect_verify_full_masking): Return false if
19612 there are no statements that need masking.
19613 (vect_active_double_reduction_p): New function.
19614 (vect_analyze_loop_operations): Use it when handling phis that
19615 are not in the loop header.
19617 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19619 PR tree-optimization/83914
19620 * tree-vect-loop.c (vectorizable_induction): Don't convert
19621 init_expr or apply the peeling adjustment for inductions
19622 that are nested within the vectorized loop.
19624 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19626 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
19629 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19633 * function.h (gimplify_parameters): Add gimple_seq * argument.
19634 * function.c: Include gimple.h and options.h.
19635 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
19636 for the added local temporaries if needed.
19637 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
19638 if there are any parameter cleanups, wrap whole body into a
19639 try/finally with the cleanups.
19641 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
19644 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
19645 Use GET_MODE_CLASS for scalar floating point.
19647 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19651 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
19652 Fix call of call_cgraph_insertion_hooks.
19654 2018-01-18 Martin Sebor <msebor@redhat.com>
19656 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
19658 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19661 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
19664 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
19667 * common.opt: (-ffile-prefix-map): New option.
19668 * opts.c (common_handle_option): Defer it.
19669 * opts-global.c (handle_common_deferred_options): Handle it.
19670 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
19671 * file-prefix-map.h: New file.
19672 (remap_debug_filename, add_debug_prefix_map): ...here.
19673 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
19674 * final.c (debug_prefix_map, add_debug_prefix_map
19675 remap_debug_filename): Move to...
19676 * file-prefix-map.c: New file.
19677 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
19678 generalize, get rid of alloca(), use strrchr() instead of strchr().
19679 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
19680 Implement in terms of add_prefix_map().
19681 (remap_macro_filename, remap_debug_filename): Implement in term of
19683 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
19684 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
19685 * dbxout.c: Include file-prefix-map.h.
19686 * varasm.c: Likewise.
19687 * vmsdbgout.c: Likewise.
19688 * xcoffout.c: Likewise.
19689 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
19690 * doc/cppopts.texi (-fmacro-prefix-map): Document.
19691 * doc/invoke.texi (-ffile-prefix-map): Document.
19692 (-fdebug-prefix-map): Update description.
19694 2018-01-18 Martin Liska <mliska@suse.cz>
19696 * config/i386/i386.c (indirect_thunk_name): Document that also
19698 (output_indirect_thunk): Document why both instructions
19699 (pause and lfence) are generated.
19701 2018-01-18 Richard Biener <rguenther@suse.de>
19703 PR tree-optimization/83887
19704 * graphite-scop-detection.c
19705 (scop_detection::get_nearest_dom_with_single_entry): Remove.
19706 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
19707 (scop_detection::merge_sese): Re-implement with a flood-fill
19708 algorithm that properly finds a SESE region if it exists.
19710 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19713 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
19714 pointer_diff optimizations use view_convert instead of convert.
19716 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19718 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19719 Generate different code for -mno-speculate-indirect-jumps.
19720 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19721 (*call_indirect_aix<mode>): Disable for
19722 -mno-speculate-indirect-jumps.
19723 (*call_indirect_aix<mode>_nospec): New define_insn.
19724 (*call_value_indirect_aix<mode>): Disable for
19725 -mno-speculate-indirect-jumps.
19726 (*call_value_indirect_aix<mode>_nospec): New define_insn.
19727 (*sibcall_nonlocal_sysv<mode>): Generate different code for
19728 -mno-speculate-indirect-jumps.
19729 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19731 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
19733 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
19734 long double type, set the flags for noting the default long double
19735 type, even if we don't pass or return a long double type.
19737 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
19740 * ipa-inline.c (flatten_function): Do not overwrite final inlining
19743 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
19745 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
19746 support for merge[hl].
19747 (fold_mergehl_helper): New helper function.
19748 (tree-vector-builder.h): New #include for tree_vector_builder usage.
19749 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
19750 (altivec_vmrglw_direct): Add xxmrglw insn.
19752 2018-01-17 Andrew Waterman <andrew@sifive.com>
19754 * config/riscv/riscv.c (riscv_conditional_register_usage): If
19755 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
19757 2018-01-17 David Malcolm <dmalcolm@redhat.com>
19760 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
19761 call the lto_location_cache before reading the
19762 DECL_SOURCE_LOCATION of the types.
19764 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
19765 Richard Sandiford <richard.sandiford@linaro.org>
19767 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
19768 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
19769 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
19770 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
19771 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
19773 * config/aarch64/constraints.md (aarch64_movti_operand):
19775 * config/aarch64/predicates.md (Uti): Add new constraint.
19777 2018-01-17 Carl Love <cel@us.ibm.com>
19779 * config/rs6000/vsx.md (define_expand xl_len_r,
19780 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
19781 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
19783 (define_expand, define_insn): Move the shift left from the
19784 define_insn to the define_expand for lxvl and stxvl instructions.
19785 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
19786 and XL_LEN_R definitions to PURE.
19788 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
19790 * config/i386/i386.c (indirect_thunk_name): Declare regno
19791 as unsigned int. Compare regno with INVALID_REGNUM.
19792 (output_indirect_thunk): Ditto.
19793 (output_indirect_thunk_function): Ditto.
19794 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
19795 in the call to output_indirect_thunk_function.
19797 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
19799 PR middle-end/83884
19800 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
19801 rather than the size of inner_type to determine the stack slot size
19802 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
19804 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
19807 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
19810 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
19812 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
19813 endian Linux systems to optionally enable multilibs for selecting
19814 the long double type if the user configured an explicit type.
19815 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
19816 have no long double multilibs if not defined.
19817 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
19818 warn if the user used -mabi={ieee,ibm}longdouble and we built
19819 multilibs for long double.
19820 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
19821 appropriate multilib option.
19822 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
19824 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
19825 for building long double multilibs.
19826 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
19828 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
19830 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
19833 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
19835 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
19838 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
19841 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
19844 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19846 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
19847 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
19849 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
19851 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
19852 different rtl trees depending on TARGET_64BIT.
19853 (rs6000_gen_lvx): Likewise.
19855 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19857 * config/visium/visium.md (nop): Tweak comment.
19858 (hazard_nop): Likewise.
19860 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19862 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
19863 -mspeculate-indirect-jumps.
19864 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
19865 for -mno-speculate-indirect-jumps.
19866 (*call_indirect_elfv2<mode>_nospec): New define_insn.
19867 (*call_value_indirect_elfv2<mode>): Disable for
19868 -mno-speculate-indirect-jumps.
19869 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
19870 (indirect_jump): Emit different RTL for
19871 -mno-speculate-indirect-jumps.
19872 (*indirect_jump<mode>): Disable for
19873 -mno-speculate-indirect-jumps.
19874 (*indirect_jump<mode>_nospec): New define_insn.
19875 (tablejump): Emit different RTL for
19876 -mno-speculate-indirect-jumps.
19877 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
19878 (tablejumpsi_nospec): New define_expand.
19879 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
19880 (tablejumpdi_nospec): New define_expand.
19881 (*tablejump<mode>_internal1): Disable for
19882 -mno-speculate-indirect-jumps.
19883 (*tablejump<mode>_internal1_nospec): New define_insn.
19884 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
19887 2018-01-16 Artyom Skrobov tyomitch@gmail.com
19889 * caller-save.c (insert_save): Drop unnecessary parameter. All
19892 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19893 Richard Biener <rguenth@suse.de>
19896 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
19897 return early, inline manually is_gimple_sizepos. Make sure if we
19898 call gimplify_expr we don't end up with a gimple constant.
19899 * tree.c (variably_modified_type_p): Don't return true for
19900 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
19901 * gimplify.h (is_gimple_sizepos): Remove.
19903 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19905 PR tree-optimization/83857
19906 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
19907 vectorizable_live_operation for pure SLP statements.
19908 (vectorizable_live_operation): Handle PHIs.
19910 2018-01-16 Richard Biener <rguenther@suse.de>
19912 PR tree-optimization/83867
19913 * tree-vect-stmts.c (vect_transform_stmt): Precompute
19914 nested_in_vect_loop_p since the scalar stmt may get invalidated.
19916 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19919 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
19920 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
19921 If off is not INTEGER_CST, issue a may not be aligned warning
19922 rather than isn't aligned. Use isn%'t rather than isn't.
19923 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
19925 <case MULT_EXPR>: Improve the case when bottom and one of the
19926 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
19927 operand, in that case check if the other operand is multiple of
19928 bottom divided by the INTEGER_CST operand.
19930 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19933 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
19934 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
19935 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
19936 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
19937 * config/pa/pa.c (pa_function_arg_advance): Likewise.
19938 (pa_function_arg, pa_arg_partial_bytes): Likewise.
19939 (pa_function_arg_size): New function.
19941 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19943 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
19944 in a separate statement.
19946 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19948 PR tree-optimization/83847
19949 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
19950 group gathers and scatters.
19952 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19954 PR rtl-optimization/86620
19955 * params.def (max-sched-ready-insns): Bump minimum value to 1.
19957 PR rtl-optimization/83213
19958 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
19959 to last if both are JUMP_INSNs.
19961 PR tree-optimization/83843
19962 * gimple-ssa-store-merging.c
19963 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
19964 store_immediate_info for bswap/nop orig_stores.
19966 2018-01-15 Andrew Waterman <andrew@sifive.com>
19968 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
19970 <UDIV>: Increase cost if !TARGET_DIV.
19972 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
19974 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
19975 (define_attr "cr_logical_3op"): New.
19976 (cceq_ior_compare): Adjust.
19977 (cceq_ior_compare_complement): Adjust.
19978 (*cceq_rev_compare): Adjust.
19979 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
19980 (is_cracked_insn): Adjust.
19981 (insn_must_be_first_in_group): Adjust.
19982 * config/rs6000/40x.md: Adjust.
19983 * config/rs6000/440.md: Adjust.
19984 * config/rs6000/476.md: Adjust.
19985 * config/rs6000/601.md: Adjust.
19986 * config/rs6000/603.md: Adjust.
19987 * config/rs6000/6xx.md: Adjust.
19988 * config/rs6000/7450.md: Adjust.
19989 * config/rs6000/7xx.md: Adjust.
19990 * config/rs6000/8540.md: Adjust.
19991 * config/rs6000/cell.md: Adjust.
19992 * config/rs6000/e300c2c3.md: Adjust.
19993 * config/rs6000/e500mc.md: Adjust.
19994 * config/rs6000/e500mc64.md: Adjust.
19995 * config/rs6000/e5500.md: Adjust.
19996 * config/rs6000/e6500.md: Adjust.
19997 * config/rs6000/mpc.md: Adjust.
19998 * config/rs6000/power4.md: Adjust.
19999 * config/rs6000/power5.md: Adjust.
20000 * config/rs6000/power6.md: Adjust.
20001 * config/rs6000/power7.md: Adjust.
20002 * config/rs6000/power8.md: Adjust.
20003 * config/rs6000/power9.md: Adjust.
20004 * config/rs6000/rs64.md: Adjust.
20005 * config/rs6000/titan.md: Adjust.
20007 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20009 * config/i386/predicates.md (indirect_branch_operand): Rewrite
20010 ix86_indirect_branch_register logic.
20012 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20014 * config/i386/constraints.md (Bs): Update
20015 ix86_indirect_branch_register check. Don't check
20016 ix86_indirect_branch_register with GOT_memory_operand.
20018 * config/i386/predicates.md (GOT_memory_operand): Don't check
20019 ix86_indirect_branch_register here.
20020 (GOT32_symbol_operand): Likewise.
20022 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20024 * config/i386/predicates.md (constant_call_address_operand):
20025 Rewrite ix86_indirect_branch_register logic.
20026 (sibcall_insn_operand): Likewise.
20028 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20030 * config/i386/constraints.md (Bs): Replace
20031 ix86_indirect_branch_thunk_register with
20032 ix86_indirect_branch_register.
20034 * config/i386/i386.md (indirect_jump): Likewise.
20035 (tablejump): Likewise.
20036 (*sibcall_memory): Likewise.
20037 (*sibcall_value_memory): Likewise.
20038 Peepholes of indirect call and jump via memory: Likewise.
20039 * config/i386/i386.opt: Likewise.
20040 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20041 (GOT_memory_operand): Likewise.
20042 (call_insn_operand): Likewise.
20043 (sibcall_insn_operand): Likewise.
20044 (GOT32_symbol_operand): Likewise.
20046 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20048 PR middle-end/83837
20049 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
20050 type rather than type addr's type points to.
20051 (expand_omp_atomic_mutex): Likewise.
20052 (expand_omp_atomic): Likewise.
20054 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20057 * config/i386/i386.c (output_indirect_thunk_function): Use
20058 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
20059 for __x86_return_thunk.
20061 2018-01-15 Richard Biener <rguenther@suse.de>
20063 PR middle-end/83850
20064 * expmed.c (extract_bit_field_1): Fix typo.
20066 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20069 * config/arm/iterators.md (VF): New mode iterator.
20070 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
20071 Remove integer-related logic from pattern.
20072 (neon_vabd<mode>_3): Likewise.
20074 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20076 PR middle-end/82694
20077 * common.opt (fstrict-overflow): No longer an alias.
20078 (fwrapv-pointer): New option.
20079 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
20080 also for pointer types based on flag_wrapv_pointer.
20081 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
20082 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
20083 opts->x_flag_wrapv got set.
20084 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
20085 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
20086 POINTER_TYPE_OVERFLOW_UNDEFINED.
20087 * match.pd: Likewise in address comparison pattern.
20088 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
20090 2018-01-15 Richard Biener <rguenther@suse.de>
20093 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
20094 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
20095 Reset type names to their identifier if their TYPE_DECL doesn't
20096 have linkage (and thus is used for ODR and devirt).
20097 (save_debug_info_for_decl): Remove.
20098 (save_debug_info_for_type): Likewise.
20099 (add_tree_to_fld_list): Adjust.
20100 * tree-pretty-print.c (dump_generic_node): Make dumping of
20101 type names more robust.
20103 2018-01-15 Richard Biener <rguenther@suse.de>
20105 * BASE-VER: Bump to 8.0.1.
20107 2018-01-14 Martin Sebor <msebor@redhat.com>
20110 * builtins.c (check_access): Avoid warning when the no-warning bit
20113 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
20115 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
20116 * ira-color (allocno_hard_regs_compare): Likewise.
20118 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
20121 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
20122 Use .pushsection/.popsection.
20124 2018-01-14 Martin Sebor <msebor@redhat.com>
20127 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
20129 2018-01-14 Jakub Jelinek <jakub@redhat.com>
20131 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
20132 entry from extra_headers.
20133 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
20134 extra_headers, make the list bitwise identical to the i?86-*-* one.
20136 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20138 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
20139 -mcmodel=large with -mindirect-branch=thunk,
20140 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
20141 -mfunction-return=thunk-extern.
20142 * doc/invoke.texi: Document -mcmodel=large is incompatible with
20143 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
20144 -mfunction-return=thunk and -mfunction-return=thunk-extern.
20146 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20148 * config/i386/i386.c (print_reg): Print the name of the full
20149 integer register without '%'.
20150 (ix86_print_operand): Handle 'V'.
20151 * doc/extend.texi: Document 'V' modifier.
20153 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20155 * config/i386/constraints.md (Bs): Disallow memory operand for
20156 -mindirect-branch-register.
20158 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20159 (GOT_memory_operand): Likewise.
20160 (call_insn_operand): Likewise.
20161 (sibcall_insn_operand): Likewise.
20162 (GOT32_symbol_operand): Likewise.
20163 * config/i386/i386.md (indirect_jump): Call convert_memory_address
20164 for -mindirect-branch-register.
20165 (tablejump): Likewise.
20166 (*sibcall_memory): Likewise.
20167 (*sibcall_value_memory): Likewise.
20168 Disallow peepholes of indirect call and jump via memory for
20169 -mindirect-branch-register.
20170 (*call_pop): Replace m with Bw.
20171 (*call_value_pop): Likewise.
20172 (*sibcall_pop_memory): Replace m with Bs.
20173 * config/i386/i386.opt (mindirect-branch-register): New option.
20174 * doc/invoke.texi: Document -mindirect-branch-register option.
20176 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20178 * config/i386/i386-protos.h (ix86_output_function_return): New.
20179 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
20180 set function_return_type.
20181 (indirect_thunk_name): Add ret_p to indicate thunk for function
20183 (output_indirect_thunk_function): Pass false to
20184 indirect_thunk_name.
20185 (ix86_output_indirect_branch_via_reg): Likewise.
20186 (ix86_output_indirect_branch_via_push): Likewise.
20187 (output_indirect_thunk_function): Create alias for function
20188 return thunk if regno < 0.
20189 (ix86_output_function_return): New function.
20190 (ix86_handle_fndecl_attribute): Handle function_return.
20191 (ix86_attribute_table): Add function_return.
20192 * config/i386/i386.h (machine_function): Add
20193 function_return_type.
20194 * config/i386/i386.md (simple_return_internal): Use
20195 ix86_output_function_return.
20196 (simple_return_internal_long): Likewise.
20197 * config/i386/i386.opt (mfunction-return=): New option.
20198 (indirect_branch): Mention -mfunction-return=.
20199 * doc/extend.texi: Document function_return function attribute.
20200 * doc/invoke.texi: Document -mfunction-return= option.
20202 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20204 * config/i386/i386-opts.h (indirect_branch): New.
20205 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20206 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20207 with local indirect jump when converting indirect call and jump.
20208 (ix86_set_indirect_branch_type): New.
20209 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20210 (indirectlabelno): New.
20211 (indirect_thunk_needed): Likewise.
20212 (indirect_thunk_bnd_needed): Likewise.
20213 (indirect_thunks_used): Likewise.
20214 (indirect_thunks_bnd_used): Likewise.
20215 (INDIRECT_LABEL): Likewise.
20216 (indirect_thunk_name): Likewise.
20217 (output_indirect_thunk): Likewise.
20218 (output_indirect_thunk_function): Likewise.
20219 (ix86_output_indirect_branch_via_reg): Likewise.
20220 (ix86_output_indirect_branch_via_push): Likewise.
20221 (ix86_output_indirect_branch): Likewise.
20222 (ix86_output_indirect_jmp): Likewise.
20223 (ix86_code_end): Call output_indirect_thunk_function if needed.
20224 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20226 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20227 (ix86_attribute_table): Add indirect_branch.
20228 * config/i386/i386.h (machine_function): Add indirect_branch_type
20229 and has_local_indirect_jump.
20230 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20232 (tablejump): Likewise.
20233 (*indirect_jump): Use ix86_output_indirect_jmp.
20234 (*tablejump_1): Likewise.
20235 (simple_return_indirect_internal): Likewise.
20236 * config/i386/i386.opt (mindirect-branch=): New option.
20237 (indirect_branch): New.
20240 (thunk-inline): Likewise.
20241 (thunk-extern): Likewise.
20242 * doc/extend.texi: Document indirect_branch function attribute.
20243 * doc/invoke.texi: Document -mindirect-branch= option.
20245 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20248 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20250 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20252 * ipa-inline.c (want_inline_small_function_p): Return false if
20253 inlining has already failed with CIF_FINAL_ERROR.
20254 (update_caller_keys): Call want_inline_small_function_p before
20256 (update_callee_keys): Likewise.
20258 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20260 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20262 (rs6000_quadword_masked_address_p): Likewise.
20263 (quad_aligned_load_p): Likewise.
20264 (quad_aligned_store_p): Likewise.
20265 (const_load_sequence_p): Add comment to describe the outer-most loop.
20266 (mimic_memory_attributes_and_flags): New function.
20267 (rs6000_gen_stvx): Likewise.
20268 (replace_swapped_aligned_store): Likewise.
20269 (rs6000_gen_lvx): Likewise.
20270 (replace_swapped_aligned_load): Likewise.
20271 (replace_swapped_load_constant): Capitalize argument name in
20272 comment describing this function.
20273 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20274 and stores that access quad-word aligned addresses and replace
20275 with stvx or lvx instructions when appropriate.
20276 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20277 New function prototype.
20278 (rs6000_quadword_masked_address_p): Likewise.
20279 (rs6000_gen_lvx): Likewise.
20280 (rs6000_gen_stvx): Likewise.
20281 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20282 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20283 when memory address is aligned.
20284 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20285 this split to select lvx instruction when memory address is aligned.
20286 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20287 instruction when memory address is aligned.
20288 (*vsx_le_perm_load_v16qi): Likewise.
20289 (four unnamed splitters): Modify to select the stvx instruction
20290 when memory is aligned.
20292 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20294 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20295 which appears in the queue multiple times.
20297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20298 Alan Hayward <alan.hayward@arm.com>
20299 David Sherwood <david.sherwood@arm.com>
20301 * tree-vectorizer.h (vec_lower_bound): New structure.
20302 (_loop_vec_info): Add check_nonzero and lower_bounds.
20303 (LOOP_VINFO_CHECK_NONZERO): New macro.
20304 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20305 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20306 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20307 fields. Make seg_len the distance travelled, not including the
20309 (dr_direction_indicator): Declare.
20310 (dr_zero_step_indicator): Likewise.
20311 (dr_known_forward_stride_p): Likewise.
20312 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20314 (runtime_alias_check_p): Allow runtime alias checks with
20316 (operator ==): Compare access_size and align.
20317 (prune_runtime_alias_test_list): Rework for new distinction between
20318 the access_size and seg_len.
20319 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20321 (get_segment_min_max): New function.
20322 (create_intersect_range_checks): Use it.
20323 (dr_step_indicator): New function.
20324 (dr_direction_indicator): Likewise.
20325 (dr_zero_step_indicator): Likewise.
20326 (dr_known_forward_stride_p): Likewise.
20327 * tree-loop-distribution.c (data_ref_segment_size): Return
20328 DR_STEP * (niters - 1).
20329 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20331 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20332 (vect_preserves_scalar_order_p): New function, split out from...
20333 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20334 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20335 (vect_vfa_access_size): New function.
20336 (vect_vfa_align): Likewise.
20337 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20338 (dump_lower_bound): New function.
20339 (vect_check_lower_bound): Likewise.
20340 (vect_small_gap_p): Likewise.
20341 (vectorizable_with_step_bound_p): Likewise.
20342 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20343 depencies if the vectorization factor is 1. Convert the checks
20344 for nonzero steps into checks on the bounds of DR_STEP. Try using
20345 a bunds check for variable steps if the minimum required step is
20346 relatively small. Update calls to the dr_with_seg_len
20347 constructor and to vect_compile_time_alias.
20348 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20350 (vect_loop_versioning): Call it.
20351 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20353 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20355 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20356 Alan Hayward <alan.hayward@arm.com>
20357 David Sherwood <david.sherwood@arm.com>
20359 * doc/sourcebuild.texi (vect_scatter_store): Document.
20360 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20362 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20364 * genopinit.c (main): Add supports_vec_scatter_store and
20365 supports_vec_scatter_store_cached to target_optabs.
20366 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20367 IFN_MASK_SCATTER_STORE.
20368 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20370 * internal-fn.h (internal_store_fn_p): Declare.
20371 (internal_fn_stored_value_index): Likewise.
20372 * internal-fn.c (scatter_store_direct): New macro.
20373 (expand_scatter_store_optab_fn): New function.
20374 (direct_scatter_store_optab_supported_p): New macro.
20375 (internal_store_fn_p): New function.
20376 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
20377 IFN_MASK_SCATTER_STORE.
20378 (internal_fn_mask_index): Likewise.
20379 (internal_fn_stored_value_index): New function.
20380 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
20381 for scatter stores.
20382 * optabs-query.h (supports_vec_scatter_store_p): Declare.
20383 * optabs-query.c (supports_vec_scatter_store_p): New function.
20384 * tree-vectorizer.h (vect_get_store_rhs): Declare.
20385 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
20386 true for scatter stores.
20387 (vect_gather_scatter_fn_p): Handle scatter stores too.
20388 (vect_check_gather_scatter): Consider using scatter stores if
20389 supports_vec_scatter_store_p.
20390 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
20391 scatter stores too.
20392 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20393 internal_fn_stored_value_index.
20394 (check_load_store_masking): Handle scatter stores too.
20395 (vect_get_store_rhs): Make public.
20396 (vectorizable_call): Use internal_store_fn_p.
20397 (vectorizable_store): Handle scatter store internal functions.
20398 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
20399 when deciding whether the end of the group has been reached.
20400 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
20401 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
20402 (mask_scatter_store<mode>): New insns.
20404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20405 Alan Hayward <alan.hayward@arm.com>
20406 David Sherwood <david.sherwood@arm.com>
20408 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
20409 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
20410 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
20412 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
20413 Use vect_truncate_gather_scatter_offset if we can't treat the
20414 operation as a normal gather load or scatter store.
20415 (get_group_load_store_type): Take the gather_scatter_info
20416 as argument. Try using a gather load or scatter store for
20417 single-element groups.
20418 (get_load_store_type): Update calls to get_group_load_store_type
20419 and vect_use_strided_gather_scatters_p.
20421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20422 Alan Hayward <alan.hayward@arm.com>
20423 David Sherwood <david.sherwood@arm.com>
20425 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
20426 optional tree argument.
20427 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
20429 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
20430 but continue to use the current value as a fallback.
20431 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
20432 to compare the updates.
20433 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
20434 (get_load_store_type): Use it when handling a strided access.
20435 (vect_get_strided_load_store_ops): New function.
20436 (vect_get_data_ptr_increment): Likewise.
20437 (vectorizable_load): Handle strided gather loads. Always pass
20438 a step to vect_create_data_ref_ptr and bump_vector_ptr.
20440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20441 Alan Hayward <alan.hayward@arm.com>
20442 David Sherwood <david.sherwood@arm.com>
20444 * doc/md.texi (gather_load@var{m}): Document.
20445 (mask_gather_load@var{m}): Likewise.
20446 * genopinit.c (main): Add supports_vec_gather_load and
20447 supports_vec_gather_load_cached to target_optabs.
20448 * optabs-tree.c (init_tree_optimization_optabs): Use
20449 ggc_cleared_alloc to allocate target_optabs.
20450 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
20451 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
20453 * internal-fn.h (internal_load_fn_p): Declare.
20454 (internal_gather_scatter_fn_p): Likewise.
20455 (internal_fn_mask_index): Likewise.
20456 (internal_gather_scatter_fn_supported_p): Likewise.
20457 * internal-fn.c (gather_load_direct): New macro.
20458 (expand_gather_load_optab_fn): New function.
20459 (direct_gather_load_optab_supported_p): New macro.
20460 (direct_internal_fn_optab): New function.
20461 (internal_load_fn_p): Likewise.
20462 (internal_gather_scatter_fn_p): Likewise.
20463 (internal_fn_mask_index): Likewise.
20464 (internal_gather_scatter_fn_supported_p): Likewise.
20465 * optabs-query.c (supports_at_least_one_mode_p): New function.
20466 (supports_vec_gather_load_p): Likewise.
20467 * optabs-query.h (supports_vec_gather_load_p): Declare.
20468 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
20469 and memory_type field.
20470 (NUM_PATTERNS): Bump to 15.
20471 * tree-vect-data-refs.c: Include internal-fn.h.
20472 (vect_gather_scatter_fn_p): New function.
20473 (vect_describe_gather_scatter_call): Likewise.
20474 (vect_check_gather_scatter): Try using internal functions for
20475 gather loads. Recognize existing calls to a gather load function.
20476 (vect_analyze_data_refs): Consider using gather loads if
20477 supports_vec_gather_load_p.
20478 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
20479 (vect_get_gather_scatter_offset_type): Likewise.
20480 (vect_convert_mask_for_vectype): Likewise.
20481 (vect_add_conversion_to_patterm): Likewise.
20482 (vect_try_gather_scatter_pattern): Likewise.
20483 (vect_recog_gather_scatter_pattern): New pattern recognizer.
20484 (vect_vect_recog_func_ptrs): Add it.
20485 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20486 internal_fn_mask_index and internal_gather_scatter_fn_p.
20487 (check_load_store_masking): Take the gather_scatter_info as an
20488 argument and handle gather loads.
20489 (vect_get_gather_scatter_ops): New function.
20490 (vectorizable_call): Check internal_load_fn_p.
20491 (vectorizable_load): Likewise. Handle gather load internal
20493 (vectorizable_store): Update call to check_load_store_masking.
20494 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
20495 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
20496 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
20497 (aarch64_gather_scale_operand_d): New predicates.
20498 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
20499 (mask_gather_load<mode>): New insns.
20501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20502 Alan Hayward <alan.hayward@arm.com>
20503 David Sherwood <david.sherwood@arm.com>
20505 * optabs.def (fold_left_plus_optab): New optab.
20506 * doc/md.texi (fold_left_plus_@var{m}): Document.
20507 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
20508 * internal-fn.c (fold_left_direct): Define.
20509 (expand_fold_left_optab_fn): Likewise.
20510 (direct_fold_left_optab_supported_p): Likewise.
20511 * fold-const-call.c (fold_const_fold_left): New function.
20512 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
20513 * tree-parloops.c (valid_reduction_p): New function.
20514 (gather_scalar_reductions): Use it.
20515 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
20516 (vect_finish_replace_stmt): Declare.
20517 * tree-vect-loop.c (fold_left_reduction_fn): New function.
20518 (needs_fold_left_reduction_p): New function, split out from...
20519 (vect_is_simple_reduction): ...here. Accept reductions that
20520 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
20521 (vect_force_simple_reduction): Also store the reduction type in
20522 the assignment's STMT_VINFO_REDUC_TYPE.
20523 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
20524 (merge_with_identity): New function.
20525 (vect_expand_fold_left): Likewise.
20526 (vectorize_fold_left_reduction): Likewise.
20527 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
20528 scalar phi in place for it. Check for target support and reject
20529 cases that would reassociate the operation. Defer the transform
20530 phase to vectorize_fold_left_reduction.
20531 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
20532 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
20533 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
20535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20537 * tree-if-conv.c (predicate_mem_writes): Remove redundant
20538 call to ifc_temp_var.
20540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20541 Alan Hayward <alan.hayward@arm.com>
20542 David Sherwood <david.sherwood@arm.com>
20544 * target.def (legitimize_address_displacement): Take the original
20545 offset as a poly_int.
20546 * targhooks.h (default_legitimize_address_displacement): Update
20548 * targhooks.c (default_legitimize_address_displacement): Likewise.
20549 * doc/tm.texi: Regenerate.
20550 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
20551 as an argument, moving assert of ad->disp == ad->disp_term to...
20552 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
20553 Try calling targetm.legitimize_address_displacement before expanding
20554 the address rather than afterwards, and adjust for the new interface.
20555 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
20556 Match the new hook interface. Handle SVE addresses.
20557 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
20558 new hook interface.
20560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20562 * Makefile.in (OBJS): Add early-remat.o.
20563 * target.def (select_early_remat_modes): New hook.
20564 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
20565 * doc/tm.texi: Regenerate.
20566 * targhooks.h (default_select_early_remat_modes): Declare.
20567 * targhooks.c (default_select_early_remat_modes): New function.
20568 * timevar.def (TV_EARLY_REMAT): New timevar.
20569 * passes.def (pass_early_remat): New pass.
20570 * tree-pass.h (make_pass_early_remat): Declare.
20571 * early-remat.c: New file.
20572 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
20574 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
20576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20577 Alan Hayward <alan.hayward@arm.com>
20578 David Sherwood <david.sherwood@arm.com>
20580 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
20581 vfm1 with a bound_epilog parameter.
20582 (vect_do_peeling): Update calls accordingly, and move the prologue
20583 call earlier in the function. Treat the base bound_epilog as 0 for
20584 fully-masked loops and retain vf - 1 for other loops. Add 1 to
20585 this base when peeling for gaps.
20586 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
20587 with fully-masked loops.
20588 (vect_estimate_min_profitable_iters): Handle the single peeled
20589 iteration in that case.
20591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20592 Alan Hayward <alan.hayward@arm.com>
20593 David Sherwood <david.sherwood@arm.com>
20595 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
20596 single-element interleaving even if the size is not a power of 2.
20597 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
20598 accesses for single-element interleaving if the group size is
20601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20602 Alan Hayward <alan.hayward@arm.com>
20603 David Sherwood <david.sherwood@arm.com>
20605 * doc/md.texi (fold_extract_last_@var{m}): Document.
20606 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
20607 * optabs.def (fold_extract_last_optab): New optab.
20608 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
20609 * internal-fn.c (fold_extract_direct): New macro.
20610 (expand_fold_extract_optab_fn): Likewise.
20611 (direct_fold_extract_optab_supported_p): Likewise.
20612 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
20613 * tree-vect-loop.c (vect_model_reduction_cost): Handle
20614 EXTRACT_LAST_REDUCTION.
20615 (get_initial_def_for_reduction): Do not create an initial vector
20616 for EXTRACT_LAST_REDUCTION reductions.
20617 (vectorizable_reduction): Leave the scalar phi in place for
20618 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
20619 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
20620 epilogue code for EXTRACT_LAST_REDUCTION and defer the
20621 transform phase to vectorizable_condition.
20622 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
20624 (vect_finish_stmt_generation): ...here.
20625 (vect_finish_replace_stmt): New function.
20626 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
20627 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
20629 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
20631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20632 Alan Hayward <alan.hayward@arm.com>
20633 David Sherwood <david.sherwood@arm.com>
20635 * doc/md.texi (extract_last_@var{m}): Document.
20636 * optabs.def (extract_last_optab): New optab.
20637 * internal-fn.def (EXTRACT_LAST): New internal function.
20638 * internal-fn.c (cond_unary_direct): New macro.
20639 (expand_cond_unary_optab_fn): Likewise.
20640 (direct_cond_unary_optab_supported_p): Likewise.
20641 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
20642 loops using EXTRACT_LAST.
20643 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
20644 (extract_last_<mode>): ...this optab.
20645 (vec_extract<mode><Vel>): Update accordingly.
20647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20648 Alan Hayward <alan.hayward@arm.com>
20649 David Sherwood <david.sherwood@arm.com>
20651 * target.def (empty_mask_is_expensive): New hook.
20652 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
20653 * doc/tm.texi: Regenerate.
20654 * targhooks.h (default_empty_mask_is_expensive): Declare.
20655 * targhooks.c (default_empty_mask_is_expensive): New function.
20656 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
20657 if the target says that empty masks are expensive.
20658 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
20660 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
20662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20663 Alan Hayward <alan.hayward@arm.com>
20664 David Sherwood <david.sherwood@arm.com>
20666 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
20667 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
20668 (vect_use_loop_mask_for_alignment_p): New function.
20669 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
20670 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
20671 niters_skip argument. Make sure that the first niters_skip elements
20672 of the first iteration are inactive.
20673 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
20674 Update call to vect_set_loop_masks_directly.
20675 (get_misalign_in_elems): New function, split out from...
20676 (vect_gen_prolog_loop_niters): ...here.
20677 (vect_update_init_of_dr): Take a code argument that specifies whether
20678 the adjustment should be added or subtracted.
20679 (vect_update_init_of_drs): Likewise.
20680 (vect_prepare_for_masked_peels): New function.
20681 (vect_do_peeling): Skip prologue peeling if we're using a mask
20682 instead. Update call to vect_update_inits_of_drs.
20683 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20685 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
20686 alignment. Do not include the number of peeled iterations in
20687 the minimum threshold in that case.
20688 (vectorizable_induction): Adjust the start value down by
20689 LOOP_VINFO_MASK_SKIP_NITERS iterations.
20690 (vect_transform_loop): Call vect_prepare_for_masked_peels.
20691 Take the number of skipped iterations into account when calculating
20693 * tree-vect-stmts.c (vect_gen_while_not): New function.
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/sourcebuild.texi (vect_fully_masked): Document.
20700 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
20701 default value to 0.
20702 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
20704 (vect_analyze_loop_2): ...here. Don't check the vectorization
20705 factor against the number of loop iterations if the loop is
20708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20709 Alan Hayward <alan.hayward@arm.com>
20710 David Sherwood <david.sherwood@arm.com>
20712 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
20713 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
20714 (dump_groups): Update accordingly.
20715 (iv_use::mem_type): New member variable.
20716 (address_p): New function.
20717 (record_use): Add a mem_type argument and initialize the new
20719 (record_group_use): Add a mem_type argument. Use address_p.
20720 Remove obsolete null checks of base_object. Update call to record_use.
20721 (find_interesting_uses_op): Update call to record_group_use.
20722 (find_interesting_uses_cond): Likewise.
20723 (find_interesting_uses_address): Likewise.
20724 (get_mem_type_for_internal_fn): New function.
20725 (find_address_like_use): Likewise.
20726 (find_interesting_uses_stmt): Try find_address_like_use before
20727 calling find_interesting_uses_op.
20728 (addr_offset_valid_p): Use the iv mem_type field as the type
20729 of the addressed memory.
20730 (add_autoinc_candidates): Likewise.
20731 (get_address_cost): Likewise.
20732 (split_small_address_groups_p): Use address_p.
20733 (split_address_groups): Likewise.
20734 (add_iv_candidate_for_use): Likewise.
20735 (autoinc_possible_for_pair): Likewise.
20736 (rewrite_groups): Likewise.
20737 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
20738 (determine_group_iv_cost): Update after split of USE_ADDRESS.
20739 (get_alias_ptr_type_for_ptr_address): New function.
20740 (rewrite_use_address): Rewrite address uses in calls that were
20741 identified by find_address_like_use.
20743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20744 Alan Hayward <alan.hayward@arm.com>
20745 David Sherwood <david.sherwood@arm.com>
20747 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
20749 * gimple-expr.h (is_gimple_addressable: Likewise.
20750 * gimple-expr.c (is_gimple_address): Likewise.
20751 * internal-fn.c (expand_call_mem_ref): New function.
20752 (expand_mask_load_optab_fn): Use it.
20753 (expand_mask_store_optab_fn): Likewise.
20755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20756 Alan Hayward <alan.hayward@arm.com>
20757 David Sherwood <david.sherwood@arm.com>
20759 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
20760 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
20761 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
20762 (cond_umax@var{mode}): Document.
20763 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
20764 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
20765 (cond_umin_optab, cond_umax_optab): New optabs.
20766 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
20767 (COND_IOR, COND_XOR): New internal functions.
20768 * internal-fn.h (get_conditional_internal_fn): Declare.
20769 * internal-fn.c (cond_binary_direct): New macro.
20770 (expand_cond_binary_optab_fn): Likewise.
20771 (direct_cond_binary_optab_supported_p): Likewise.
20772 (get_conditional_internal_fn): New function.
20773 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
20774 Cope with reduction statements that are vectorized as calls rather
20776 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
20777 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
20778 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
20779 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
20780 (UNSPEC_COND_EOR): New unspecs.
20781 (optab): Add mappings for them.
20782 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
20783 (sve_int_op, sve_fp_op): New int attributes.
20785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20786 Alan Hayward <alan.hayward@arm.com>
20787 David Sherwood <david.sherwood@arm.com>
20789 * optabs.def (while_ult_optab): New optab.
20790 * doc/md.texi (while_ult@var{m}@var{n}): Document.
20791 * internal-fn.def (WHILE_ULT): New internal function.
20792 * internal-fn.h (direct_internal_fn_supported_p): New override
20793 that takes two types as argument.
20794 * internal-fn.c (while_direct): New macro.
20795 (expand_while_optab_fn): New function.
20796 (convert_optab_supported_p): Likewise.
20797 (direct_while_optab_supported_p): New macro.
20798 * wide-int.h (wi::udiv_ceil): New function.
20799 * tree-vectorizer.h (rgroup_masks): New structure.
20800 (vec_loop_masks): New typedef.
20801 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
20802 and fully_masked_p.
20803 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
20804 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
20805 (vect_max_vf): New function.
20806 (slpeel_make_loop_iterate_ntimes): Delete.
20807 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
20808 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
20809 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
20810 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
20811 internal-fn.h, stor-layout.h and optabs-query.h.
20812 (vect_set_loop_mask): New function.
20813 (add_preheader_seq): Likewise.
20814 (add_header_seq): Likewise.
20815 (interleave_supported_p): Likewise.
20816 (vect_maybe_permute_loop_masks): Likewise.
20817 (vect_set_loop_masks_directly): Likewise.
20818 (vect_set_loop_condition_masked): Likewise.
20819 (vect_set_loop_condition_unmasked): New function, split out from
20820 slpeel_make_loop_iterate_ntimes.
20821 (slpeel_make_loop_iterate_ntimes): Rename to..
20822 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
20823 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
20824 (vect_do_peeling): Update call accordingly.
20825 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
20827 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20828 mask_compare_type, can_fully_mask_p and fully_masked_p.
20829 (release_vec_loop_masks): New function.
20830 (_loop_vec_info): Use it to free the loop masks.
20831 (can_produce_all_loop_masks_p): New function.
20832 (vect_get_max_nscalars_per_iter): Likewise.
20833 (vect_verify_full_masking): Likewise.
20834 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
20835 retries, and free the mask rgroups before retrying. Check loop-wide
20836 reasons for disallowing fully-masked loops. Make the final decision
20837 about whether use a fully-masked loop or not.
20838 (vect_estimate_min_profitable_iters): Do not assume that peeling
20839 for the number of iterations will be needed for fully-masked loops.
20840 (vectorizable_reduction): Disable fully-masked loops.
20841 (vectorizable_live_operation): Likewise.
20842 (vect_halve_mask_nunits): New function.
20843 (vect_double_mask_nunits): Likewise.
20844 (vect_record_loop_mask): Likewise.
20845 (vect_get_loop_mask): Likewise.
20846 (vect_transform_loop): Handle the case in which the final loop
20847 iteration might handle a partial vector. Call vect_set_loop_condition
20848 instead of slpeel_make_loop_iterate_ntimes.
20849 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
20850 (check_load_store_masking): New function.
20851 (prepare_load_store_mask): Likewise.
20852 (vectorizable_store): Handle fully-masked loops.
20853 (vectorizable_load): Likewise.
20854 (supportable_widening_operation): Use vect_halve_mask_nunits for
20856 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
20857 (vect_gen_while): New function.
20858 * config/aarch64/aarch64.md (umax<mode>3): New expander.
20859 (aarch64_uqdec<mode>): New insn.
20861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20862 Alan Hayward <alan.hayward@arm.com>
20863 David Sherwood <david.sherwood@arm.com>
20865 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
20866 (reduc_xor_scal_optab): New optabs.
20867 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
20868 (reduc_xor_scal_@var{m}): Document.
20869 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
20870 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
20871 internal functions.
20872 * fold-const-call.c (fold_const_call): Handle them.
20873 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
20874 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
20875 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
20876 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
20877 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
20878 (UNSPEC_XORV): New unspecs.
20879 (optab): Add entries for them.
20880 (BITWISEV): New int iterator.
20881 (bit_reduc_op): New int attributes.
20883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20884 Alan Hayward <alan.hayward@arm.com>
20885 David Sherwood <david.sherwood@arm.com>
20887 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
20888 * internal-fn.def (VEC_SHL_INSERT): New internal function.
20889 * optabs.def (vec_shl_insert_optab): New optab.
20890 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
20891 (duplicate_and_interleave): Likewise.
20892 * tree-vect-loop.c: Include internal-fn.h.
20893 (neutral_op_for_slp_reduction): New function, split out from
20894 get_initial_defs_for_reduction.
20895 (get_initial_def_for_reduction): Handle option 2 for variable-length
20896 vectors by loading the neutral value into a vector and then shifting
20897 the initial value into element 0.
20898 (get_initial_defs_for_reduction): Replace the code argument with
20899 the neutral value calculated by neutral_op_for_slp_reduction.
20900 Use gimple_build_vector for constant-length vectors.
20901 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
20902 but the first group_size elements have a neutral value.
20903 Use duplicate_and_interleave otherwise.
20904 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
20905 Update call to get_initial_defs_for_reduction. Handle SLP
20906 reductions for variable-length vectors by creating one vector
20907 result for each scalar result, with the elements associated
20908 with other scalar results stubbed out with the neutral value.
20909 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
20910 Require IFN_VEC_SHL_INSERT for double reductions on
20911 variable-length vectors, or SLP reductions that have
20912 a neutral value. Require can_duplicate_and_interleave_p
20913 support for variable-length unchained SLP reductions if there
20914 is no neutral value, such as for MIN/MAX reductions. Also require
20915 the number of vector elements to be a multiple of the number of
20916 SLP statements when doing variable-length unchained SLP reductions.
20917 Update call to vect_create_epilog_for_reduction.
20918 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
20919 and remove initial values.
20920 (duplicate_and_interleave): Make public.
20921 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
20922 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
20924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20925 Alan Hayward <alan.hayward@arm.com>
20926 David Sherwood <david.sherwood@arm.com>
20928 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
20929 (can_duplicate_and_interleave_p): New function.
20930 (vect_get_and_check_slp_defs): Take the vector of statements
20931 rather than just the current one. Remove excess parentheses.
20932 Restriction rejectinon of vect_constant_def and vect_external_def
20933 for variable-length vectors to boolean types, or types for which
20934 can_duplicate_and_interleave_p is false.
20935 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
20936 (duplicate_and_interleave): New function.
20937 (vect_get_constant_vectors): Use gimple_build_vector for
20938 constant-length vectors and suitable variable-length constant
20939 vectors. Use duplicate_and_interleave for other variable-length
20940 vectors. Don't defer the update when inserting new statements.
20942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20943 Alan Hayward <alan.hayward@arm.com>
20944 David Sherwood <david.sherwood@arm.com>
20946 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
20947 min_profitable_iters doesn't go negative.
20949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20950 Alan Hayward <alan.hayward@arm.com>
20951 David Sherwood <david.sherwood@arm.com>
20953 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
20954 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
20955 * optabs.def (vec_mask_load_lanes_optab): New optab.
20956 (vec_mask_store_lanes_optab): Likewise.
20957 * internal-fn.def (MASK_LOAD_LANES): New internal function.
20958 (MASK_STORE_LANES): Likewise.
20959 * internal-fn.c (mask_load_lanes_direct): New macro.
20960 (mask_store_lanes_direct): Likewise.
20961 (expand_mask_load_optab_fn): Handle masked operations.
20962 (expand_mask_load_lanes_optab_fn): New macro.
20963 (expand_mask_store_optab_fn): Handle masked operations.
20964 (expand_mask_store_lanes_optab_fn): New macro.
20965 (direct_mask_load_lanes_optab_supported_p): Likewise.
20966 (direct_mask_store_lanes_optab_supported_p): Likewise.
20967 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
20969 (vect_load_lanes_supported): Likewise.
20970 * tree-vect-data-refs.c (strip_conversion): New function.
20971 (can_group_stmts_p): Likewise.
20972 (vect_analyze_data_ref_accesses): Use it instead of checking
20973 for a pair of assignments.
20974 (vect_store_lanes_supported): Take a masked_p parameter.
20975 (vect_load_lanes_supported): Likewise.
20976 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
20977 vect_store_lanes_supported and vect_load_lanes_supported.
20978 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
20979 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
20980 parameter. Don't allow gaps for masked accesses.
20981 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
20982 and vect_load_lanes_supported.
20983 (get_load_store_type): Take a masked_p parameter and update
20984 call to get_group_load_store_type.
20985 (vectorizable_store): Update call to get_load_store_type.
20986 Handle IFN_MASK_STORE_LANES.
20987 (vectorizable_load): Update call to get_load_store_type.
20988 Handle IFN_MASK_LOAD_LANES.
20990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20991 Alan Hayward <alan.hayward@arm.com>
20992 David Sherwood <david.sherwood@arm.com>
20994 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
20996 * config/aarch64/aarch64-protos.h
20997 (aarch64_sve_struct_memory_operand_p): Declare.
20998 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
20999 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
21000 (VPRED, vpred): Handle SVE structure modes.
21001 * config/aarch64/constraints.md (Utx): New constraint.
21002 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
21003 (aarch64_sve_struct_nonimmediate_operand): New predicates.
21004 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
21005 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
21006 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
21007 structure modes. Split into pieces after RA.
21008 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
21009 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
21011 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
21012 SVE structure modes.
21013 (aarch64_classify_address): Likewise.
21014 (sizetochar): Move earlier in file.
21015 (aarch64_print_operand): Handle SVE register lists.
21016 (aarch64_array_mode): New function.
21017 (aarch64_sve_struct_memory_operand_p): Likewise.
21018 (TARGET_ARRAY_MODE): Redefine.
21020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21021 Alan Hayward <alan.hayward@arm.com>
21022 David Sherwood <david.sherwood@arm.com>
21024 * target.def (array_mode): New target hook.
21025 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
21026 * doc/tm.texi: Regenerate.
21027 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
21028 * hooks.c (hook_optmode_mode_uhwi_none): New function.
21029 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
21030 targetm.array_mode.
21031 * stor-layout.c (mode_for_array): Likewise. Support polynomial
21034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21035 Alan Hayward <alan.hayward@arm.com>
21036 David Sherwood <david.sherwood@arm.com>
21038 * fold-const.c (fold_binary_loc): Check the argument types
21039 rather than the result type when testing for a vector operation.
21041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21043 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
21044 * doc/tm.texi: Regenerate.
21046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21047 Alan Hayward <alan.hayward@arm.com>
21048 David Sherwood <david.sherwood@arm.com>
21050 * doc/invoke.texi (-msve-vector-bits=): Document new option.
21051 (sve): Document new AArch64 extension.
21052 * doc/md.texi (w): Extend the description of the AArch64
21053 constraint to include SVE vectors.
21054 (Upl, Upa): Document new AArch64 predicate constraints.
21055 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
21057 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
21058 (msve-vector-bits=): New option.
21059 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
21060 SVE when these are disabled.
21061 (sve): New extension.
21062 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
21063 modes. Adjust their number of units based on aarch64_sve_vg.
21064 (MAX_BITSIZE_MODE_ANY_MODE): Define.
21065 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
21066 aarch64_addr_query_type.
21067 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
21068 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
21069 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
21070 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
21071 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
21072 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
21073 (aarch64_simd_imm_zero_p): Delete.
21074 (aarch64_check_zero_based_sve_index_immediate): Declare.
21075 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21076 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21077 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21078 (aarch64_sve_float_mul_immediate_p): Likewise.
21079 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21080 rather than an rtx.
21081 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
21082 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
21083 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
21084 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
21085 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
21086 (aarch64_regmode_natural_size): Likewise.
21087 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
21088 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
21090 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
21091 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
21092 for VG and the SVE predicate registers.
21093 (V_ALIASES): Add a "z"-prefixed alias.
21094 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
21095 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
21096 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
21097 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
21098 (REG_CLASS_NAMES): Add entries for them.
21099 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
21100 and the predicate registers.
21101 (aarch64_sve_vg): Declare.
21102 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
21103 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
21104 (REGMODE_NATURAL_SIZE): Define.
21105 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
21107 * config/aarch64/aarch64.c: Include cfgrtl.h.
21108 (simd_immediate_info): Add a constructor for series vectors,
21109 and an associated step field.
21110 (aarch64_sve_vg): New variable.
21111 (aarch64_dbx_register_number): Handle VG and the predicate registers.
21112 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
21113 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
21114 (VEC_ANY_DATA, VEC_STRUCT): New constants.
21115 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
21116 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
21117 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
21118 (aarch64_get_mask_mode): New functions.
21119 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
21120 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
21121 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
21122 predicate modes and predicate registers. Explicitly restrict
21123 GPRs to modes of 16 bytes or smaller. Only allow FP registers
21124 to store a vector mode if it is recognized by
21125 aarch64_classify_vector_mode.
21126 (aarch64_regmode_natural_size): New function.
21127 (aarch64_hard_regno_caller_save_mode): Return the original mode
21129 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
21130 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
21131 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
21132 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
21134 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
21135 does not overlap dest if the function is frame-related. Handle
21137 (aarch64_split_add_offset): New function.
21138 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
21139 them aarch64_add_offset.
21140 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
21141 and update call to aarch64_sub_sp.
21142 (aarch64_add_cfa_expression): New function.
21143 (aarch64_expand_prologue): Pass extra temporary registers to the
21144 functions above. Handle the case in which we need to emit new
21145 DW_CFA_expressions for registers that were originally saved
21146 relative to the stack pointer, but now have to be expressed
21147 relative to the frame pointer.
21148 (aarch64_output_mi_thunk): Pass extra temporary registers to the
21150 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
21151 IP0 and IP1 values for SVE frames.
21152 (aarch64_expand_vec_series): New function.
21153 (aarch64_expand_sve_widened_duplicate): Likewise.
21154 (aarch64_expand_sve_const_vector): Likewise.
21155 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
21156 Handle SVE constants. Use emit_move_insn to move a force_const_mem
21157 into the register, rather than emitting a SET directly.
21158 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
21159 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
21160 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
21161 (offset_9bit_signed_scaled_p): New functions.
21162 (aarch64_replicate_bitmask_imm): New function.
21163 (aarch64_bitmask_imm): Use it.
21164 (aarch64_cannot_force_const_mem): Reject expressions involving
21165 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
21166 (aarch64_classify_index): Handle SVE indices, by requiring
21167 a plain register index with a scale that matches the element size.
21168 (aarch64_classify_address): Handle SVE addresses. Assert that
21169 the mode of the address is VOIDmode or an integer mode.
21170 Update call to aarch64_classify_symbol.
21171 (aarch64_classify_symbolic_expression): Update call to
21172 aarch64_classify_symbol.
21173 (aarch64_const_vec_all_in_range_p): New function.
21174 (aarch64_print_vector_float_operand): Likewise.
21175 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
21176 "vN" for FP registers with SVE modes. Handle (const ...) vectors
21177 and the FP immediates 1.0 and 0.5.
21178 (aarch64_print_address_internal): Handle SVE addresses.
21179 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
21180 (aarch64_regno_regclass): Handle predicate registers.
21181 (aarch64_secondary_reload): Handle big-endian reloads of SVE
21183 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
21184 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
21185 (aarch64_convert_sve_vector_bits): New function.
21186 (aarch64_override_options): Use it to handle -msve-vector-bits=.
21187 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21188 rather than an rtx.
21189 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21190 Handle SVE vector and predicate modes. Accept VL-based constants
21191 that need only one temporary register, and VL offsets that require
21192 no temporary registers.
21193 (aarch64_conditional_register_usage): Mark the predicate registers
21194 as fixed if SVE isn't available.
21195 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21196 Return true for SVE vector and predicate modes.
21197 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21198 rather than an unsigned int. Handle SVE modes.
21199 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21201 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21203 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21204 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21205 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21206 (aarch64_sve_float_mul_immediate_p): New functions.
21207 (aarch64_sve_valid_immediate): New function.
21208 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21209 Explicitly reject structure modes. Check for INDEX constants.
21210 Handle PTRUE and PFALSE constants.
21211 (aarch64_check_zero_based_sve_index_immediate): New function.
21212 (aarch64_simd_imm_zero_p): Delete.
21213 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21214 vector modes. Accept constants in the range of CNT[BHWD].
21215 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21216 ask for an Advanced SIMD mode.
21217 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21218 (aarch64_simd_vector_alignment): Handle SVE predicates.
21219 (aarch64_vectorize_preferred_vector_alignment): New function.
21220 (aarch64_simd_vector_alignment_reachable): Use it instead of
21222 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21223 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21225 (MAX_VECT_LEN): Delete.
21226 (expand_vec_perm_d): Add a vec_flags field.
21227 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21228 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21229 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21231 (aarch64_evpc_rev): Rename to...
21232 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21233 (aarch64_evpc_rev_global): New function.
21234 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21235 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21237 (aarch64_evpc_sve_tbl): New function.
21238 (aarch64_expand_vec_perm_const_1): Update after rename of
21239 aarch64_evpc_rev. Handle SVE permutes too, trying
21240 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21241 than aarch64_evpc_tbl.
21242 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21243 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21244 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21245 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21246 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21247 (aarch64_expand_sve_vcond): New functions.
21248 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21249 of aarch64_vector_mode_p.
21250 (aarch64_dwarf_poly_indeterminate_value): New function.
21251 (aarch64_compute_pressure_classes): Likewise.
21252 (aarch64_can_change_mode_class): Likewise.
21253 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21254 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21255 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21256 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21257 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21258 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21259 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21260 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21262 (Dn, Dl, Dr): Accept const as well as const_vector.
21263 (Dz): Likewise. Compare against CONST0_RTX.
21264 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21265 of "vector" where appropriate.
21266 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21267 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21268 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21269 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21270 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21271 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21272 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21273 (v_int_equiv): Extend to SVE modes.
21274 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21276 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21277 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21278 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21279 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21280 (SVE_COND_FP_CMP): New int iterators.
21281 (perm_hilo): Handle the new unpack unspecs.
21282 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21284 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21285 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21286 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21287 (aarch64_equality_operator, aarch64_constant_vector_operand)
21288 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21289 (aarch64_sve_nonimmediate_operand): Likewise.
21290 (aarch64_sve_general_operand): Likewise.
21291 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21292 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21293 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21294 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21295 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21296 (aarch64_sve_float_arith_immediate): Likewise.
21297 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21298 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21299 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21300 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21301 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21302 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21303 (aarch64_sve_float_arith_operand): Likewise.
21304 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21305 (aarch64_sve_float_mul_operand): Likewise.
21306 (aarch64_sve_vec_perm_operand): Likewise.
21307 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21308 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21309 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21310 as well as const_vector.
21311 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21312 in file. Use CONST0_RTX and CONSTM1_RTX.
21313 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21314 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21315 Use aarch64_simd_imm_zero.
21316 * config/aarch64/aarch64-sve.md: New file.
21317 * config/aarch64/aarch64.md: Include it.
21318 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21319 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21320 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21321 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21322 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21323 (sve): New attribute.
21324 (enabled): Disable instructions with the sve attribute unless
21326 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21327 aarch64_expand_mov_immediate.
21328 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21329 CNT[BHSD] immediates.
21330 (movti): Split CONST_POLY_INT moves into two halves.
21331 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21332 Split additions that need a temporary here if the destination
21333 is the stack pointer.
21334 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21335 (*add<mode>3_poly_1): New instruction.
21336 (set_clobber_cc): New expander.
21338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21340 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21341 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21342 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21343 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21344 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21345 Change innermode from fixed_mode_size to machine_mode.
21346 (simplify_subreg): Update call accordingly. Handle a constant-sized
21347 subreg of a variable-length CONST_VECTOR.
21349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21350 Alan Hayward <alan.hayward@arm.com>
21351 David Sherwood <david.sherwood@arm.com>
21353 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21354 (add_offset_to_base): New function, split out from...
21355 (create_mem_ref): ...here. When handling a scale other than 1,
21356 check first whether the address is valid without the offset.
21357 Add it into the base if so, leaving the index and scale as-is.
21359 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21362 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21363 fold_for_warn before checking if arg2 is INTEGER_CST.
21365 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21367 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21368 (store_multiple_operation): Delete.
21369 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21370 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21371 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21372 guarded by TARGET_STRING.
21373 (rs6000_output_load_multiple): Delete.
21374 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21375 OPTION_MASK_STRING / TARGET_STRING handling.
21376 (print_operand) <'N', 'O'>: Add comment that these are unused now.
21377 (const rs6000_opt_masks) <"string">: Change mask to 0.
21378 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
21379 (MASK_STRING): Delete.
21380 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
21382 (load_multiple): Delete.
21389 (store_multiple): Delete.
21396 (movmemsi_8reg): Delete.
21397 (corresponding unnamed define_insn): Delete.
21398 (movmemsi_6reg): Delete.
21399 (corresponding unnamed define_insn): Delete.
21400 (movmemsi_4reg): Delete.
21401 (corresponding unnamed define_insn): Delete.
21402 (movmemsi_2reg): Delete.
21403 (corresponding unnamed define_insn): Delete.
21404 (movmemsi_1reg): Delete.
21405 (corresponding unnamed define_insn): Delete.
21406 * config/rs6000/rs6000.opt (mno-string): New.
21407 (mstring): Replace by deprecation warning stub.
21408 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
21410 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21412 * regrename.c (regrename_do_replace): If replacing the same
21413 reg multiple times, try to reuse last created gen_raw_REG.
21416 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
21417 main to workaround a bug in GDB.
21419 2018-01-12 Tom de Vries <tom@codesourcery.com>
21422 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
21424 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
21426 PR rtl-optimization/80481
21427 * ira-color.c (get_cap_member): New function.
21428 (allocnos_conflict_by_live_ranges_p): Use it.
21429 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
21430 (setup_slot_coalesced_allocno_live_ranges): Ditto.
21432 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
21435 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
21436 (*saddl_se_1): Ditto.
21437 (*ssubsi_1): Ditto.
21438 (*ssubl_se_1): Ditto.
21440 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21442 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
21443 rather than wi::to_widest for DR_INITs.
21444 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
21445 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
21446 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
21448 (vect_analyze_group_access_1): Note that here.
21450 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21452 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
21453 polynomial type sizes.
21455 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21457 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
21458 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
21459 (gimple_add_tmp_var): Likewise.
21461 2018-01-12 Martin Liska <mliska@suse.cz>
21463 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
21464 (gimple_alloc_sizes): Likewise.
21465 (dump_gimple_statistics): Use PRIu64 in printf format.
21466 * gimple.h: Change uint64_t to int.
21468 2018-01-12 Martin Liska <mliska@suse.cz>
21470 * tree-core.h: Use uint64_t instead of int.
21471 * tree.c (tree_node_counts): Likewise.
21472 (tree_node_sizes): Likewise.
21473 (dump_tree_statistics): Use PRIu64 in printf format.
21475 2018-01-12 Martin Liska <mliska@suse.cz>
21477 * Makefile.in: As qsort_chk is implemented in vec.c, add
21478 vec.o to linkage of gencfn-macros.
21479 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
21480 passing the info to record_node_allocation_statistics.
21481 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
21483 * ggc-common.c (struct ggc_usage): Add operator== and use
21484 it in operator< and compare function.
21485 * mem-stats.h (struct mem_usage): Likewise.
21486 * vec.c (struct vec_usage): Remove operator< and compare
21487 function. Can be simply inherited.
21489 2018-01-12 Martin Jambor <mjambor@suse.cz>
21492 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
21493 * tree-ssa-math-opts.c: Include domwalk.h.
21494 (convert_mult_to_fma_1): New function.
21495 (fma_transformation_info): New type.
21496 (fma_deferring_state): Likewise.
21497 (cancel_fma_deferring): New function.
21498 (result_of_phi): Likewise.
21499 (last_fma_candidate_feeds_initial_phi): Likewise.
21500 (convert_mult_to_fma): Added deferring logic, split actual
21501 transformation to convert_mult_to_fma_1.
21502 (math_opts_dom_walker): New type.
21503 (math_opts_dom_walker::after_dom_children): New method, body moved
21504 here from pass_optimize_widening_mul::execute, added deferring logic
21506 (pass_optimize_widening_mul::execute): Moved most of code to
21507 math_opts_dom_walker::after_dom_children.
21508 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
21509 * config/i386/i386.c (ix86_option_override_internal): Added
21510 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
21512 2018-01-12 Richard Biener <rguenther@suse.de>
21515 * dwarf2out.c (gen_variable_die): Do not reset old_die for
21516 inline instance vars.
21518 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
21521 * config/rx/rx.c (rx_is_restricted_memory_address):
21522 Handle SUBREG case.
21524 2018-01-12 Richard Biener <rguenther@suse.de>
21526 PR tree-optimization/80846
21527 * target.def (split_reduction): New target hook.
21528 * targhooks.c (default_split_reduction): New function.
21529 * targhooks.h (default_split_reduction): Declare.
21530 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
21531 target requests first reduce vectors by combining low and high
21533 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
21534 (get_vectype_for_scalar_type_and_size): Export.
21535 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
21536 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
21537 * doc/tm.texi: Regenerate.
21538 * config/i386/i386.c (ix86_split_reduction): Implement
21539 TARGET_VECTORIZE_SPLIT_REDUCTION.
21541 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21544 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
21545 in PIC mode except for TARGET_VXWORKS_RTP.
21546 * config/sparc/sparc.c: Include cfgrtl.h.
21547 (TARGET_INIT_PIC_REG): Define.
21548 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
21549 (sparc_pic_register_p): New predicate.
21550 (sparc_legitimate_address_p): Use it.
21551 (sparc_legitimize_pic_address): Likewise.
21552 (sparc_delegitimize_address): Likewise.
21553 (sparc_mode_dependent_address_p): Likewise.
21554 (gen_load_pcrel_sym): Remove 4th parameter.
21555 (load_got_register): Adjust call to above. Remove obsolete stuff.
21556 (sparc_expand_prologue): Do not call load_got_register here.
21557 (sparc_flat_expand_prologue): Likewise.
21558 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
21559 (sparc_use_pseudo_pic_reg): New function.
21560 (sparc_init_pic_reg): Likewise.
21561 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21562 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
21564 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
21566 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
21567 Add item for branch_cost.
21569 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21571 PR rtl-optimization/83565
21572 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
21573 not extend the result to a larger mode for rotate operations.
21574 (num_sign_bit_copies1): Likewise.
21576 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21579 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
21581 Use values-Xc.o for -pedantic.
21582 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
21584 2018-01-12 Martin Liska <mliska@suse.cz>
21587 * ipa-devirt.c (final_warning_record::grow_type_warnings):
21589 (possible_polymorphic_call_targets): Use it.
21590 (ipa_devirt): Likewise.
21592 2018-01-12 Martin Liska <mliska@suse.cz>
21594 * profile-count.h (enum profile_quality): Use 0 as invalid
21595 enum value of profile_quality.
21597 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
21599 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
21600 -mext-string options.
21602 2018-01-12 Richard Biener <rguenther@suse.de>
21604 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
21605 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
21606 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21608 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21610 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
21612 * configure.ac (--with-long-double-format): Add support for the
21613 configuration option to change the default long double format on
21615 * config.gcc (powerpc*-linux*-*): Likewise.
21616 * configure: Regenerate.
21617 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
21618 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
21619 used without modification.
21621 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21623 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
21624 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
21625 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
21626 MISC_BUILTIN_SPEC_BARRIER.
21627 (rs6000_init_builtins): Likewise.
21628 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
21630 (speculation_barrier): New define_insn.
21631 * doc/extend.texi: Document __builtin_speculation_barrier.
21633 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21636 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
21637 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
21638 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
21640 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
21641 integral modes instead of "ss" and "sd".
21642 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
21643 vectors with 32-bit and 64-bit elements.
21644 (vecdupssescalarmodesuffix): New mode attribute.
21645 (vec_dup<mode>): Use it.
21647 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
21650 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
21651 frame if argument is passed on stack.
21653 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21656 * ree.c (combine_reaching_defs): Optimize also
21657 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
21658 reg2=any_extend(exp); reg1=reg2;, formatting fix.
21660 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21662 PR middle-end/83189
21663 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
21665 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21667 PR middle-end/83718
21668 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
21669 after they are computed.
21671 2018-01-11 Bin Cheng <bin.cheng@arm.com>
21673 PR tree-optimization/83695
21674 * gimple-loop-linterchange.cc
21675 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
21676 reset cached scev information after interchange.
21677 (pass_linterchange::execute): Remove call to scev_reset_htab.
21679 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21681 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
21682 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
21683 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
21684 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
21685 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
21686 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
21687 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
21688 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
21689 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
21690 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
21691 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
21692 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
21693 (V_lane_reg): Likewise.
21694 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
21696 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
21697 (vfmal_lane_low<mode>_intrinsic,
21698 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
21699 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
21700 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
21701 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
21702 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
21703 vfmsl_lane_high<mode>_intrinsic): New define_insns.
21705 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21707 * config/arm/arm-cpus.in (fp16fml): New feature.
21708 (ALL_SIMD): Add fp16fml.
21709 (armv8.2-a): Add fp16fml as an option.
21710 (armv8.3-a): Likewise.
21711 (armv8.4-a): Add fp16fml as part of fp16.
21712 * config/arm/arm.h (TARGET_FP16FML): Define.
21713 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
21715 * config/arm/arm-modes.def (V2HF): Define.
21716 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
21717 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
21718 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
21719 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
21720 vfmsl_low, vfmsl_high): New set of builtins.
21721 * config/arm/iterators.md (PLUSMINUS): New code iterator.
21722 (vfml_op): New code attribute.
21723 (VFMLHALVES): New int iterator.
21724 (VFML, VFMLSEL): New mode attributes.
21725 (V_reg): Define mapping for V2HF.
21726 (V_hi, V_lo): New mode attributes.
21727 (VF_constraint): Likewise.
21728 (vfml_half, vfml_half_selector): New int attributes.
21729 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
21731 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
21732 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
21734 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
21735 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
21736 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
21737 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
21739 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
21740 Document new effective target and option set.
21742 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21744 * config/arm/arm-cpus.in (armv8_4): New feature.
21745 (ARMv8_4a): New fgroup.
21746 (armv8.4-a): New arch.
21747 * config/arm/arm-tables.opt: Regenerate.
21748 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
21749 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
21750 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
21751 Add matching rules for -march=armv8.4-a and extensions.
21752 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
21754 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
21757 * config/rx/rx.md (BW): New mode attribute.
21758 (sync_lock_test_and_setsi): Add mode suffix to insn output.
21760 2018-01-11 Richard Biener <rguenther@suse.de>
21762 PR tree-optimization/83435
21763 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
21764 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
21765 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
21767 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21768 Alan Hayward <alan.hayward@arm.com>
21769 David Sherwood <david.sherwood@arm.com>
21771 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
21773 (aarch64_classify_address): Initialize it. Track polynomial offsets.
21774 (aarch64_print_address_internal): Use it to check for a zero offset.
21776 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21777 Alan Hayward <alan.hayward@arm.com>
21778 David Sherwood <david.sherwood@arm.com>
21780 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
21781 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
21782 Return a poly_int64 rather than a HOST_WIDE_INT.
21783 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
21784 rather than a HOST_WIDE_INT.
21785 * config/aarch64/aarch64.h (aarch64_frame): Protect with
21786 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
21787 hard_fp_offset, frame_size, initial_adjust, callee_offset and
21788 final_offset from HOST_WIDE_INT to poly_int64.
21789 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
21790 to_constant when getting the number of units in an Advanced SIMD
21792 (aarch64_builtin_vectorized_function): Check for a constant number
21794 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
21796 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
21797 attribute instead of GET_MODE_NUNITS.
21798 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
21799 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
21800 GET_MODE_SIZE for fixed-size registers.
21801 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
21802 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
21803 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
21804 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
21805 (aarch64_print_operand, aarch64_print_address_internal)
21806 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
21807 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
21808 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
21809 Handle polynomial GET_MODE_SIZE.
21810 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
21811 wider than SImode without modification.
21812 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
21813 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
21814 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
21815 passing and returning SVE modes.
21816 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
21817 rather than GEN_INT.
21818 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
21819 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
21820 (aarch64_allocate_and_probe_stack_space): Likewise.
21821 (aarch64_layout_frame): Cope with polynomial offsets.
21822 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
21823 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
21824 polynomial offsets.
21825 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
21826 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
21827 poly_int64 rather than a HOST_WIDE_INT.
21828 (aarch64_get_separate_components, aarch64_process_components)
21829 (aarch64_expand_prologue, aarch64_expand_epilogue)
21830 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
21831 (aarch64_anchor_offset): New function, split out from...
21832 (aarch64_legitimize_address): ...here.
21833 (aarch64_builtin_vectorization_cost): Handle polynomial
21834 TYPE_VECTOR_SUBPARTS.
21835 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
21837 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
21838 number of elements from the PARALLEL rather than the mode.
21839 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
21840 rather than GET_MODE_BITSIZE.
21841 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
21842 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
21843 (aarch64_expand_vec_perm_const_1): Handle polynomial
21844 d->perm.length () and d->perm elements.
21845 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
21846 Apply to_constant to d->perm elements.
21847 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
21848 polynomial CONST_VECTOR_NUNITS.
21849 (aarch64_move_pointer): Take amount as a poly_int64 rather
21851 (aarch64_progress_pointer): Avoid temporary variable.
21852 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
21853 the mode attribute instead of GET_MODE.
21855 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21856 Alan Hayward <alan.hayward@arm.com>
21857 David Sherwood <david.sherwood@arm.com>
21859 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
21860 x exists before using it.
21861 (aarch64_add_constant_internal): Rename to...
21862 (aarch64_add_offset_1): ...this. Replace regnum with separate
21863 src and dest rtxes. Handle the case in which they're different,
21864 including when the offset is zero. Replace scratchreg with an rtx.
21865 Use 2 additions if there is no spare register into which we can
21866 move a 16-bit constant.
21867 (aarch64_add_constant): Delete.
21868 (aarch64_add_offset): Replace reg with separate src and dest
21869 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
21870 Use aarch64_add_offset_1.
21871 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
21872 an rtx rather than an int. Take the delta as a poly_int64
21873 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
21874 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
21875 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
21876 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
21877 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
21878 and aarch64_add_sp.
21879 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
21880 aarch64_add_constant.
21882 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21884 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
21885 Use scalar_float_mode.
21887 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21889 * config/aarch64/aarch64-simd.md
21890 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
21891 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
21892 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
21893 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
21894 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
21895 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
21896 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
21897 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
21898 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
21899 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
21901 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21904 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
21905 targ_options->x_arm_arch_string is non NULL.
21907 2018-01-11 Tamar Christina <tamar.christina@arm.com>
21909 * config/aarch64/aarch64.h
21910 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
21912 2018-01-11 Sudakshina Das <sudi.das@arm.com>
21915 * expmed.c (emit_store_flag_force): Swap if const op0
21916 and change VOIDmode to mode of op0.
21918 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21920 PR rtl-optimization/83761
21921 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
21922 than bytes to mode_for_size.
21924 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21926 PR middle-end/83189
21927 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
21928 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
21931 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21933 PR middle-end/83575
21934 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
21935 when in layout mode.
21936 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
21937 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
21940 2018-01-10 Michael Collison <michael.collison@arm.com>
21942 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
21943 * config/aarch64/aarch64-option-extension.def: Add
21944 AARCH64_OPT_EXTENSION of 'fp16fml'.
21945 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21946 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
21947 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
21948 * config/aarch64/constraints.md (Ui7): New constraint.
21949 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
21950 (VFMLA_SEL_W): Ditto.
21953 (VFMLA16_LOW): New int iterator.
21954 (VFMLA16_HIGH): Ditto.
21955 (UNSPEC_FMLAL): New unspec.
21956 (UNSPEC_FMLSL): Ditto.
21957 (UNSPEC_FMLAL2): Ditto.
21958 (UNSPEC_FMLSL2): Ditto.
21959 (f16mac): New code attribute.
21960 * config/aarch64/aarch64-simd-builtins.def
21961 (aarch64_fmlal_lowv2sf): Ditto.
21962 (aarch64_fmlsl_lowv2sf): Ditto.
21963 (aarch64_fmlalq_lowv4sf): Ditto.
21964 (aarch64_fmlslq_lowv4sf): Ditto.
21965 (aarch64_fmlal_highv2sf): Ditto.
21966 (aarch64_fmlsl_highv2sf): Ditto.
21967 (aarch64_fmlalq_highv4sf): Ditto.
21968 (aarch64_fmlslq_highv4sf): Ditto.
21969 (aarch64_fmlal_lane_lowv2sf): Ditto.
21970 (aarch64_fmlsl_lane_lowv2sf): Ditto.
21971 (aarch64_fmlal_laneq_lowv2sf): Ditto.
21972 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
21973 (aarch64_fmlalq_lane_lowv4sf): Ditto.
21974 (aarch64_fmlsl_lane_lowv4sf): Ditto.
21975 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
21976 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
21977 (aarch64_fmlal_lane_highv2sf): Ditto.
21978 (aarch64_fmlsl_lane_highv2sf): Ditto.
21979 (aarch64_fmlal_laneq_highv2sf): Ditto.
21980 (aarch64_fmlsl_laneq_highv2sf): Ditto.
21981 (aarch64_fmlalq_lane_highv4sf): Ditto.
21982 (aarch64_fmlsl_lane_highv4sf): Ditto.
21983 (aarch64_fmlalq_laneq_highv4sf): Ditto.
21984 (aarch64_fmlsl_laneq_highv4sf): Ditto.
21985 * config/aarch64/aarch64-simd.md:
21986 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
21987 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21988 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
21989 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21990 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
21991 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
21992 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
21993 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
21994 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
21995 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
21996 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
21997 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
21998 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
21999 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
22000 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
22001 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
22002 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
22003 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
22004 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
22005 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
22006 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
22007 (vfmlsl_low_u32): Ditto.
22008 (vfmlalq_low_u32): Ditto.
22009 (vfmlslq_low_u32): Ditto.
22010 (vfmlal_high_u32): Ditto.
22011 (vfmlsl_high_u32): Ditto.
22012 (vfmlalq_high_u32): Ditto.
22013 (vfmlslq_high_u32): Ditto.
22014 (vfmlal_lane_low_u32): Ditto.
22015 (vfmlsl_lane_low_u32): Ditto.
22016 (vfmlal_laneq_low_u32): Ditto.
22017 (vfmlsl_laneq_low_u32): Ditto.
22018 (vfmlalq_lane_low_u32): Ditto.
22019 (vfmlslq_lane_low_u32): Ditto.
22020 (vfmlalq_laneq_low_u32): Ditto.
22021 (vfmlslq_laneq_low_u32): Ditto.
22022 (vfmlal_lane_high_u32): Ditto.
22023 (vfmlsl_lane_high_u32): Ditto.
22024 (vfmlal_laneq_high_u32): Ditto.
22025 (vfmlsl_laneq_high_u32): Ditto.
22026 (vfmlalq_lane_high_u32): Ditto.
22027 (vfmlslq_lane_high_u32): Ditto.
22028 (vfmlalq_laneq_high_u32): Ditto.
22029 (vfmlslq_laneq_high_u32): Ditto.
22030 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
22031 (AARCH64_FL_FOR_ARCH8_4): New.
22032 (AARCH64_ISA_F16FML): New ISA flag.
22033 (TARGET_F16FML): New feature flag for fp16fml.
22034 (doc/invoke.texi): Document new fp16fml option.
22036 2018-01-10 Michael Collison <michael.collison@arm.com>
22038 * config/aarch64/aarch64-builtins.c:
22039 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
22040 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22041 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
22042 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
22043 (AARCH64_ISA_SHA3): New ISA flag.
22044 (TARGET_SHA3): New feature flag for sha3.
22045 * config/aarch64/iterators.md (sha512_op): New int attribute.
22046 (CRYPTO_SHA512): New int iterator.
22047 (UNSPEC_SHA512H): New unspec.
22048 (UNSPEC_SHA512H2): Ditto.
22049 (UNSPEC_SHA512SU0): Ditto.
22050 (UNSPEC_SHA512SU1): Ditto.
22051 * config/aarch64/aarch64-simd-builtins.def
22052 (aarch64_crypto_sha512hqv2di): New builtin.
22053 (aarch64_crypto_sha512h2qv2di): Ditto.
22054 (aarch64_crypto_sha512su0qv2di): Ditto.
22055 (aarch64_crypto_sha512su1qv2di): Ditto.
22056 (aarch64_eor3qv8hi): Ditto.
22057 (aarch64_rax1qv2di): Ditto.
22058 (aarch64_xarqv2di): Ditto.
22059 (aarch64_bcaxqv8hi): Ditto.
22060 * config/aarch64/aarch64-simd.md:
22061 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
22062 (aarch64_crypto_sha512su0qv2di): Ditto.
22063 (aarch64_crypto_sha512su1qv2di): Ditto.
22064 (aarch64_eor3qv8hi): Ditto.
22065 (aarch64_rax1qv2di): Ditto.
22066 (aarch64_xarqv2di): Ditto.
22067 (aarch64_bcaxqv8hi): Ditto.
22068 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
22069 (vsha512h2q_u64): Ditto.
22070 (vsha512su0q_u64): Ditto.
22071 (vsha512su1q_u64): Ditto.
22072 (veor3q_u16): Ditto.
22073 (vrax1q_u64): Ditto.
22074 (vxarq_u64): Ditto.
22075 (vbcaxq_u16): Ditto.
22076 * config/arm/types.md (crypto_sha512): New type attribute.
22077 (crypto_sha3): Ditto.
22078 (doc/invoke.texi): Document new sha3 option.
22080 2018-01-10 Michael Collison <michael.collison@arm.com>
22082 * config/aarch64/aarch64-builtins.c:
22083 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
22084 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22085 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
22086 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
22087 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
22088 (AARCH64_ISA_SM4): New ISA flag.
22089 (TARGET_SM4): New feature flag for sm4.
22090 * config/aarch64/aarch64-simd-builtins.def
22091 (aarch64_sm3ss1qv4si): Ditto.
22092 (aarch64_sm3tt1aq4si): Ditto.
22093 (aarch64_sm3tt1bq4si): Ditto.
22094 (aarch64_sm3tt2aq4si): Ditto.
22095 (aarch64_sm3tt2bq4si): Ditto.
22096 (aarch64_sm3partw1qv4si): Ditto.
22097 (aarch64_sm3partw2qv4si): Ditto.
22098 (aarch64_sm4eqv4si): Ditto.
22099 (aarch64_sm4ekeyqv4si): Ditto.
22100 * config/aarch64/aarch64-simd.md:
22101 (aarch64_sm3ss1qv4si): Ditto.
22102 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
22103 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
22104 (aarch64_sm4eqv4si): Ditto.
22105 (aarch64_sm4ekeyqv4si): Ditto.
22106 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
22107 (sm3part_op): Ditto.
22108 (CRYPTO_SM3TT): Ditto.
22109 (CRYPTO_SM3PART): Ditto.
22110 (UNSPEC_SM3SS1): New unspec.
22111 (UNSPEC_SM3TT1A): Ditto.
22112 (UNSPEC_SM3TT1B): Ditto.
22113 (UNSPEC_SM3TT2A): Ditto.
22114 (UNSPEC_SM3TT2B): Ditto.
22115 (UNSPEC_SM3PARTW1): Ditto.
22116 (UNSPEC_SM3PARTW2): Ditto.
22117 (UNSPEC_SM4E): Ditto.
22118 (UNSPEC_SM4EKEY): Ditto.
22119 * config/aarch64/constraints.md (Ui2): New constraint.
22120 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
22121 * config/arm/types.md (crypto_sm3): New type attribute.
22122 (crypto_sm4): Ditto.
22123 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
22124 (vsm3tt1aq_u32): Ditto.
22125 (vsm3tt1bq_u32): Ditto.
22126 (vsm3tt2aq_u32): Ditto.
22127 (vsm3tt2bq_u32): Ditto.
22128 (vsm3partw1q_u32): Ditto.
22129 (vsm3partw2q_u32): Ditto.
22130 (vsm4eq_u32): Ditto.
22131 (vsm4ekeyq_u32): Ditto.
22132 (doc/invoke.texi): Document new sm4 option.
22134 2018-01-10 Michael Collison <michael.collison@arm.com>
22136 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
22137 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
22138 (AARCH64_FL_FOR_ARCH8_4): New.
22139 (AARCH64_FL_V8_4): New flag.
22140 (doc/invoke.texi): Document new armv8.4-a option.
22142 2018-01-10 Michael Collison <michael.collison@arm.com>
22144 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22145 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
22146 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
22147 * config/aarch64/aarch64-option-extension.def: Add
22148 AARCH64_OPT_EXTENSION of 'sha2'.
22149 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
22150 (crypto): Disable sha2 and aes if crypto disabled.
22151 (crypto): Enable aes and sha2 if enabled.
22152 (simd): Disable sha2 and aes if simd disabled.
22153 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
22155 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
22156 (TARGET_SHA2): New feature flag for sha2.
22157 (TARGET_AES): New feature flag for aes.
22158 * config/aarch64/aarch64-simd.md:
22159 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
22160 conditional on TARGET_AES.
22161 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
22162 (aarch64_crypto_sha1hsi): Make pattern conditional
22164 (aarch64_crypto_sha1hv4si): Ditto.
22165 (aarch64_be_crypto_sha1hv4si): Ditto.
22166 (aarch64_crypto_sha1su1v4si): Ditto.
22167 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
22168 (aarch64_crypto_sha1su0v4si): Ditto.
22169 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
22170 (aarch64_crypto_sha256su0v4si): Ditto.
22171 (aarch64_crypto_sha256su1v4si): Ditto.
22172 (doc/invoke.texi): Document new aes and sha2 options.
22174 2018-01-10 Martin Sebor <msebor@redhat.com>
22176 PR tree-optimization/83781
22177 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
22180 2018-01-11 Martin Sebor <msebor@gmail.com>
22181 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22183 PR tree-optimization/83501
22184 PR tree-optimization/81703
22186 * tree-ssa-strlen.c (get_string_cst): Rename...
22187 (get_string_len): ...to this. Handle global constants.
22188 (handle_char_store): Adjust.
22190 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22191 Jim Wilson <jimw@sifive.com>
22193 * config/riscv/riscv-protos.h (riscv_output_return): New.
22194 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22195 (riscv_attribute_table, riscv_output_return),
22196 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22197 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22198 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22200 (riscv_expand_prologue): Add early return for naked function.
22201 (riscv_expand_epilogue): Likewise.
22202 (riscv_function_ok_for_sibcall): Return false for naked function.
22203 (riscv_set_current_function): New.
22204 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22205 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22206 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22207 * doc/extend.texi (RISC-V Function Attributes): New.
22209 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22211 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22212 check for 128-bit long double before checking TCmode.
22213 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22214 128-bit long doubles before checking TFmode or TCmode.
22215 (FLOAT128_IBM_P): Likewise.
22217 2018-01-10 Martin Sebor <msebor@redhat.com>
22219 PR tree-optimization/83671
22220 * builtins.c (c_strlen): Unconditionally return zero for the empty
22222 Use -Warray-bounds for warnings.
22223 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22224 for non-constant array indices with COMPONENT_REF, arrays of
22225 arrays, and pointers to arrays.
22226 (gimple_fold_builtin_strlen): Determine and set length range for
22227 non-constant character arrays.
22229 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22231 PR middle-end/81897
22232 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22235 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22237 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22239 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22242 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22243 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22244 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22245 indexed_or_indirect_operand predicate.
22246 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22247 (*vsx_le_perm_load_v8hi): Likewise.
22248 (*vsx_le_perm_load_v16qi): Likewise.
22249 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22250 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22251 (*vsx_le_perm_store_v8hi): Likewise.
22252 (*vsx_le_perm_store_v16qi): Likewise.
22253 (eight unnamed splitters): Likewise.
22255 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22257 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22258 * config/rs6000/emmintrin.h: Likewise.
22259 * config/rs6000/mmintrin.h: Likewise.
22260 * config/rs6000/xmmintrin.h: Likewise.
22262 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22265 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22267 * tree.c (tree_nop_conversion): Return true for location wrapper
22269 (maybe_wrap_with_location): New function.
22270 (selftest::check_strip_nops): New function.
22271 (selftest::test_location_wrappers): New function.
22272 (selftest::tree_c_tests): Call it.
22273 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22274 (maybe_wrap_with_location): New decl.
22275 (EXPR_LOCATION_WRAPPER_P): New macro.
22276 (location_wrapper_p): New inline function.
22277 (tree_strip_any_location_wrapper): New inline function.
22279 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22282 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22283 stack_realign_offset for the largest alignment of stack slot
22285 (ix86_find_max_used_stack_alignment): New function.
22286 (ix86_finalize_stack_frame_flags): Use it. Set
22287 max_used_stack_alignment if we don't realign stack.
22288 * config/i386/i386.h (machine_function): Add
22289 max_used_stack_alignment.
22291 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22293 * config/arm/arm.opt (-mbranch-cost): New option.
22294 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22297 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22300 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22301 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22303 2018-01-10 Richard Biener <rguenther@suse.de>
22306 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22307 early out so it also covers the case where we have a non-NULL
22310 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22312 PR tree-optimization/83753
22313 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22314 for non-strided grouped accesses if the number of elements is 1.
22316 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22319 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22320 * i386.h (TARGET_USE_GATHER): Define.
22321 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22323 2018-01-10 Martin Liska <mliska@suse.cz>
22326 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22327 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22329 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22330 CLEANUP_NO_PARTITIONING is not set.
22332 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22334 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22335 for vectors, as a partial revert of r254296.
22336 * rtl.h (const_vec_p): Delete.
22337 (const_vec_duplicate_p): Don't test for vector CONSTs.
22338 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22339 * expmed.c (make_tree): Likewise.
22342 * common.md (E, F): Use CONSTANT_P instead of checking for
22344 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22345 checking for CONST_VECTOR.
22347 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22349 PR middle-end/83575
22350 * predict.c (force_edge_cold): Handle in more sane way edges
22351 with no prediction.
22353 2018-01-09 Carl Love <cel@us.ibm.com>
22355 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22357 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22358 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22359 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22360 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22361 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22362 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22363 * config/rs6000/rs6000-protos.h: Add extern defition for
22364 rs6000_generate_float2_double_code.
22365 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22367 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22368 (float2_v2df): Add define_expand.
22370 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22373 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22374 op_mode in the force_to_mode call.
22376 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22378 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
22379 instead of checking each element individually.
22380 (aarch64_evpc_uzp): Likewise.
22381 (aarch64_evpc_zip): Likewise.
22382 (aarch64_evpc_ext): Likewise.
22383 (aarch64_evpc_rev): Likewise.
22384 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
22385 instead of checking each element individually. Return true without
22387 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
22388 whether all selected elements come from the same input, instead of
22389 checking each element individually. Remove calls to gen_rtx_REG,
22390 start_sequence and end_sequence and instead assert that no rtl is
22393 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22395 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
22396 order of HIGH and CONST checks.
22398 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22400 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
22401 if the destination isn't an SSA_NAME.
22403 2018-01-09 Richard Biener <rguenther@suse.de>
22405 PR tree-optimization/83668
22406 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
22408 (canonicalize_loop_form): ... here, renamed from ...
22409 (canonicalize_loop_closed_ssa_form): ... this and amended to
22410 swap successor edges for loop exit blocks to make us use
22411 the RPO order we need for initial schedule generation.
22413 2018-01-09 Joseph Myers <joseph@codesourcery.com>
22415 PR tree-optimization/64811
22416 * match.pd: When optimizing comparisons with Inf, avoid
22417 introducing or losing exceptions from comparisons with NaN.
22419 2018-01-09 Martin Liska <mliska@suse.cz>
22422 * asan.c (shadow_mem_size): Add gcc_assert.
22424 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
22426 Don't save registers in main().
22429 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
22430 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
22431 * config/avr/avr.c (avr_set_current_function): Don't error if
22432 naked, OS_task or OS_main are specified at the same time.
22433 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
22435 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
22437 * common/config/avr/avr-common.c (avr_option_optimization_table):
22438 Switch on -mmain-is-OS_task for optimizing compilations.
22440 2018-01-09 Richard Biener <rguenther@suse.de>
22442 PR tree-optimization/83572
22443 * graphite.c: Include cfganal.h.
22444 (graphite_transform_loops): Connect infinite loops to exit
22445 and remove fake edges at the end.
22447 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22449 * ipa-inline.c (edge_badness): Revert accidental checkin.
22451 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22454 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
22455 symbols; not inline clones.
22457 2018-01-09 Jakub Jelinek <jakub@redhat.com>
22460 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
22461 hard registers. Formatting fixes.
22463 PR preprocessor/83722
22464 * gcc.c (try_generate_repro): Pass
22465 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
22466 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
22469 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
22470 Kito Cheng <kito.cheng@gmail.com>
22472 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
22473 (riscv_leaf_function_p): Delete.
22474 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
22476 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22478 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
22480 (do_ifelse): New function.
22481 (do_isel): New function.
22482 (do_sub3): New function.
22483 (do_add3): New function.
22484 (do_load_mask_compare): New function.
22485 (do_overlap_load_compare): New function.
22486 (expand_compare_loop): New function.
22487 (expand_block_compare): Call expand_compare_loop() when appropriate.
22488 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
22489 option description.
22490 (-mblock-compare-inline-loop-limit): New option.
22492 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22495 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
22496 Reverse order of second and third operands in first alternative.
22497 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
22498 of first and second elements in UNSPEC_VPERMR vector.
22499 (altivec_expand_vec_perm_le): Likewise.
22501 2018-01-08 Jeff Law <law@redhat.com>
22503 PR rtl-optimizatin/81308
22504 * tree-switch-conversion.c (cfg_altered): New file scoped static.
22505 (process_switch): If group_case_labels makes a change, then set
22507 (pass_convert_switch::execute): If a switch is converted, then
22508 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
22510 PR rtl-optimization/81308
22511 * recog.c (split_all_insns): Conditionally cleanup the CFG after
22514 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
22516 PR target/83663 - Revert r255946
22517 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
22518 generation for cases where splatting a value is not useful.
22519 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
22520 across a vec_duplicate and a paradoxical subreg forming a vector
22521 mode to a vec_concat.
22523 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22525 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
22526 -march=armv8.3-a variants.
22527 * config/arm/t-multilib: Likewise.
22528 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
22530 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22532 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
22534 (cceq_ior_compare_complement): Give it a name so I can use it, and
22535 change boolean_or_operator predicate to boolean_operator so it can
22536 be used to generate a crand.
22537 (eqne): New code iterator.
22538 (bd/bd_neg): New code_attrs.
22539 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
22540 a single define_insn.
22541 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
22542 decrement (bdnzt/bdnzf/bdzt/bdzf).
22543 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
22544 with the new names of the branch decrement patterns, and added the
22545 names of the branch decrement conditional patterns.
22547 2018-01-08 Richard Biener <rguenther@suse.de>
22549 PR tree-optimization/83563
22550 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
22553 2018-01-08 Richard Biener <rguenther@suse.de>
22555 PR middle-end/83713
22556 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
22558 2018-01-08 Richard Biener <rguenther@suse.de>
22560 PR tree-optimization/83685
22561 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
22562 references to abnormals.
22564 2018-01-08 Richard Biener <rguenther@suse.de>
22567 * dwarf2out.c (output_indirect_strings): Handle empty
22568 skeleton_debug_str_hash.
22569 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
22571 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22573 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
22574 (emit_store_direct): Likewise.
22575 (arc_trampoline_adjust_address): Likewise.
22576 (arc_asm_trampoline_template): New function.
22577 (arc_initialize_trampoline): Use asm_trampoline_template.
22578 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
22579 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
22580 * config/arc/arc.md (flush_icache): Delete pattern.
22582 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22584 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
22585 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
22588 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22591 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
22592 by not USED_FOR_TARGET.
22593 (make_pass_resolve_sw_modes): Likewise.
22595 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22597 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
22600 2018-01-08 Richard Biener <rguenther@suse.de>
22602 PR middle-end/83580
22603 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
22605 2018-01-08 Richard Biener <rguenther@suse.de>
22607 PR middle-end/83517
22608 * match.pd ((t * 2) / 2) -> t): Add missing :c.
22610 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
22612 PR middle-end/81897
22613 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
22614 basic blocks with a small number of successors.
22615 (convert_control_dep_chain_into_preds): Improve handling of
22617 (dump_predicates): Split apart into...
22618 (dump_pred_chain): ...here...
22619 (dump_pred_info): ...and here.
22620 (can_one_predicate_be_invalidated_p): Add debugging printfs.
22621 (can_chain_union_be_invalidated_p): Improve check for invalidation
22623 (uninit_uses_cannot_happen): Avoid unnecessary if
22624 convert_control_dep_chain_into_preds yielded nothing.
22626 2018-01-06 Martin Sebor <msebor@redhat.com>
22628 PR tree-optimization/83640
22629 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
22630 subtracting negative offset from size.
22631 (builtin_access::overlap): Adjust offset bounds of the access to fall
22632 within the size of the object if possible.
22634 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
22636 PR rtl-optimization/83699
22637 * expmed.c (extract_bit_field_1): Restrict the vector usage of
22638 extract_bit_field_as_subreg to cases in which the extracted
22639 value is also a vector.
22641 * lra-constraints.c (process_alt_operands): Test for the equivalence
22642 substitutions when detecting a possible reload cycle.
22644 2018-01-06 Jakub Jelinek <jakub@redhat.com>
22647 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
22648 by default if flag_selective_schedling{,2}. Formatting fixes.
22650 PR rtl-optimization/83682
22651 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
22652 if it has non-VECTOR_MODE element mode.
22653 (vec_duplicate_p): Likewise.
22655 PR middle-end/83694
22656 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
22657 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
22659 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22662 * config/i386/i386-builtin.def
22663 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
22664 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
22665 Require also OPTION_MASK_ISA_AVX512F in addition to
22666 OPTION_MASK_ISA_GFNI.
22667 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
22668 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
22669 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
22670 to OPTION_MASK_ISA_GFNI.
22671 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
22672 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
22673 OPTION_MASK_ISA_AVX512BW.
22674 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
22675 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
22676 addition to OPTION_MASK_ISA_GFNI.
22677 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
22678 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
22679 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
22680 to OPTION_MASK_ISA_GFNI.
22681 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
22682 a requirement for all ISAs rather than any of them with a few
22684 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
22686 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
22687 bitmasks to be enabled with 3 exceptions, instead of requiring any
22688 enabled ISA with lots of exceptions.
22689 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
22690 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
22691 Change avx512bw in isa attribute to avx512f.
22692 * config/i386/sgxintrin.h: Add license boilerplate.
22693 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
22694 to __AVX512F__ and __AVX512VL to __AVX512VL__.
22695 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
22696 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
22698 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
22699 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
22700 temporarily sse2 rather than sse if not enabled already.
22703 * config/i386/sse.md (VI248_VLBW): Rename to ...
22704 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
22705 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
22706 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
22707 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
22708 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
22709 mode iterator instead of VI248_VLBW.
22711 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
22713 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
22714 (record_modified): Skip clobbers; add debug output.
22715 (param_change_prob): Use sreal frequencies.
22717 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
22719 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
22720 punt for user-aligned variables.
22722 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
22724 * tree-chrec.c (chrec_contains_symbols): Return true for
22727 2018-01-05 Sudakshina Das <sudi.das@arm.com>
22730 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
22731 of (x|y) == x for BICS pattern.
22733 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22735 PR tree-optimization/83605
22736 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
22737 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
22740 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
22742 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
22743 * config/epiphany/rtems.h: New file.
22745 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22746 Uros Bizjak <ubizjak@gmail.com>
22749 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
22750 QIreg_operand instead of register_operand predicate.
22751 * config/i386/i386.c (ix86_rop_should_change_byte_p,
22752 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
22753 comments instead of -fmitigate[-_]rop.
22755 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22758 * cgraphunit.c (symbol_table::compile): Switch to text_section
22759 before calling assembly_start debug hook.
22760 * run-rtl-passes.c (run_rtl_passes): Likewise.
22763 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22765 * tree-vrp.c (extract_range_from_binary_expr_1): Check
22766 range_int_cst_p rather than !symbolic_range_p before calling
22767 extract_range_from_multiplicative_op_1.
22769 2018-01-04 Jeff Law <law@redhat.com>
22771 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
22772 redundant test in assertion.
22774 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22776 * doc/rtl.texi: Document machine_mode wrapper classes.
22778 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22780 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
22781 using tree_to_uhwi.
22783 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22785 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
22786 the VEC_PERM_EXPR fold to fail.
22788 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22791 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
22792 to switched_sections.
22794 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22797 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
22798 test for d.testing.
22800 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
22803 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
22804 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
22806 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22809 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
22810 is BLKmode and bitpos not zero or mode change is needed.
22812 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22815 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
22818 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
22821 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
22822 instead of MULT rtx. Update all corresponding splitters.
22823 (*saddl_se): Ditto.
22824 (*ssub<modesuffix>): Ditto.
22825 (*ssubl_se): Ditto.
22826 (*cmp_sadd_di): Update split patterns.
22827 (*cmp_sadd_si): Ditto.
22828 (*cmp_sadd_sidi): Ditto.
22829 (*cmp_ssub_di): Ditto.
22830 (*cmp_ssub_si): Ditto.
22831 (*cmp_ssub_sidi): Ditto.
22832 * config/alpha/predicates.md (const23_operand): New predicate.
22833 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
22834 Look for ASHIFT, not MULT inner operand.
22835 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
22837 2018-01-04 Martin Liska <mliska@suse.cz>
22839 PR gcov-profile/83669
22840 * gcov.c (output_intermediate_file): Add version to intermediate
22842 * doc/gcov.texi: Document new field 'version' in intermediate
22843 file format. Fix location of '-k' option of gcov command.
22845 2018-01-04 Martin Liska <mliska@suse.cz>
22848 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
22850 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22852 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
22854 2018-01-03 Martin Sebor <msebor@redhat.com>
22856 PR tree-optimization/83655
22857 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
22858 checking calls with invalid arguments.
22860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22862 * tree-vect-stmts.c (vect_get_store_rhs): New function.
22863 (vectorizable_mask_load_store): Delete.
22864 (vectorizable_call): Return false for masked loads and stores.
22865 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
22866 instead of gimple_assign_rhs1.
22867 (vectorizable_load): Handle IFN_MASK_LOAD.
22868 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
22870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22872 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
22874 (vectorizable_mask_load_store): ...here.
22875 (vectorizable_load): ...and here.
22877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22879 * tree-vect-stmts.c (vect_build_all_ones_mask)
22880 (vect_build_zero_merge_argument): New functions, split out from...
22881 (vectorizable_load): ...here.
22883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22885 * tree-vect-stmts.c (vect_check_store_rhs): New function,
22887 (vectorizable_mask_load_store): ...here.
22888 (vectorizable_store): ...and here.
22890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22892 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
22894 (vectorizable_mask_load_store): ...here.
22896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22898 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
22899 (vect_model_store_cost): Take a vec_load_store_type instead of a
22901 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
22902 (vect_model_store_cost): Take a vec_load_store_type instead of a
22904 (vectorizable_mask_load_store): Update accordingly.
22905 (vectorizable_store): Likewise.
22906 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
22908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22910 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
22911 IFN_MASK_LOAD calls here rather than...
22912 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
22914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22915 Alan Hayward <alan.hayward@arm.com>
22916 David Sherwood <david.sherwood@arm.com>
22918 * expmed.c (extract_bit_field_1): For vector extracts,
22919 fall back to extract_bit_field_as_subreg if vec_extract
22922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22923 Alan Hayward <alan.hayward@arm.com>
22924 David Sherwood <david.sherwood@arm.com>
22926 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
22927 they are variable or constant sized.
22928 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
22929 slots for constant-sized data.
22931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22932 Alan Hayward <alan.hayward@arm.com>
22933 David Sherwood <david.sherwood@arm.com>
22935 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
22936 handling COND_EXPRs with boolean comparisons, try to find a better
22937 basis for the mask type than the boolean itself.
22939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22941 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
22942 is calculated and how it can be overridden.
22943 * genmodes.c (max_bitsize_mode_any_mode): New variable.
22944 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
22946 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
22949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22950 Alan Hayward <alan.hayward@arm.com>
22951 David Sherwood <david.sherwood@arm.com>
22953 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
22954 Remove the mode argument.
22955 (aarch64_simd_valid_immediate): Remove the mode and inverse
22957 * config/aarch64/iterators.md (bitsize): New iterator.
22958 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
22959 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
22960 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
22961 aarch64_simd_valid_immediate.
22962 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
22963 (aarch64_reg_or_bic_imm): Likewise.
22964 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
22965 with an insn_type enum and msl with a modifier_type enum.
22966 Replace element_width with a scalar_mode. Change the shift
22967 to unsigned int. Add constructors for scalar_float_mode and
22968 scalar_int_mode elements.
22969 (aarch64_vect_float_const_representable_p): Delete.
22970 (aarch64_can_const_movi_rtx_p)
22971 (aarch64_simd_scalar_immediate_valid_for_move)
22972 (aarch64_simd_make_constant): Update call to
22973 aarch64_simd_valid_immediate.
22974 (aarch64_advsimd_valid_immediate_hs): New function.
22975 (aarch64_advsimd_valid_immediate): Likewise.
22976 (aarch64_simd_valid_immediate): Remove mode and inverse
22977 arguments. Rewrite to use the above. Use const_vec_duplicate_p
22978 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
22979 and aarch64_float_const_representable_p on the result.
22980 (aarch64_output_simd_mov_immediate): Remove mode argument.
22981 Update call to aarch64_simd_valid_immediate and use of
22982 simd_immediate_info.
22983 (aarch64_output_scalar_simd_mov_immediate): Update call
22986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22987 Alan Hayward <alan.hayward@arm.com>
22988 David Sherwood <david.sherwood@arm.com>
22990 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
22991 (mode_nunits): Likewise CONST_MODE_NUNITS.
22992 * machmode.def (ADJUST_NUNITS): Document.
22993 * genmodes.c (mode_data::need_nunits_adj): New field.
22994 (blank_mode): Update accordingly.
22995 (adj_nunits): New variable.
22996 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
22998 (emit_mode_size_inline): Set need_bytesize_adj for all modes
22999 listed in adj_nunits.
23000 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
23001 listed in adj_nunits. Don't emit case statements for such modes.
23002 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
23003 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
23004 nothing if adj_nunits is nonnull.
23005 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
23006 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
23007 (emit_mode_fbit): Update use of print_maybe_const_decl.
23008 (emit_move_size): Likewise. Treat the array as non-const
23010 (emit_mode_adjustments): Handle adj_nunits.
23012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23014 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
23015 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
23016 (VECTOR_MODES): Use it.
23017 (make_vector_modes): Take the prefix as an argument.
23019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23020 Alan Hayward <alan.hayward@arm.com>
23021 David Sherwood <david.sherwood@arm.com>
23023 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
23024 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
23025 for MODE_VECTOR_BOOL.
23026 * machmode.def (VECTOR_BOOL_MODE): Document.
23027 * genmodes.c (VECTOR_BOOL_MODE): New macro.
23028 (make_vector_bool_mode): New function.
23029 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
23031 * lto-streamer-in.c (lto_input_mode_table): Likewise.
23032 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
23034 * stor-layout.c (int_mode_for_mode): Likewise.
23035 * tree.c (build_vector_type_for_mode): Likewise.
23036 * varasm.c (output_constant_pool_2): Likewise.
23037 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
23038 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
23039 for MODE_VECTOR_BOOL.
23040 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
23041 of mode class checks.
23042 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
23043 instead of a list of mode class checks.
23044 (expand_vector_scalar_condition): Likewise.
23045 (type_for_widest_vector_mode): Handle BImode as an inner mode.
23047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23048 Alan Hayward <alan.hayward@arm.com>
23049 David Sherwood <david.sherwood@arm.com>
23051 * machmode.h (mode_size): Change from unsigned short to
23053 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
23054 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23055 or if measurement_type is not polynomial.
23056 (fixed_size_mode::includes_p): Check for constant-sized modes.
23057 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
23058 return a poly_uint16 rather than an unsigned short.
23059 (emit_mode_size): Change the type of mode_size from unsigned short
23060 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
23061 (emit_mode_adjustments): Cope with polynomial vector sizes.
23062 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23064 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23066 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
23067 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
23068 * caller-save.c (setup_save_areas): Likewise.
23069 (replace_reg_with_saved_mem): Likewise.
23070 * calls.c (emit_library_call_value_1): Likewise.
23071 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
23072 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
23073 (gen_lowpart_for_combine): Likewise.
23074 * convert.c (convert_to_integer_1): Likewise.
23075 * cse.c (equiv_constant, cse_insn): Likewise.
23076 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
23077 (cselib_subst_to_values): Likewise.
23078 * dce.c (word_dce_process_block): Likewise.
23079 * df-problems.c (df_word_lr_mark_ref): Likewise.
23080 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
23081 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
23082 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
23083 (rtl_for_decl_location): Likewise.
23084 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
23085 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
23086 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
23087 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
23088 (expand_expr_real_1): Likewise.
23089 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
23090 (pad_below): Likewise.
23091 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23092 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
23093 * ira.c (get_subreg_tracking_sizes): Likewise.
23094 * ira-build.c (ira_create_allocno_objects): Likewise.
23095 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
23096 (ira_sort_regnos_for_alter_reg): Likewise.
23097 * ira-costs.c (record_operand_costs): Likewise.
23098 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
23099 (resolve_simple_move): Likewise.
23100 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
23101 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
23102 (lra_constraints): Likewise.
23103 (CONST_POOL_OK_P): Reject variable-sized modes.
23104 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
23105 (add_pseudo_to_slot, lra_spill): Likewise.
23106 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23107 * optabs-query.c (get_best_extraction_insn): Likewise.
23108 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23109 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
23110 (expand_mult_highpart, valid_multiword_target_p): Likewise.
23111 * recog.c (offsettable_address_addr_space_p): Likewise.
23112 * regcprop.c (maybe_mode_change): Likewise.
23113 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
23114 * regrename.c (build_def_use): Likewise.
23115 * regstat.c (dump_reg_info): Likewise.
23116 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
23117 (find_reloads, find_reloads_subreg_address): Likewise.
23118 * reload1.c (eliminate_regs_1): Likewise.
23119 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
23120 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
23121 (simplify_binary_operation_1, simplify_subreg): Likewise.
23122 * targhooks.c (default_function_arg_padding): Likewise.
23123 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
23124 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
23125 (verify_gimple_assign_ternary): Likewise.
23126 * tree-inline.c (estimate_move_cost): Likewise.
23127 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23128 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
23129 (get_address_cost_ainc): Likewise.
23130 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
23131 (vect_supportable_dr_alignment): Likewise.
23132 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23133 (vectorizable_reduction): Likewise.
23134 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
23135 (vectorizable_operation, vectorizable_load): Likewise.
23136 * tree.c (build_same_sized_truth_vector_type): Likewise.
23137 * valtrack.c (cleanup_auto_inc_dec): Likewise.
23138 * var-tracking.c (emit_note_insn_var_location): Likewise.
23139 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
23140 (ADDR_VEC_ALIGN): Likewise.
23142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23143 Alan Hayward <alan.hayward@arm.com>
23144 David Sherwood <david.sherwood@arm.com>
23146 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
23148 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23149 or if measurement_type is polynomial.
23150 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
23151 * combine.c (make_extraction): Likewise.
23152 * dse.c (find_shift_sequence): Likewise.
23153 * dwarf2out.c (mem_loc_descriptor): Likewise.
23154 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
23155 (extract_bit_field, extract_low_bits): Likewise.
23156 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
23157 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
23158 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
23159 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
23160 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23161 * reload.c (find_reloads): Likewise.
23162 * reload1.c (alter_reg): Likewise.
23163 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
23164 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
23165 * tree-if-conv.c (predicate_mem_writes): Likewise.
23166 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
23167 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
23168 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23169 * valtrack.c (dead_debug_insert_temp): Likewise.
23170 * varasm.c (mergeable_constant_section): Likewise.
23171 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
23173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23174 Alan Hayward <alan.hayward@arm.com>
23175 David Sherwood <david.sherwood@arm.com>
23177 * expr.c (expand_assignment): Cope with polynomial mode sizes
23178 when assigning to a CONCAT.
23180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23181 Alan Hayward <alan.hayward@arm.com>
23182 David Sherwood <david.sherwood@arm.com>
23184 * machmode.h (mode_precision): Change from unsigned short to
23186 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23188 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23189 or if measurement_type is not polynomial.
23190 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23191 in which the mode is already known to be a scalar_int_mode.
23192 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23193 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23195 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23196 for GET_MODE_PRECISION.
23197 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23198 for GET_MODE_PRECISION.
23199 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23201 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23202 (expand_field_assignment, make_extraction): Likewise.
23203 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23204 (get_last_value): Likewise.
23205 * convert.c (convert_to_integer_1): Likewise.
23206 * cse.c (cse_insn): Likewise.
23207 * expr.c (expand_expr_real_1): Likewise.
23208 * lra-constraints.c (simplify_operand_subreg): Likewise.
23209 * optabs-query.c (can_atomic_load_p): Likewise.
23210 * optabs.c (expand_atomic_load): Likewise.
23211 (expand_atomic_store): Likewise.
23212 * ree.c (combine_reaching_defs): Likewise.
23213 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23214 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23215 * tree.h (type_has_mode_precision_p): Likewise.
23216 * ubsan.c (instrument_si_overflow): Likewise.
23218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23219 Alan Hayward <alan.hayward@arm.com>
23220 David Sherwood <david.sherwood@arm.com>
23222 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23223 polynomial numbers of units.
23224 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23225 (valid_vector_subparts_p): New function.
23226 (build_vector_type): Remove temporary shim and take the number
23227 of units as a poly_uint64 rather than an int.
23228 (build_opaque_vector_type): Take the number of units as a
23229 poly_uint64 rather than an int.
23230 * tree.c (build_vector_from_ctor): Handle polynomial
23231 TYPE_VECTOR_SUBPARTS.
23232 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23233 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23234 (build_vector_from_val): If the number of units is variable,
23235 use build_vec_duplicate_cst for constant operands and
23236 VEC_DUPLICATE_EXPR otherwise.
23237 (make_vector_type): Remove temporary is_constant ().
23238 (build_vector_type, build_opaque_vector_type): Take the number of
23239 units as a poly_uint64 rather than an int.
23240 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23242 * cfgexpand.c (expand_debug_expr): Likewise.
23243 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23244 (store_constructor, expand_expr_real_1): Likewise.
23245 (const_scalar_mask_from_tree): Likewise.
23246 * fold-const-call.c (fold_const_reduction): Likewise.
23247 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23248 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23249 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23250 (fold_relational_const): Likewise.
23251 (native_interpret_vector): Likewise. Change the size from an
23252 int to an unsigned int.
23253 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23254 TYPE_VECTOR_SUBPARTS.
23255 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23256 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23257 duplicating a non-constant operand into a variable-length vector.
23258 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23259 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23260 * ipa-icf.c (sem_variable::equals): Likewise.
23261 * match.pd: Likewise.
23262 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23263 * print-tree.c (print_node): Likewise.
23264 * stor-layout.c (layout_type): Likewise.
23265 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23266 * tree-cfg.c (verify_gimple_comparison): Likewise.
23267 (verify_gimple_assign_binary): Likewise.
23268 (verify_gimple_assign_ternary): Likewise.
23269 (verify_gimple_assign_single): Likewise.
23270 * tree-pretty-print.c (dump_generic_node): Likewise.
23271 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23272 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23273 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23274 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23275 (vect_shift_permute_load_chain): Likewise.
23276 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23277 (expand_vector_condition, optimize_vector_constructor): Likewise.
23278 (lower_vec_perm, get_compute_type): Likewise.
23279 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23280 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23281 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23282 (vect_recog_mask_conversion_pattern): Likewise.
23283 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23284 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23285 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23286 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23287 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23288 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23289 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23290 (supportable_widening_operation): Likewise.
23291 (supportable_narrowing_operation): Likewise.
23292 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23294 * varasm.c (output_constant): Likewise.
23296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23297 Alan Hayward <alan.hayward@arm.com>
23298 David Sherwood <david.sherwood@arm.com>
23300 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23301 so that both the length == 3 and length != 3 cases set up their
23302 own permute vectors. Add comments explaining why we know the
23303 number of elements is constant.
23304 (vect_permute_load_chain): Likewise.
23306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23307 Alan Hayward <alan.hayward@arm.com>
23308 David Sherwood <david.sherwood@arm.com>
23310 * machmode.h (mode_nunits): Change from unsigned char to
23312 (ONLY_FIXED_SIZE_MODES): New macro.
23313 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23314 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23315 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23317 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23318 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23319 or if measurement_type is not polynomial.
23320 * genmodes.c (ZERO_COEFFS): New macro.
23321 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23323 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23324 Use ZERO_COEFFS when emitting initializers.
23325 * data-streamer.h (bp_pack_poly_value): New function.
23326 (bp_unpack_poly_value): Likewise.
23327 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23328 for GET_MODE_NUNITS.
23329 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23330 for GET_MODE_NUNITS.
23331 * tree.c (make_vector_type): Remove temporary shim and make
23332 the real function take the number of units as a poly_uint64
23333 rather than an int.
23334 (build_vector_type_for_mode): Handle polynomial nunits.
23335 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23336 * emit-rtl.c (const_vec_series_p_1): Likewise.
23337 (gen_rtx_CONST_VECTOR): Likewise.
23338 * fold-const.c (test_vec_duplicate_folding): Likewise.
23339 * genrecog.c (validate_pattern): Likewise.
23340 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23341 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23342 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23343 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23344 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23345 * rtlanal.c (subreg_get_info): Likewise.
23346 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23347 (vect_grouped_load_supported): Likewise.
23348 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23349 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23350 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23351 (simplify_const_unary_operation, simplify_binary_operation_1)
23352 (simplify_const_binary_operation, simplify_ternary_operation)
23353 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23354 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23355 instead of CONST_VECTOR_NUNITS.
23356 * varasm.c (output_constant_pool_2): Likewise.
23357 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23358 explicit-encoded elements in the XVEC for variable-length vectors.
23360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23362 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23365 Alan Hayward <alan.hayward@arm.com>
23366 David Sherwood <david.sherwood@arm.com>
23368 * coretypes.h (fixed_size_mode): Declare.
23369 (fixed_size_mode_pod): New typedef.
23370 * builtins.h (target_builtins::x_apply_args_mode)
23371 (target_builtins::x_apply_result_mode): Change type to
23372 fixed_size_mode_pod.
23373 * builtins.c (apply_args_size, apply_result_size, result_vector)
23374 (expand_builtin_apply_args_1, expand_builtin_apply)
23375 (expand_builtin_return): Update accordingly.
23377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23379 * cse.c (hash_rtx_cb): Hash only the encoded elements.
23380 * cselib.c (cselib_hash_rtx): Likewise.
23381 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
23382 CONST_VECTOR encoding.
23384 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23385 Jeff Law <law@redhat.com>
23388 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
23389 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
23390 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
23391 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
23394 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
23395 explicitly probe *sp in a noreturn function if there were any callee
23396 register saves or frame pointer is needed.
23398 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23401 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
23402 BLKmode for ternary, binary or unary expressions.
23405 * var-tracking.c (delete_vta_debug_insn): New inline function.
23406 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
23407 insns from get_insns () to NULL instead of each bb separately.
23408 Use delete_vta_debug_insn. No longer static.
23409 (vt_debug_insns_local, variable_tracking_main_1): Adjust
23410 delete_vta_debug_insns callers.
23411 * rtl.h (delete_vta_debug_insns): Declare.
23412 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
23413 instead of variable_tracking_main.
23415 2018-01-03 Martin Sebor <msebor@redhat.com>
23417 PR tree-optimization/83603
23418 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
23419 arguments past the endof the argument list in functions declared
23420 without a prototype.
23421 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
23422 Avoid checking when arguments are null.
23424 2018-01-03 Martin Sebor <msebor@redhat.com>
23427 * doc/extend.texi (attribute const): Fix a typo.
23428 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
23429 issuing -Wsuggest-attribute for void functions.
23431 2018-01-03 Martin Sebor <msebor@redhat.com>
23433 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
23434 offset_int::from instead of wide_int::to_shwi.
23435 (maybe_diag_overlap): Remove assertion.
23436 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
23437 * gimple-ssa-sprintf.c (format_directive): Same.
23438 (parse_directive): Same.
23439 (sprintf_dom_walker::compute_format_length): Same.
23440 (try_substitute_return_value): Same.
23442 2018-01-03 Jeff Law <law@redhat.com>
23444 PR middle-end/83654
23445 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
23446 non-constant residual for zero at runtime and avoid probing in
23447 that case. Reorganize code for trailing problem to mirror handling
23450 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23452 PR tree-optimization/83501
23453 * tree-ssa-strlen.c (get_string_cst): New.
23454 (handle_char_store): Call get_string_cst.
23456 2018-01-03 Martin Liska <mliska@suse.cz>
23458 PR tree-optimization/83593
23459 * tree-ssa-strlen.c: Include tree-cfg.h.
23460 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
23461 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
23462 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
23464 (strlen_dom_walker::before_dom_children): Call
23465 gimple_purge_dead_eh_edges. Dump tranformation with details
23467 (strlen_dom_walker::before_dom_children): Update call by adding
23468 new argument cleanup_eh.
23469 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
23471 2018-01-03 Martin Liska <mliska@suse.cz>
23474 * cif-code.def (VARIADIC_THUNK): New enum value.
23475 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
23478 2018-01-03 Jan Beulich <jbeulich@suse.com>
23480 * sse.md (mov<mode>_internal): Tighten condition for when to use
23481 vmovdqu<ssescalarsize> for TI and OI modes.
23483 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23485 Update copyright years.
23487 2018-01-03 Martin Liska <mliska@suse.cz>
23490 * ipa-visibility.c (function_and_variable_visibility): Skip
23491 functions with noipa attribure.
23493 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23495 * gcc.c (process_command): Update copyright notice dates.
23496 * gcov-dump.c (print_version): Ditto.
23497 * gcov.c (print_version): Ditto.
23498 * gcov-tool.c (print_version): Ditto.
23499 * gengtype.c (create_file): Ditto.
23500 * doc/cpp.texi: Bump @copying's copyright year.
23501 * doc/cppinternals.texi: Ditto.
23502 * doc/gcc.texi: Ditto.
23503 * doc/gccint.texi: Ditto.
23504 * doc/gcov.texi: Ditto.
23505 * doc/install.texi: Ditto.
23506 * doc/invoke.texi: Ditto.
23508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23510 * vector-builder.h (vector_builder::m_full_nelts): Change from
23511 unsigned int to poly_uint64.
23512 (vector_builder::full_nelts): Update prototype accordingly.
23513 (vector_builder::new_vector): Likewise.
23514 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
23515 (vector_builder::operator ==): Likewise.
23516 (vector_builder::finalize): Likewise.
23517 * int-vector-builder.h (int_vector_builder::int_vector_builder):
23518 Take the number of elements as a poly_uint64 rather than an
23520 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
23521 from unsigned int to poly_uint64.
23522 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
23523 (vec_perm_indices::new_vector): Likewise.
23524 (vec_perm_indices::length): Likewise.
23525 (vec_perm_indices::nelts_per_input): Likewise.
23526 (vec_perm_indices::input_nelts): Likewise.
23527 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
23528 number of elements per input as a poly_uint64 rather than an
23529 unsigned int. Use the original encoding for variable-length
23530 vectors, rather than clamping each individual element.
23531 For the second and subsequent elements in each pattern,
23532 clamp the step and base before clamping their sum.
23533 (vec_perm_indices::series_p): Handle polynomial element counts.
23534 (vec_perm_indices::all_in_range_p): Likewise.
23535 (vec_perm_indices_to_tree): Likewise.
23536 (vec_perm_indices_to_rtx): Likewise.
23537 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
23538 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
23539 (tree_vector_builder::new_binary_operation): Handle polynomial
23540 element counts. Return false if we need to know the number
23541 of elements at compile time.
23542 * fold-const.c (fold_vec_perm): Punt if the number of elements
23543 isn't known at compile time.
23545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23547 * vec-perm-indices.h (vec_perm_builder): Change element type
23548 from HOST_WIDE_INT to poly_int64.
23549 (vec_perm_indices::element_type): Update accordingly.
23550 (vec_perm_indices::clamp): Handle polynomial element_types.
23551 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23552 (vec_perm_indices::all_in_range_p): Likewise.
23553 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
23555 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
23556 polynomial vec_perm_indices element types.
23557 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
23558 * fold-const.c (fold_vec_perm): Likewise.
23559 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
23560 * tree-vect-generic.c (lower_vec_perm): Likewise.
23561 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23562 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
23563 element type to HOST_WIDE_INT.
23565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23566 Alan Hayward <alan.hayward@arm.com>
23567 David Sherwood <david.sherwood@arm.com>
23569 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
23570 rather than an int. Use plus_constant.
23571 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
23572 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
23574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23575 Alan Hayward <alan.hayward@arm.com>
23576 David Sherwood <david.sherwood@arm.com>
23578 * calls.c (emit_call_1, expand_call): Change struct_value_size from
23579 a HOST_WIDE_INT to a poly_int64.
23581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23582 Alan Hayward <alan.hayward@arm.com>
23583 David Sherwood <david.sherwood@arm.com>
23585 * calls.c (load_register_parameters): Cope with polynomial
23586 mode sizes. Require a constant size for BLKmode parameters
23587 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
23588 forces a parameter to be padded at the lsb end in order to
23589 fill a complete number of words, require the parameter size
23590 to be ordered wrt UNITS_PER_WORD.
23592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23593 Alan Hayward <alan.hayward@arm.com>
23594 David Sherwood <david.sherwood@arm.com>
23596 * reload1.c (spill_stack_slot_width): Change element type
23597 from unsigned int to poly_uint64_pod.
23598 (alter_reg): Treat mode sizes as polynomial.
23600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23601 Alan Hayward <alan.hayward@arm.com>
23602 David Sherwood <david.sherwood@arm.com>
23604 * reload.c (complex_word_subreg_p): New function.
23605 (reload_inner_reg_of_subreg, push_reload): Use it.
23607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23608 Alan Hayward <alan.hayward@arm.com>
23609 David Sherwood <david.sherwood@arm.com>
23611 * lra-constraints.c (process_alt_operands): Reject matched
23612 operands whose sizes aren't ordered.
23613 (match_reload): Refer to this check here.
23615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23616 Alan Hayward <alan.hayward@arm.com>
23617 David Sherwood <david.sherwood@arm.com>
23619 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
23620 that the mode size is in the set {1, 2, 4, 8, 16}.
23622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23623 Alan Hayward <alan.hayward@arm.com>
23624 David Sherwood <david.sherwood@arm.com>
23626 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
23627 Use plus_constant instead of gen_rtx_PLUS.
23629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23630 Alan Hayward <alan.hayward@arm.com>
23631 David Sherwood <david.sherwood@arm.com>
23633 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
23634 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
23635 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
23636 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
23637 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
23638 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
23639 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
23640 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
23641 * config/i386/i386.c (ix86_push_rounding): ...this new function.
23642 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
23644 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
23645 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
23646 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
23647 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
23648 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
23649 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
23650 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
23651 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
23652 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
23653 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
23655 * expr.c (emit_move_resolve_push): Treat the input and result
23656 of PUSH_ROUNDING as a poly_int64.
23657 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
23658 (emit_push_insn): Likewise.
23659 * lra-eliminations.c (mark_not_eliminable): Likewise.
23660 * recog.c (push_operand): Likewise.
23661 * reload1.c (elimination_effects): Likewise.
23662 * rtlanal.c (nonzero_bits1): Likewise.
23663 * calls.c (store_one_arg): Likewise. Require the padding to be
23664 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
23671 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
23678 rather than an int.
23680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23681 Alan Hayward <alan.hayward@arm.com>
23682 David Sherwood <david.sherwood@arm.com>
23684 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
23685 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
23686 via stack temporaries. Treat the mode size as polynomial too.
23688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23689 Alan Hayward <alan.hayward@arm.com>
23690 David Sherwood <david.sherwood@arm.com>
23692 * expr.c (expand_expr_real_2): When handling conversions involving
23693 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
23694 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
23695 as a poly_uint64 too.
23697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23698 Alan Hayward <alan.hayward@arm.com>
23699 David Sherwood <david.sherwood@arm.com>
23701 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
23703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23704 Alan Hayward <alan.hayward@arm.com>
23705 David Sherwood <david.sherwood@arm.com>
23707 * combine.c (can_change_dest_mode): Handle polynomial
23708 REGMODE_NATURAL_SIZE.
23709 * expmed.c (store_bit_field_1): Likewise.
23710 * expr.c (store_constructor): Likewise.
23711 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
23712 and polynomial REGMODE_NATURAL_SIZE.
23713 (gen_lowpart_common): Likewise.
23714 * reginfo.c (record_subregs_of_mode): Likewise.
23715 * rtlanal.c (read_modify_subreg_p): Likewise.
23717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23718 Alan Hayward <alan.hayward@arm.com>
23719 David Sherwood <david.sherwood@arm.com>
23721 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
23722 numbers of elements.
23724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23725 Alan Hayward <alan.hayward@arm.com>
23726 David Sherwood <david.sherwood@arm.com>
23728 * match.pd: Cope with polynomial numbers of vector elements.
23730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23731 Alan Hayward <alan.hayward@arm.com>
23732 David Sherwood <david.sherwood@arm.com>
23734 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
23735 in a POINTER_PLUS_EXPR.
23737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23738 Alan Hayward <alan.hayward@arm.com>
23739 David Sherwood <david.sherwood@arm.com>
23741 * omp-simd-clone.c (simd_clone_subparts): New function.
23742 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
23743 (ipa_simd_modify_function_body): Likewise.
23745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23746 Alan Hayward <alan.hayward@arm.com>
23747 David Sherwood <david.sherwood@arm.com>
23749 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
23750 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
23751 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
23752 (expand_vector_condition, vector_element): Likewise.
23753 (subparts_gt): New function.
23754 (get_compute_type): Use subparts_gt.
23755 (count_type_subparts): Delete.
23756 (expand_vector_operations_1): Use subparts_gt instead of
23757 count_type_subparts.
23759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23760 Alan Hayward <alan.hayward@arm.com>
23761 David Sherwood <david.sherwood@arm.com>
23763 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
23764 (vect_compile_time_alias): ...this new function. Do the calculation
23765 on poly_ints rather than trees.
23766 (vect_prune_runtime_alias_test_list): Update call accordingly.
23768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23769 Alan Hayward <alan.hayward@arm.com>
23770 David Sherwood <david.sherwood@arm.com>
23772 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
23774 (vect_schedule_slp_instance): Likewise.
23776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23777 Alan Hayward <alan.hayward@arm.com>
23778 David Sherwood <david.sherwood@arm.com>
23780 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
23781 constant and extern definitions for variable-length vectors.
23782 (vect_get_constant_vectors): Note that the number of units
23783 is known to be constant.
23785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23786 Alan Hayward <alan.hayward@arm.com>
23787 David Sherwood <david.sherwood@arm.com>
23789 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
23790 of units as polynomial. Choose between WIDE and NARROW based
23793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23794 Alan Hayward <alan.hayward@arm.com>
23795 David Sherwood <david.sherwood@arm.com>
23797 * tree-vect-stmts.c (simd_clone_subparts): New function.
23798 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
23800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23801 Alan Hayward <alan.hayward@arm.com>
23802 David Sherwood <david.sherwood@arm.com>
23804 * tree-vect-stmts.c (vectorizable_call): Treat the number of
23805 vectors as polynomial. Use build_index_vector for
23806 IFN_GOMP_SIMD_LANE.
23808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23809 Alan Hayward <alan.hayward@arm.com>
23810 David Sherwood <david.sherwood@arm.com>
23812 * tree-vect-stmts.c (get_load_store_type): Treat the number of
23813 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
23814 for variable-length vectors.
23815 (vectorizable_mask_load_store): Treat the number of units as
23816 polynomial, asserting that it is constant if the condition has
23817 already been enforced.
23818 (vectorizable_store, vectorizable_load): Likewise.
23820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23821 Alan Hayward <alan.hayward@arm.com>
23822 David Sherwood <david.sherwood@arm.com>
23824 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
23825 of units as polynomial. Punt if we can't tell at compile time
23826 which vector contains the final result.
23828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23829 Alan Hayward <alan.hayward@arm.com>
23830 David Sherwood <david.sherwood@arm.com>
23832 * tree-vect-loop.c (vectorizable_induction): Treat the number
23833 of units as polynomial. Punt on SLP inductions. Use an integer
23834 VEC_SERIES_EXPR for variable-length integer reductions. Use a
23835 cast of such a series for variable-length floating-point
23838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23839 Alan Hayward <alan.hayward@arm.com>
23840 David Sherwood <david.sherwood@arm.com>
23842 * tree.h (build_index_vector): Declare.
23843 * tree.c (build_index_vector): New function.
23844 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
23845 of units as polynomial, forcibly converting it to a constant if
23846 vectorizable_reduction has already enforced the condition.
23847 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
23848 to create a {1,2,3,...} vector.
23849 (vectorizable_reduction): Treat the number of units as polynomial.
23850 Choose vectype_in based on the largest scalar element size rather
23851 than the smallest number of units. Enforce the restrictions
23854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23855 Alan Hayward <alan.hayward@arm.com>
23856 David Sherwood <david.sherwood@arm.com>
23858 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
23859 number of units as polynomial.
23861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23862 Alan Hayward <alan.hayward@arm.com>
23863 David Sherwood <david.sherwood@arm.com>
23865 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
23866 * target.def (autovectorize_vector_sizes): Return the vector sizes
23867 by pointer, using vector_sizes rather than a bitmask.
23868 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
23869 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
23870 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
23872 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
23873 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
23874 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
23875 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
23876 * omp-general.c (omp_max_vf): Likewise.
23877 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23878 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
23879 * tree-vect-loop.c (vect_analyze_loop): Likewise.
23880 * tree-vect-slp.c (vect_slp_bb): Likewise.
23881 * doc/tm.texi: Regenerate.
23882 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23884 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
23885 the vector size as a poly_uint64 rather than an unsigned int.
23886 (current_vector_size): Change from an unsigned int to a poly_uint64.
23887 (get_vectype_for_scalar_type): Update accordingly.
23888 * tree.h (build_truth_vector_type): Take the size and number of
23889 units as a poly_uint64 rather than an unsigned int.
23890 (build_vector_type): Add a temporary overload that takes
23891 the number of units as a poly_uint64 rather than an unsigned int.
23892 * tree.c (make_vector_type): Likewise.
23893 (build_truth_vector_type): Take the number of units as a poly_uint64
23894 rather than an unsigned int.
23896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23897 Alan Hayward <alan.hayward@arm.com>
23898 David Sherwood <david.sherwood@arm.com>
23900 * target.def (get_mask_mode): Take the number of units and length
23901 as poly_uint64s rather than unsigned ints.
23902 * targhooks.h (default_get_mask_mode): Update accordingly.
23903 * targhooks.c (default_get_mask_mode): Likewise.
23904 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
23905 * doc/tm.texi: Regenerate.
23907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23908 Alan Hayward <alan.hayward@arm.com>
23909 David Sherwood <david.sherwood@arm.com>
23911 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
23912 * omp-general.c (omp_max_vf): Likewise.
23913 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
23914 (expand_omp_simd): Handle polynomial safelen.
23915 * omp-low.c (omplow_simd_context): Add a default constructor.
23916 (omplow_simd_context::max_vf): Change from int to poly_uint64.
23917 (lower_rec_simd_input_clauses): Update accordingly.
23918 (lower_rec_input_clauses): Likewise.
23920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23921 Alan Hayward <alan.hayward@arm.com>
23922 David Sherwood <david.sherwood@arm.com>
23924 * tree-vectorizer.h (vect_nunits_for_cost): New function.
23925 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
23926 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
23927 (vect_analyze_slp_cost): Likewise.
23928 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
23929 (vect_model_load_cost): Likewise.
23931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23932 Alan Hayward <alan.hayward@arm.com>
23933 David Sherwood <david.sherwood@arm.com>
23935 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
23936 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
23937 from an unsigned int * to a poly_uint64_pod *.
23938 (calculate_unrolling_factor): New function.
23939 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
23941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23942 Alan Hayward <alan.hayward@arm.com>
23943 David Sherwood <david.sherwood@arm.com>
23945 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
23946 from an unsigned int to a poly_uint64.
23947 (_loop_vec_info::slp_unrolling_factor): Likewise.
23948 (_loop_vec_info::vectorization_factor): Change from an int
23950 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
23951 (vect_get_num_vectors): New function.
23952 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
23953 (vect_get_num_copies): Use vect_get_num_vectors.
23954 (vect_analyze_data_ref_dependences): Change max_vf from an int *
23955 to an unsigned int *.
23956 (vect_analyze_data_refs): Change min_vf from an int * to a
23958 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23959 than an unsigned HOST_WIDE_INT.
23960 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
23961 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23962 to an unsigned int *.
23963 (vect_analyze_data_ref_dependences): Likewise.
23964 (vect_compute_data_ref_alignment): Handle polynomial vf.
23965 (vect_enhance_data_refs_alignment): Likewise.
23966 (vect_prune_runtime_alias_test_list): Likewise.
23967 (vect_shift_permute_load_chain): Likewise.
23968 (vect_supportable_dr_alignment): Likewise.
23969 (dependence_distance_ge_vf): Take the vectorization factor as a
23970 poly_uint64 rather than an unsigned HOST_WIDE_INT.
23971 (vect_analyze_data_refs): Change min_vf from an int * to a
23973 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
23974 vfm1 as a poly_uint64 rather than an int. Make the same change
23975 for the returned bound_scalar.
23976 (vect_gen_vector_loop_niters): Handle polynomial vf.
23977 (vect_do_peeling): Likewise. Update call to
23978 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
23979 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
23981 * tree-vect-loop.c (vect_determine_vectorization_factor)
23982 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
23983 (vect_get_known_peeling_cost): Likewise.
23984 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
23985 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
23986 (vect_transform_loop): Likewise. Use the lowest possible VF when
23987 updating the upper bounds of the loop.
23988 (vect_min_worthwhile_factor): Make static. Return an unsigned int
23989 rather than an int.
23990 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
23991 polynomial unroll factors.
23992 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
23993 (vect_make_slp_decision): Likewise.
23994 (vect_supported_load_permutation_p): Likewise, and polynomial
23996 (vect_analyze_slp_cost): Handle polynomial vf.
23997 (vect_slp_analyze_node_operations): Likewise.
23998 (vect_slp_analyze_bb_1): Likewise.
23999 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24000 than an unsigned HOST_WIDE_INT.
24001 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
24002 (vectorizable_load): Handle polynomial vf.
24003 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
24005 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
24007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24008 Alan Hayward <alan.hayward@arm.com>
24009 David Sherwood <david.sherwood@arm.com>
24011 * match.pd: Handle bit operations involving three constants
24012 and try to fold one pair.
24014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24016 * tree-vect-loop-manip.c: Include gimple-fold.h.
24017 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
24018 niters_maybe_zero parameters. Handle other cases besides a step of 1.
24019 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
24020 Add a path that uses a step of VF instead of 1, but disable it
24022 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
24023 and niters_no_overflow parameters. Update calls to
24024 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
24025 Create a new SSA name if the latter choses to use a ste other
24026 than zero, and return it via niters_vector_mult_vf_var.
24027 * tree-vect-loop.c (vect_transform_loop): Update calls to
24028 vect_do_peeling, vect_gen_vector_loop_niters and
24029 slpeel_make_loop_iterate_ntimes.
24030 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
24031 (vect_gen_vector_loop_niters): Update declarations after above changes.
24033 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
24035 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
24036 128-bit round to integer instructions.
24037 (ceil<mode>2): Likewise.
24038 (btrunc<mode>2): Likewise.
24039 (round<mode>2): Likewise.
24041 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24043 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
24044 unaligned VSX load/store on P8/P9.
24045 (expand_block_clear): Allow the use of unaligned VSX
24046 load/store on P8/P9.
24048 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24050 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
24052 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
24053 swap associated with both a load and a store.
24055 2018-01-02 Andrew Waterman <andrew@sifive.com>
24057 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
24058 * config/riscv/riscv.md (clear_cache): Use it.
24060 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
24062 * web.c: Remove out-of-date comment.
24064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24066 * expr.c (fixup_args_size_notes): Check that any existing
24067 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
24068 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
24069 (emit_single_push_insn): ...here.
24071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24073 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
24074 (const_vector_encoded_nelts): New function.
24075 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
24076 (const_vector_int_elt, const_vector_elt): Declare.
24077 * emit-rtl.c (const_vector_int_elt_1): New function.
24078 (const_vector_elt): Likewise.
24079 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
24080 of CONST_VECTOR_ELT.
24082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24084 * expr.c: Include rtx-vector-builder.h.
24085 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
24086 directly on the tree encoding.
24087 (const_vector_from_tree): Likewise.
24088 * optabs.c: Include rtx-vector-builder.h.
24089 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
24090 sequence of "u" values.
24091 * vec-perm-indices.c: Include rtx-vector-builder.h.
24092 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
24093 directly on the vec_perm_indices encoding.
24095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24097 * doc/rtl.texi (const_vector): Describe new encoding scheme.
24098 * Makefile.in (OBJS): Add rtx-vector-builder.o.
24099 * rtx-vector-builder.h: New file.
24100 * rtx-vector-builder.c: Likewise.
24101 * rtl.h (rtx_def::u2): Add a const_vector field.
24102 (CONST_VECTOR_NPATTERNS): New macro.
24103 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
24104 (CONST_VECTOR_DUPLICATE_P): Likewise.
24105 (CONST_VECTOR_STEPPED_P): Likewise.
24106 (CONST_VECTOR_ENCODED_ELT): Likewise.
24107 (const_vec_duplicate_p): Check for a duplicated vector encoding.
24108 (unwrap_const_vec_duplicate): Likewise.
24109 (const_vec_series_p): Check for a non-duplicated vector encoding.
24110 Say that the function only returns true for integer vectors.
24111 * emit-rtl.c: Include rtx-vector-builder.h.
24112 (gen_const_vec_duplicate_1): Delete.
24113 (gen_const_vector): Call gen_const_vec_duplicate instead of
24114 gen_const_vec_duplicate_1.
24115 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
24116 (gen_const_vec_duplicate): Use rtx_vector_builder.
24117 (gen_const_vec_series): Likewise.
24118 (gen_rtx_CONST_VECTOR): Likewise.
24119 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
24120 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24121 Build a new vector rather than modifying a CONST_VECTOR in-place.
24122 (handle_special_swappables): Update call accordingly.
24123 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
24124 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24125 Build a new vector rather than modifying a CONST_VECTOR in-place.
24126 (handle_special_swappables): Update call accordingly.
24128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24130 * simplify-rtx.c (simplify_const_binary_operation): Use
24131 CONST_VECTOR_ELT instead of XVECEXP.
24133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24135 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
24136 the selector elements to be different from the data elements
24137 if the selector is a VECTOR_CST.
24138 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
24139 ssizetype for the selector.
24141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24143 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
24144 before testing each element individually.
24145 * tree-vect-generic.c (lower_vec_perm): Likewise.
24147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24149 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
24150 * selftest-run-tests.c (selftest::run_tests): Call it.
24151 * vector-builder.h (vector_builder::operator ==): New function.
24152 (vector_builder::operator !=): Likewise.
24153 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
24154 (vec_perm_indices::all_from_input_p): New function.
24155 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24156 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
24157 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
24158 instead of reading the VECTOR_CST directly. Detect whether both
24159 vector inputs are the same before constructing the vec_perm_indices,
24160 and update the number of inputs argument accordingly. Use the
24161 utility functions added above. Only construct sel2 if we need to.
24163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24165 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
24166 the broadcast of the low byte.
24167 (expand_mult_highpart): Use an explicit encoding for the permutes.
24168 * optabs-query.c (can_mult_highpart_p): Likewise.
24169 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
24170 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24171 (vectorizable_bswap): Likewise.
24172 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
24173 explicit encoding for the power-of-2 permutes.
24174 (vect_permute_store_chain): Likewise.
24175 (vect_grouped_load_supported): Likewise.
24176 (vect_permute_load_chain): Likewise.
24178 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24180 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
24181 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
24182 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
24183 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24184 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
24185 (vect_gen_perm_mask_any): Likewise.
24187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24189 * int-vector-builder.h: New file.
24190 * vec-perm-indices.h: Include int-vector-builder.h.
24191 (vec_perm_indices): Redefine as an int_vector_builder.
24192 (auto_vec_perm_indices): Delete.
24193 (vec_perm_builder): Redefine as a stand-alone class.
24194 (vec_perm_indices::vec_perm_indices): New function.
24195 (vec_perm_indices::clamp): Likewise.
24196 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24197 (vec_perm_indices::new_vector): New function.
24198 (vec_perm_indices::new_expanded_vector): Update for new
24199 vec_perm_indices class.
24200 (vec_perm_indices::rotate_inputs): New function.
24201 (vec_perm_indices::all_in_range_p): Operate directly on the
24202 encoded form, without computing elided elements.
24203 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24204 encoding. Update for new vec_perm_indices class.
24205 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24206 the given vec_perm_builder.
24207 (expand_vec_perm_var): Update vec_perm_builder constructor.
24208 (expand_mult_highpart): Use vec_perm_builder instead of
24209 auto_vec_perm_indices.
24210 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24211 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24212 or double series encoding as appropriate.
24213 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24214 vec_perm_indices instead of auto_vec_perm_indices.
24215 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24216 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24217 (vect_permute_store_chain): Likewise.
24218 (vect_grouped_load_supported): Likewise.
24219 (vect_permute_load_chain): Likewise.
24220 (vect_shift_permute_load_chain): Likewise.
24221 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24222 (vect_transform_slp_perm_load): Likewise.
24223 (vect_schedule_slp_instance): Likewise.
24224 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24225 (vectorizable_mask_load_store): Likewise.
24226 (vectorizable_bswap): Likewise.
24227 (vectorizable_store): Likewise.
24228 (vectorizable_load): Likewise.
24229 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24230 vec_perm_indices instead of auto_vec_perm_indices. Use
24231 tree_to_vec_perm_builder to read the vector from a tree.
24232 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24233 vec_perm_builder instead of a vec_perm_indices.
24234 (have_whole_vector_shift): Use vec_perm_builder and
24235 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24236 truncation to calc_vec_perm_mask_for_shift.
24237 (vect_create_epilog_for_reduction): Likewise.
24238 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24239 from auto_vec_perm_indices to vec_perm_indices.
24240 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24241 instead of changing individual elements.
24242 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24243 the vector in d.perm.
24244 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24245 from auto_vec_perm_indices to vec_perm_indices.
24246 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24247 instead of changing individual elements.
24248 (arm_vectorize_vec_perm_const): Use new_vector to install
24249 the vector in d.perm.
24250 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24251 Update vec_perm_builder constructor.
24252 (rs6000_expand_interleave): Likewise.
24253 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24254 (rs6000_expand_interleave): Likewise.
24256 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24258 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24259 to qimode could truncate the indices.
24260 * optabs.c (expand_vec_perm_var): Likewise.
24262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24264 * Makefile.in (OBJS): Add vec-perm-indices.o.
24265 * vec-perm-indices.h: New file.
24266 * vec-perm-indices.c: Likewise.
24267 * target.h (vec_perm_indices): Replace with a forward class
24269 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24270 * optabs.h: Include vec-perm-indices.h.
24271 (expand_vec_perm): Delete.
24272 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24273 (expand_vec_perm_const): Declare.
24274 * target.def (vec_perm_const_ok): Replace with...
24275 (vec_perm_const): ...this new hook.
24276 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24277 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24278 * doc/tm.texi: Regenerate.
24279 * optabs.def (vec_perm_const): Delete.
24280 * doc/md.texi (vec_perm_const): Likewise.
24281 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24282 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24283 expand_vec_perm for constant permutation vectors. Assert that
24284 the mode of variable permutation vectors is the integer equivalent
24285 of the mode that is being permuted.
24286 * optabs-query.h (selector_fits_mode_p): Declare.
24287 * optabs-query.c: Include vec-perm-indices.h.
24288 (selector_fits_mode_p): New function.
24289 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24290 is defined, instead of checking whether the vec_perm_const_optab
24291 exists. Use targetm.vectorize.vec_perm_const instead of
24292 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24293 fit in the vector mode before using a variable permute.
24294 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24295 vec_perm_indices instead of an rtx.
24296 (expand_vec_perm): Replace with...
24297 (expand_vec_perm_const): ...this new function. Take the selector
24298 as a vec_perm_indices rather than an rtx. Also take the mode of
24299 the selector. Update call to shift_amt_for_vec_perm_mask.
24300 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24301 Use vec_perm_indices::new_expanded_vector to expand the original
24302 selector into bytes. Check whether the indices fit in the vector
24303 mode before using a variable permute.
24304 (expand_vec_perm_var): Make global.
24305 (expand_mult_highpart): Use expand_vec_perm_const.
24306 * fold-const.c: Includes vec-perm-indices.h.
24307 * tree-ssa-forwprop.c: Likewise.
24308 * tree-vect-data-refs.c: Likewise.
24309 * tree-vect-generic.c: Likewise.
24310 * tree-vect-loop.c: Likewise.
24311 * tree-vect-slp.c: Likewise.
24312 * tree-vect-stmts.c: Likewise.
24313 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24315 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24316 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24317 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24318 (aarch64_vectorize_vec_perm_const): ...this new function.
24319 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24320 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24321 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24322 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24323 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24324 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24325 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24327 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24328 check for NEON modes.
24329 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24330 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24331 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24332 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24334 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24335 the old VEC_PERM_CONST conditions.
24336 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24337 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24338 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24339 (ia64_vectorize_vec_perm_const_ok): Merge into...
24340 (ia64_vectorize_vec_perm_const): ...this new function.
24341 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24342 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24343 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24344 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24345 * config/mips/mips.c (mips_expand_vec_perm_const)
24346 (mips_vectorize_vec_perm_const_ok): Merge into...
24347 (mips_vectorize_vec_perm_const): ...this new function.
24348 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24349 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24350 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24351 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24352 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24353 (rs6000_expand_vec_perm_const): Delete.
24354 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24356 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24357 (altivec_expand_vec_perm_const_le): Take each operand individually.
24358 Operate on constant selectors rather than rtxes.
24359 (altivec_expand_vec_perm_const): Likewise. Update call to
24360 altivec_expand_vec_perm_const_le.
24361 (rs6000_expand_vec_perm_const): Delete.
24362 (rs6000_vectorize_vec_perm_const_ok): Delete.
24363 (rs6000_vectorize_vec_perm_const): New function.
24364 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24365 an element count and rtx array.
24366 (rs6000_expand_extract_even): Update call accordingly.
24367 (rs6000_expand_interleave): Likewise.
24368 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24369 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24370 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24371 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24372 (rs6000_expand_vec_perm_const): Delete.
24373 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24374 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24375 (altivec_expand_vec_perm_const_le): Take each operand individually.
24376 Operate on constant selectors rather than rtxes.
24377 (altivec_expand_vec_perm_const): Likewise. Update call to
24378 altivec_expand_vec_perm_const_le.
24379 (rs6000_expand_vec_perm_const): Delete.
24380 (rs6000_vectorize_vec_perm_const_ok): Delete.
24381 (rs6000_vectorize_vec_perm_const): New function. Remove stray
24382 reference to the SPE evmerge intructions.
24383 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24384 an element count and rtx array.
24385 (rs6000_expand_extract_even): Update call accordingly.
24386 (rs6000_expand_interleave): Likewise.
24387 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
24388 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
24390 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24392 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24394 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
24395 vector mode and that that mode matches the mode of the data
24397 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
24398 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
24399 directly using expand_vec_perm_1 when forcing selectors into
24401 (expand_vec_perm_var): New function, split out from expand_vec_perm.
24403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24405 * optabs-query.h (can_vec_perm_p): Delete.
24406 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
24407 * optabs-query.c (can_vec_perm_p): Split into...
24408 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
24409 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
24410 particular selector is valid.
24411 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24412 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24413 (vect_grouped_load_supported): Likewise.
24414 (vect_shift_permute_load_chain): Likewise.
24415 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24416 (vect_transform_slp_perm_load): Likewise.
24417 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24418 (vectorizable_bswap): Likewise.
24419 (vect_gen_perm_mask_checked): Likewise.
24420 * fold-const.c (fold_ternary_loc): Likewise. Don't take
24421 implementations of variable permutation vectors into account
24422 when deciding which selector to use.
24423 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
24424 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
24425 with a false third argument.
24426 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
24427 to test whether the constant selector is valid and can_vec_perm_var_p
24428 to test whether a variable selector is valid.
24430 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24432 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
24433 * optabs-query.c (can_vec_perm_p): Likewise.
24434 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
24435 instead of vec_perm_indices.
24436 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
24437 (vect_gen_perm_mask_checked): Likewise,
24438 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
24439 (vect_gen_perm_mask_checked): Likewise,
24441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24443 * optabs-query.h (qimode_for_vec_perm): Declare.
24444 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
24445 (qimode_for_vec_perm): ...this new function.
24446 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
24448 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24450 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
24451 does not have a conditional at the top.
24453 2018-01-02 Richard Biener <rguenther@suse.de>
24455 * ipa-inline.c (big_speedup_p): Fix expression.
24457 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24460 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
24463 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24467 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
24468 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
24469 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
24470 cond_taken_branch_cost 3->4.
24472 2018-01-01 Jakub Jelinek <jakub@redhat.com>
24474 PR tree-optimization/83581
24475 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
24476 TODO_cleanup_cfg if any changes have been made.
24478 PR middle-end/83608
24479 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
24480 convert_modes if target mode has the right side, but different mode
24483 PR middle-end/83609
24484 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
24485 last argument when extracting from CONCAT. If either from_real or
24486 from_imag is NULL, use expansion through memory. If result is not
24487 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
24488 the parts directly to inner mode, if even that fails, use expansion
24491 PR middle-end/83623
24492 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
24493 check for bswap in mode rather than HImode and use that in expand_unop
24496 Copyright (C) 2018 Free Software Foundation, Inc.
24498 Copying and distribution of this file, with or without modification,
24499 are permitted in any medium without royalty provided the copyright
24500 notice and this notice are preserved.