1 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
5 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
9 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11 * tree-vect-stmts.c (vectorizable_call): Make sure that we
12 use the stmt_vec_info of the original bb statement for the
13 new zero assignment, even if the call is part of a pattern.
15 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
17 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
18 that the sequence is attached to the original statement rather
19 than the pattern statement.
20 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
21 PATTERN_DEF_SEQ from the original statement rather than
22 the main pattern statement.
23 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
24 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
25 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
27 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
29 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
30 definition statements before the early exit for statements that aren't
32 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
34 (vect_transform_loop): ...here. Process pattern definition
35 statements without first checking whether the main pattern
36 statement is live or relevant.
38 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
40 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
41 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
43 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
45 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
46 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
47 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
48 (expand_block_compare): Change select_block_compare_mode call.
49 (expand_strncmp_align_check): Use new functions, fix comment.
50 (emit_final_str_compare_gpr): New function.
51 (expand_strn_compare): Refactor and clean up code.
52 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
54 2018-06-19 Tony Reix <tony.reix@atos.com>
55 Damien Bergamini <damien.bergamini@atos.com>
56 David Edelsohn <dje.gcc@gmail.com>
58 * collect2.c (static_obj): New variable.
59 (static_libs): New variable.
60 (is_in_list): Uncomment declaration.
61 (main): Track AIX libraries linked statically.
62 (is_in_list): Uncomment definition.
63 (scan_prog_file): Don't add AIX shared libraries initializer
64 to constructor list if linking statically.
66 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
68 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
70 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
72 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
74 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
77 2018-06-19 Martin Liska <mliska@suse.cz>
79 * config/i386/i386.c (ix86_can_inline_p): Do not use
80 ipa_fn_summaries::get_create.
81 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
83 (devirtualization_time_bonus): Likewise.
84 (ipcp_propagate_stage): Likewise.
85 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
86 (edge_set_predicate): Likewise.
87 (evaluate_conditions_for_known_args): Likewise.
88 (evaluate_properties_for_edge): Likewise.
89 (ipa_call_summary::reset): Tranform to ...
90 (ipa_call_summary::~ipa_call_summary): ... this.
91 (ipa_fn_summary::reset): Transform to ...
92 (ipa_fn_summary::~ipa_fn_summary): ... this.
93 (ipa_fn_summary_t::remove): Rename to ...
94 (ipa_fn_summary_t::remove_callees): ... this.
95 (ipa_fn_summary_t::duplicate): Use placement new
96 instead of memory copy.
97 (ipa_call_summary_t::duplicate): Likewise.
98 (ipa_call_summary_t::remove): Remove.
99 (dump_ipa_call_summary): Change get_create to get.
100 (ipa_dump_fn_summary): Dump only when summary exists.
101 (analyze_function_body): Use symbol_summary::get instead
103 (compute_fn_summary): Likewise.
104 (estimate_edge_devirt_benefit): Likewise.
105 (estimate_edge_size_and_time): Likewise.
106 (inline_update_callee_summaries): Likewise.
107 (remap_edge_change_prob): Likewise.
108 (remap_edge_summaries): Likewise.
109 (ipa_merge_fn_summary_after_inlining): Likewise.
110 (write_ipa_call_summary): Likewise.
111 (ipa_fn_summary_write): Likewise.
112 (ipa_free_fn_summary): Likewise.
113 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
114 (struct ipa_call_summary): Likewise.
115 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
117 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
118 (estimate_size_after_inlining): Likewise.
119 (estimate_growth): Likewise.
120 (growth_likely_positive): Likewise.
121 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
122 (inline_call): Likewise.
123 * ipa-inline.c (caller_growth_limits): Likewise.
124 (can_inline_edge_p): Likewise.
125 (can_inline_edge_by_limits_p): Likewise.
126 (compute_uninlined_call_time): Likewise.
127 (compute_inlined_call_time): Likewise.
128 (want_inline_small_function_p): Likewise.
129 (edge_badness): Likewise.
130 (update_caller_keys): Likewise.
131 (update_callee_keys): Likewise.
132 (inline_small_functions): Likewise.
133 (inline_to_all_callers_1): Likewise.
134 (dump_overall_stats): Likewise.
135 (early_inline_small_functions): Likewise.
136 (early_inliner): Likewise.
137 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
138 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
139 * ipa-pure-const.c (malloc_candidate_p): Likewise.
140 * ipa-split.c (execute_split_functions): Likewise.
141 * symbol-summary.h: Likewise.
142 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
144 2018-06-19 Richard Biener <rguenther@suse.de>
146 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
147 (vectorize_loops): ... here. Fix dbgcnt handling.
148 (try_vectorize_loop): Wrap try_vectorize_loop_1.
150 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
153 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
154 ieee128 argument takes up only one (vector) register, not two (floating
157 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
159 * gimplify.c (gimplify_init_constructor): Really never clear for an
160 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
162 2018-06-19 Richard Biener <rguenther@suse.de>
164 PR tree-optimization/86179
165 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
166 after failed recognition.
168 2018-06-18 Martin Sebor <msebor@redhat.com>
171 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
172 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
173 Handle integer subtraction.
174 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
175 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
177 2018-06-18 David Malcolm <dmalcolm@redhat.com>
179 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
180 param from rtx to rtx_insn *.
181 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
183 (frv_ifcvt_modify_insn): Likwise.
184 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
185 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
186 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
187 as_a <rtx_insn *> cast to local "unprotected_region" once
188 it's been established that it's not NULL or pc_rtx.
189 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
190 param "sethi" from rtx to rtx_insn *.
191 (nds32_group_float_insns): Likewise for param "insn".
192 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
194 (vax_output_int_subtract): Likewise.
195 * config/vax/vax.c (vax_output_int_add): Likewise for param
197 (vax_output_int_subtract): Likewise.
198 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
199 (emit_pattern_after): Likewise for param "after".
200 (emit_insn_after): Likewise.
201 (emit_jump_insn_after): Likewise.
202 (emit_call_insn_after): Likewise.
203 (emit_debug_insn_after): Likewise.
204 (emit_pattern_before): Likewise for param "before".
205 (emit_insn_before): Likewise.
206 (emit_jump_insn_before): Likewise.
207 * final.c (get_insn_template): Likewise for param "insn", removing
209 * output.h (get_insn_template): Likewise for 2nd param.
210 * rtl.h (emit_insn_before): Likewise.
211 (emit_jump_insn_before): Likewise.
212 (emit_debug_insn_before_noloc): Likewise.
213 (emit_insn_after): Likewise.
214 (emit_jump_insn_after): Likewise.
215 (emit_call_insn_after): Likewise.
216 (emit_debug_insn_after): Likewise.
217 (set_insn_deleted): Likewise for param.
219 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
222 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
223 floating point modes, so that IFmode is numerically greater than
224 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
225 to declare the ordering. This prevents IFmode from being
226 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
227 machine. Include rs6000-modes.h to share the fractional values
228 between genmodes* and the rest of the compiler.
232 * config/rs6000/rs6000-modes.h: New file.
233 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
234 meaning of rs6000_long_double_size so that 126..128 selects an
235 appropriate 128-bit floating point type.
236 (rs6000_option_override_internal): Likewise.
237 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
238 (TARGET_LONG_DOUBLE_128): Change the meaning of
239 rs6000_long_double_size so that 126..128 selects an appropriate
240 128-bit floating point type.
241 (LONG_DOUBLE_TYPE_SIZE): Update comment.
242 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
243 source and destination to match the standard usage.
244 (truncifkf2): Likewise.
245 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
246 ISA 2.07 to use an explicit clobber, instead of passing in a
248 (copysign<mode>3_soft): Likewise.
250 2018-06-18 David Malcolm <dmalcolm@redhat.com>
252 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
253 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
254 (vect_slp_analyze_instance_dependence): Likewise.
255 (vect_enhance_data_refs_alignment): Likewise.
256 (vect_analyze_data_refs_alignment): Likewise.
257 (vect_slp_analyze_and_verify_instance_alignment
258 (vect_analyze_data_ref_accesses): Likewise.
259 (vect_prune_runtime_alias_test_list): Likewise.
260 (vect_analyze_data_refs): Likewise.
261 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
262 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
263 (vect_analyze_scalar_cycles_1): Likewise.
264 (vect_get_loop_niters): Likewise.
265 (vect_analyze_loop_form_1): Likewise.
266 (vect_update_vf_for_slp): Likewise.
267 (vect_analyze_loop_operations): Likewise.
268 (vect_analyze_loop): Likewise.
269 (vectorizable_induction): Likewise.
270 (vect_transform_loop): Likewise.
271 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
272 * tree-vect-slp.c (vect_analyze_slp): Likewise.
273 (vect_make_slp_decision): Likewise.
274 (vect_detect_hybrid_slp): Likewise.
275 (vect_slp_analyze_operations): Likewise.
276 (vect_slp_bb): Likewise.
277 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
278 (vectorizable_bswap): Likewise.
279 (vectorizable_call): Likewise.
280 (vectorizable_simd_clone_call): Likewise.
281 (vectorizable_conversion): Likewise.
282 (vectorizable_assignment): Likewise.
283 (vectorizable_shift): Likewise.
284 (vectorizable_operation): Likewise.
285 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
287 2018-06-18 Martin Sebor <msebor@redhat.com>
289 PR tree-optimization/81384
290 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
291 * builtins.c (expand_builtin_strnlen): New function.
292 (expand_builtin): Call it.
293 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
294 * builtins.def (BUILT_IN_STRNLEN): New.
295 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
296 Warn for bounds in excess of maximum object size.
297 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
298 single-value ranges. Handle strnlen.
299 (handle_builtin_strlen): Handle strnlen.
300 (strlen_check_and_optimize_stmt): Same.
301 * doc/extend.texi (Other Builtins): Document strnlen.
303 2018-06-18 Maya Rashish <coypu@sdf.org>
305 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
306 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
307 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
309 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
311 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
313 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
315 * tree.c (escaped_string::escape): Replace cast to char * by
316 const_cast<char *> (unescaped).
318 2018-06-18 Nick Clifton <nickc@redhat.com>
321 * tree.c (escaped_string): New class. Converts an unescaped
322 string into its escaped equivalent.
323 (warn_deprecated_use): Use the new class to convert the
324 deprecation message, if present.
325 (test_escaped_strings): New self test.
326 (test_c_tests): Add test_escaped_strings.
327 * doc/extend.texi (deprecated): Add a note that the
328 deprecation message is affected by the -fmessage-length
329 option, and that control characters will be escaped.
330 (#pragma GCC error): Document this pragma.
331 (#pragma GCC warning): Likewise.
332 * doc/invoke.texi (-fmessage-length): Document this option's
333 effect on the #warning and #error preprocessor directives and
334 the deprecated attribute.
336 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
338 * tree.c (decl_value_expr_lookup): Revert latest change.
339 (decl_value_expr_insert): Likewise.
341 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
343 * gimplify.c (nonlocal_vlas): Delete.
344 (nonlocal_vla_vars): Likewise.
345 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
347 (gimplify_body): Do not create and destroy nonlocal_vlas.
348 * tree-nested.c: Include diagnostic.h.
349 (use_pointer_in_frame): Tweak.
350 (lookup_field_for_decl): Add assertion and declare the transformation.
351 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
352 internal error when the reference is in a wrong context. Do not
353 create a debug decl by default.
354 (note_nonlocal_block_vlas): Delete.
355 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
356 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
357 create a debug decl by default.
358 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
359 call is in a wrong context.
360 (fixup_vla_decls): New function.
361 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
362 debug variables were created.
363 * tree.c (decl_value_expr_lookup): Add checking assertion.
364 (decl_value_expr_insert): Likewise.
366 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
369 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
370 * tree-scalar-evolution.c (interpret_expr): Likewise.
371 (expression_expensive_p): Likewise.
372 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
373 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
374 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
375 (ssa_defined_by_minus_one_stmt_p): New.
377 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
380 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
381 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
382 * dojump.c (do_jump): Likewise.
383 * expr.c (expand_expr_real_2): Check operand type's sign.
384 * fold-const.c (const_unop): Handle ABSU_EXPR.
385 (fold_abs_const): Likewise.
386 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
387 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
388 (strip_sign_op_1): Likesise.
389 * match.pd: Add new pattern to generate ABSU_EXPR.
390 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
391 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
392 * tree-eh.c (operation_could_trap_helper_p): Likewise.
393 * tree-inline.c (estimate_operator_cost): Likewise.
394 * tree-pretty-print.c (dump_generic_node): Likewise.
395 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
396 * tree.def (ABSU_EXPR): New.
398 2018-06-16 Jakub Jelinek <jakub@redhat.com>
401 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
402 documented as preserved for backward compatibility only.
403 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
405 PR rtl-optimization/86108
406 * bb-reorder.c (create_forwarder_block): Renamed to ...
407 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
408 jump from new landing pad to the second part.
409 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
412 2018-06-15 Jakub Jelinek <jakub@redhat.com>
415 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
416 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
417 Only call store_expr for halves if the mode is the same.
420 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
421 Fix up comment formatting.
423 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
425 * typed-splay-tree.h (typed_splay_tree::remove): New function.
426 (typed_splay_tree::closure,
427 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
428 (typed_splay_tree::typed_splay_tree,
429 typed_splay_tree::operator =): Declared private.
430 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
431 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
432 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
433 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
434 typed_splay_tree::splay_tree_splay,
435 typed_splay_tree::splay_tree_foreach_helper,
436 typed_splay_tree::splay_tree_insert,
437 typed_splay_tree::splay_tree_remove,
438 typed_splay_tree::splay_tree_lookup,
439 typed_splay_tree::splay_tree_predecessor,
440 typed_splay_tree::splay_tree_successor,
441 typed_splay_tree::splay_tree_min,
442 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
443 (typed_splay_tree::root, typed_splay_tree::comp,
444 typed_splay_tree::delete_key,
445 typed_splay_tree::delete_value): New data members.
446 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
447 typed_splay_tree::remove.
449 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
451 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
452 -mginv and -mno-ginv to the assembler.
453 * config/mips/mips.opt (-mcrc): New option.
455 * doc/invoke.text (-mcrc): Document.
458 2018-06-15 Nick Clifton <nickc@redhat.com>
461 * tree.c (escaped_string): New class. Converts an unescaped
462 string into its escaped equivalent.
463 (warn_deprecated_use): Use the new class to convert the
464 deprecation message, if present.
465 (test_escaped_strings): New self test.
466 (test_c_tests): Add test_escaped_strings.
467 * doc/extend.texi (deprecated): Add a note that the
468 deprecation message is affected by the -fmessage-length
469 option, and that control characters will be escaped.
470 (#pragma GCC error): Document this pragma.
471 (#pragma GCC warning): Likewise.
472 * doc/invoke.texi (-fmessage-length): Document this option's
473 effect on the #warning and #error preprocessor directives and
474 the deprecated attribute.
476 2018-06-15 Richard Biener <rguenther@suse.de>
478 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
479 here, also noting vector size used.
480 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
481 size used in MSG_OPTIMIZED_LOCATIONS dump.
482 (pass_slp_vectorize::execute): Adjust.
484 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
487 * config/arc/arc.c (arc_return_address_register): Fix
490 2018-06-15 Richard Biener <rguenther@suse.de>
493 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
494 leave useless conversion stripping to force_gimple_operand_gsi.
495 (gimplify_build2): Likewise.
496 (gimplify_build1): Likewise.
498 2018-06-15 Richard Biener <rguenther@suse.de>
501 * tree-cfg.c (move_stmt_op): unshare invariant addresses
502 before adjusting their block.
504 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
506 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
507 multilibs for *-*-rtems*.
508 * config/riscv/t-rtems: New file.
510 2018-06-14 Jakub Jelinek <jakub@redhat.com>
513 * match.pd ((A +- CST1) +- CST2): Punt if last resort
514 unsigned_type_for returns NULL.
517 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
518 subregs of multi-word pseudos unless the float mode has word size.
520 2018-06-14 Richard Biener <rguenther@suse.de>
523 * tree-vect-generic.c (build_word_mode_vector_type): Remove
524 duplicate and harmful type_hash_canon.
525 * tree.c (type_hash_canon): Assert we didn't find ourselves.
527 2018-06-14 Richard Biener <rguenther@suse.de>
530 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
533 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
535 * config/rtems.h (STDINT_LONG32): Define.
537 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
538 Prachi Godbole <prachi.godbole@imgtec.com>
540 * config/mips/mips-cpus.def: Define P6600.
541 * config/mips/mips-tables.opt: Regenerate.
542 * config/mips/mips.c (mips_ucbranch_type): New enum.
543 (mips_rtx_cost_data): Add support for P6600.
544 (mips_issue_rate): Likewise.
545 (mips_multipass_dfa_lookahead): Likewise.
546 (mips_avoid_hazard): Likewise.
547 (mips_reorg_process_insns): Likewise.
548 (mips_classify_branch_p6600): New function.
549 * config/mips/mips.h (TUNE_P6600): New define.
550 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
551 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
552 * config/mips/mips.md: Include p6600.md.
553 (processor): Add p6600.
554 * config/mips/p6600.md: New file.
555 * doc/invoke.texi: Add p6600 to supported architectures.
557 2018-06-13 Martin Sebor <msebor@redhat.com>
559 PR tree-optimization/86114
560 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
562 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
564 2018-06-13 Richard Biener <rguenther@suse.de>
566 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
567 Properly set vector type of the intermediate stmt.
568 * tree-vect-stmts.c (vectorizable_operation): The destination
569 var always has vectype_out type.
571 2018-06-13 Jeff Law <law@redhat.com>
573 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
574 integer 0 for argument to print_rtl_with_bb.
575 (rl78_reorg): Likewise.
577 2018-06-13 David Malcolm <dmalcolm@redhat.com>
579 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
580 from rtx to rtx_insn *.
581 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
583 (add_sched_insns_for_speculation): Likewise for local "target",
584 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
585 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
586 from rtx_insn ** to rtx_code_label **.
587 (reorg_emit_nops): Likewise.
588 (c6x_reorg): Likewise for local "call_labels".
589 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
591 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
592 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
593 the loops over LABEL_REFS.
594 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
596 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
597 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
598 (split_branches): Strengthen local "olabel" from rtx to
599 rtx_insn *, adding a safe_as_a cast.
600 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
602 (add_insn_after): Likewise for first two params.
603 (add_insn_before): Likewise.
604 (remove_insn): Likewise for param.
605 (emit_pattern_before_noloc): Likewise for second and third params.
606 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
607 (emit_call_insn_before_noloc): Likewise.
608 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
610 (emit_barrier_before): Likewise.
611 (emit_label_before): Strengthen "label" param from "rtx" to
612 "rtx_code_label *". Strengthen "before" param from "rtx" to
614 (emit_insn_after_1): Strengthen "after" param from "rtx" to
616 (emit_pattern_after_noloc): Likewise.
617 (emit_insn_after_noloc): Likewise.
618 (emit_jump_insn_after_noloc): Likewise.
619 (emit_call_insn_after_noloc): Likewise.
620 (emit_debug_insn_after_noloc): Likewise.
621 (emit_barrier_after): Likewise.
622 (emit_label_after): Likewise for both params.
623 (emit_pattern_after_setloc): Likewise for "after" param. Convert
624 "loc" param from "int" to "location_t".
625 (emit_insn_after_setloc): Likewise.
626 (emit_jump_insn_after_setloc): Likewise.
627 (emit_call_insn_after_setloc): Likewise.
628 (emit_debug_insn_after_setloc): Likewise.
629 (emit_pattern_before_setloc): Likewise for "before" param. Convert
630 "loc" param from "int" to "location_t".
631 (emit_pattern_before): Convert NULL_RTX to NULL.
632 (emit_insn_before_setloc): Convert "loc" param from "int" to
634 (emit_jump_insn_before_setloc): Likewise.
635 (emit_call_insn_before_setloc): Likewise.
636 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
637 rtx_insn *. Convert "loc" param from "int" to "location_t".
638 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
639 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
640 Convert 3rd param from "int" to "location_t".
641 (emit_barrier_before, emit_barrier_after, next_real_insn):
642 Strengthen param from rtx to rtx_insn *.
643 (emit_label_before): Strengthen 1st param from "rtx" to
644 "rtx_code_label *". Strengthen 2nd param from "rtx" to
646 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
647 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
648 Strengthen 2nd param from "rtx" to "rtx_insn *".
649 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
650 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
651 Likewise. Convert 3rd param from "int" to "location_t".
652 (emit_label_after): Strengthen 1st param from "rtx" to
654 (next_real_insn, remove_insn): Strengthen param from "rtx" to
656 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
657 from "rtx" to "rtx_insn *".
659 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
661 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
662 bodies streamed in with -Q.
663 * dumpfile.c (dump_files): Add lto-stream-out dump file.
664 * dumpfile.h (tree_dump_index): Add lto_stream_out.
665 * gimple-streamer-out.c: Include gimple-pretty-print.h
666 (output_bb): Dump stmts streamed.
667 * lto-section-out.c: Include print-tree.h
668 (lto_begin_section): Dump sections created.
669 (lto_output_decl_index): Dump decl encoded.
670 * lto-streamer-out.c: Include print-tree.h
671 (create_output_block): Dump output block created.
672 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
673 (output_function): Dump function output.
674 (output_constructor): Dump constructor streamed.
675 (write_global_stream): Output indexes encoded.
676 (produce_asm_for_decls): Dump streams encoded.
677 * lto-streamer.c (streamer_dump_file): New global var.
678 * lto-streamer.h (streamer_dump_file): Declare.
679 * passes.c (ipa_write_summaries): Initialize streamer dump.
680 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
683 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
686 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
687 offsets for register save directives. Emit a second batch of save
688 directives, if need be, when the function accesses prior frames.
690 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
692 * config/arc/fpu.md (fmasf4): Force operand to register.
695 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
697 * config/arc/arc-protos.h (arc_pad_return): Remove.
698 * config/arc/arc.c (machine_function): Remove force_short_suffix
700 (arc_print_operand): Adjust printing of '&'.
701 (arc_verify_short): Remove conditional printing of short suffix.
702 (arc_final_prescan_insn): Remove reference to size_reason.
703 (pad_return): New function.
704 (arc_reorg): Call pad_return.
705 (arc_pad_return): Remove.
706 (arc_init_machine_status): Remove reference to force_short_suffix.
707 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
708 (attr length): When attribute iscompact is true force to 2
709 regardless; in the case of maybe check if we want to force the
710 instruction to have 4 bytes length.
711 (nopv): Change it to generate 4 byte long nop as well.
712 (blockage): New pattern.
713 (simple_return): Remove call to arc_pad_return.
714 (p_return_i): Likewise.
716 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
718 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
720 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
722 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
725 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
727 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
728 for ARC700 and ARCv2.
730 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
733 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
734 operands[2] instead of operands[1].
737 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
739 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
740 case, check whether the outer register overlaps an unallocatable
741 register, not just whether it fits the required class.
743 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
745 * poly-int.h (can_div_trunc_p): Add new overload in which all values
747 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
748 (memrefs_conflict_p): Likewise.
749 (init_alias_analysis): Likewise.
750 * cfgexpand.c (expand_debug_expr): Likewise.
751 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
752 * cse.c (fold_rtx): Likewise.
753 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
754 * expr.c (emit_block_move_hints): Likewise.
755 (clear_storage_hints, push_block, emit_push_insn): Likewise.
756 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
757 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
758 (emit_group_store): Likewise.
759 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
760 to read the PRE/POST_MODIFY increment.
761 * calls.c (store_one_arg): Use strip_offset.
762 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
764 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
766 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
767 (simplify_binary_operation_1): Extend CONST_INT handling to
769 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
770 than a HOST_WIDE_INT.
771 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
773 (adjust_mems, add_stores): Update accodingly.
774 (vt_canonicalize_addr): Track polynomial offsets.
775 (emit_note_insn_var_location): Likewise.
776 (vt_add_function_parameter): Likewise.
777 (vt_initialize): Likewise.
779 2018-06-12 Jeff Law <law@redhat.com>
781 * config.gcc (alpha*-*-freebsd*): Remove.
782 * config/alpha/freebsd.h: Remove.
784 2018-06-12 David Malcolm <dmalcolm@redhat.com>
787 * spellcheck-tree.c (levenshtein_distance): Rename to...
788 (get_edit_distance): ...this, and update for underlying renaming.
789 * spellcheck-tree.h (levenshtein_distance): Rename to...
790 (get_edit_distance): ...this.
791 * spellcheck.c (levenshtein_distance): Rename to...
792 (get_edit_distance): ...this. Convert from Levenshtein distance
793 to Damerau-Levenshtein distance by supporting transpositions of
794 adjacent characters. Rename "v1" to "v_next" and "v0" to
796 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
797 (selftest::test_edit_distance_unit_test_oneway): ...this, and
798 update for underlying renaming.
799 (selftest::levenshtein_distance_unit_test): Rename to...
800 (selftest::test_get_edit_distance_unit): ...this, and update for
802 (selftest::test_find_closest_string): Add example from PR 69968
803 where transposition helps
804 (selftest::test_metric_conditions): Update for renaming.
805 (selftest::test_metric_conditions): Likewise.
806 (selftest::spellcheck_c_tests): Likewise.
807 * spellcheck.h (levenshtein_distance): Rename both overloads to...
808 (get_edit_distance): ...this.
809 (best_match::consider): Update for renaming.
811 2018-06-12 Martin Sebor <msebor@redhat.com>
813 PR tree-optimization/85259
814 * builtins.c (compute_objsize): Handle constant offsets.
815 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
816 true iff a warning has been issued.
817 * gimple.h (gimple_nonartificial_location): New function.
818 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
819 gimple_nonartificial_location and handle -Wno-system-headers.
820 (handle_builtin_stxncpy): Same.
822 2018-06-12 Martin Sebor <msebor@redhat.com>
825 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
827 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
829 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
830 BUILTIN_VEC_XST entries for pointer to double and long long.
832 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
835 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
837 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
840 2018-06-12 Martin Liska <mliska@suse.cz>
842 * doc/options.texi: Document IntegerRange.
844 2018-06-12 Martin Liska <mliska@suse.cz>
846 * config/i386/i386.opt: Make MPX-related options as Deprecated.
847 * opt-functions.awk: Handle Deprecated flag.
848 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
850 (read_cmdline_option): Report warning for a deprecated option.
851 * opts.h (struct cl_option): Add new field cl_deprecated.
852 (CL_ERR_DEPRECATED): New.
854 2018-06-12 Martin Liska <mliska@suse.cz>
856 * doc/options.texi: Document Deprecated option flag.
858 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
860 * config/arc/arc-arch.h (arc_extras): New enum.
861 (arc_cpu_t):Add field extra.
862 (arc_cpu_types): Consider the extras.
863 * config/arc/arc-cpus.def: Add extras info.
864 * config/arc/arc-opts.h (processor_type): Consider extra field.
865 * config/arc/arc.c (arc_override_options): Handle extra field.
867 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
869 * config/arc/arc-arch.h: Update ARC_OPTX macro.
870 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
872 * config/arc/arc.c (arc_init): Update pic warning.
873 (irq_range): Update irq range parsing warnings.
874 (arc_override_options): Update various warning messages.
875 (arc_handle_aux_attribute): Likewise.
877 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
879 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
881 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
883 * doc/sourcebuild.texi: Document usage of line number 0 in verify
884 compiler messages directives.
886 2018-06-12 Matthew Fortune <mfortune@gmail.com>
888 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
889 * config/mips/mips-tables.opt: Regenerate.
890 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
892 * doc/invoke.texi: Document -march=i6500.
894 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
896 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
897 (i6400_gpmul): Add cpu_unit.
898 (i6400_gpdiv): Likewise.
899 (i6400_msa_add_d): Update reservations.
900 (i6400_msa_int_add) Likewise.
901 (i6400_msa_short_logic3) Likewise.
902 (i6400_msa_short_logic2) Likewise.
903 (i6400_msa_short_logic) Likewise.
904 (i6400_msa_move) Likewise.
905 (i6400_msa_cmp) Likewise.
906 (i6400_msa_short_float2) Likewise.
907 (i6400_msa_div_d) Likewise.
908 (i6400_msa_long_logic1) Likewise.
909 (i6400_msa_long_logic2) Likewise.
910 (i6400_msa_mult) Likewise.
911 (i6400_msa_long_float2) Likewise.
912 (i6400_msa_long_float4) Likewise.
913 (i6400_msa_long_float5) Likewise.
914 (i6400_msa_long_float8) Likewise.
915 (i6400_fpu_fadd): Include frint type.
916 (i6400_fpu_store): New define_insn_reservation.
917 (i6400_fpu_load): Likewise.
918 (i6400_fpu_move): Likewise.
919 (i6400_fpu_fcmp): Likewise.
920 (i6400_fpu_fmadd): Likewise.
921 (i6400_int_mult): Include imul3nc type and update reservation.
922 (i6400_int_div): Include idiv3 type and update reservation.
923 (i6400_int_load): Update to check type not move_type.
924 (i6400_int_store): Likewise.
925 (i6400_int_prefetch): Set zero latency.
927 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
929 * gcc.c: Document new %@{...} sequence.
930 (LINK_COMMAND_SPEC): Use it for the -L switches.
931 (cpp_unique_options): Use it for the -I switches.
932 (at_file_argbuf): New global variable.
933 (in_at_file): Likewise.
934 (alloc_args): Create at_file_argbuf.
935 (clear_args): Truncate at_file_argbuf.
936 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
937 (open_at_file): New function.
938 (close_at_file): Likewise.
939 (create_at_file): Delete.
940 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
943 (validate_switches_from_spec): Deal with %@{...} sequence.
944 (validate_switches): Likewise.
945 (driver::finalize): Call clear_args.
947 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
949 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
951 2018-06-11 Martin Sebor <msebor@redhat.com>
953 * doc/invoke.texi (-Wall): List -Wc++17-compat.
954 (Wno-class-memaccess): Add @opindex.
955 (Wno-templates, Wno-multiple-inheritance): Same.
956 (Wno-virtual-inheritance, Wno-namespaces): Same.
957 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
958 (Wno-format-overflow, Wno-format-truncation): Same.
959 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
960 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
961 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
962 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
963 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
964 (Wno-misspelled-isr): Same.
966 2018-06-11 Martin Sebor <msebor@redhat.com>
968 * PR tree-optimization/86083
969 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
971 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
973 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
975 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
978 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
979 on the correct operand.
980 (*movdi_internal64): Ditto.
982 2018-06-11 Martin Liska <mliska@suse.cz>
984 PR tree-optimization/86089
985 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
987 2018-06-11 Julia Koval <julia.koval@intel.com>
989 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
990 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
991 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
993 2018-06-11 Olivier Hainque <hainque@adacore.com>
995 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
996 for Ada with strict dwarf2.
998 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1001 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1004 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1006 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1008 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1010 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1013 2018-06-08 Martin Liska <mliska@suse.cz>
1015 * tree-cfg.h (debug_function): Fix argument type to match
1018 2018-06-08 Martin Liska <mliska@suse.cz>
1020 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1021 Remove usage of MPX-related (and removed) fields.
1022 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1024 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1026 * cfg.c (debug): Use TDF_NONE rather than 0.
1027 * cfghooks.c (debug): Likewise.
1028 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1029 (struct dump_option_value_info): Convert to...
1030 (struct kv_pair): ...this template type.
1031 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1033 (optinfo_verbosity_options): Likewise.
1034 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1036 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1037 than int for "optgroup_flags" param.
1038 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1040 (dump_dec): Likewise.
1041 (dump_finish): Use TDF_NONE rather than 0.
1042 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1043 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1044 than 0. Update for change to option_ptr.
1045 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1046 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1047 0. Update for changes to optinfo_verbosity_options and
1049 (opt_info_switch_p): Convert optgroup_flags from int to
1051 (dump_basic_block): Use dump_flags_t rather than int
1052 for "dump_kind" param.
1053 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1054 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1055 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1056 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1057 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1058 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1059 TDF_NONE): Convert from macros to...
1060 (enum dump_flag): ...this new enum.
1061 (dump_flags_t): Update to use enum.
1062 (operator|, operator&, operator~, operator|=, operator&=):
1063 Implement for dump_flags_t.
1064 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1065 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1066 Convert from macros to...
1067 (enum optgroup_flag): ...this new enum.
1068 (optgroup_flags_t): New typedef.
1069 (operator|, operator|=): Implement for optgroup_flags_t.
1070 (struct dump_file_info): Convert field "alt_flags" to
1071 dump_flags_t. Convert field "optgroup_flags" to
1073 (dump_basic_block): Use dump_flags_t rather than int for param.
1074 (dump_generic_expr_loc): Likewise.
1075 (dump_dec): Likewise.
1076 (dump_register): Convert param "optgroup_flags" to
1078 (opt_info_enable_passes): Likewise.
1079 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1081 * gimple-pretty-print.c (debug): Likewise.
1082 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1083 (merged_store_group::apply_stores): Likewise.
1084 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1085 * gimple.c (verify_gimple_pp): Likewise.
1086 * graphite-poly.c (print_pbb_body): Likewise.
1087 * passes.c (pass_manager::register_one_dump_file): Convert
1088 local "optgroup_flags" to optgroup_flags_t.
1089 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1091 (debug_body): Likewise.
1092 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1093 to optgroup_flags_t.
1094 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1096 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1097 (convert_mult_to_fma): Likewise.
1098 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1099 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1100 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1101 "dump_kind" to dump_flags_t.
1103 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1105 * config/rs6000/rs6000.c (min, max): Delete.
1107 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1109 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1110 -mabi=spe and -mabi=no-spe.
1112 2018-06-08 Martin Liska <mliska@suse.cz>
1114 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1115 where we expect an existing summary.
1117 2018-06-08 Martin Liska <mliska@suse.cz>
1119 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1120 * ipa-inline.h (estimate_edge_growth): Likewise.
1122 2018-06-08 Martin Liska <mliska@suse.cz>
1124 * cgraph.c (function_version_hasher::hash): Use
1125 cgraph_node::get_uid ().
1126 (function_version_hasher::equal):
1127 * cgraph.h (cgraph_node::get_uid): New method.
1128 * ipa-inline.c (update_caller_keys): Use
1129 cgraph_node::get_uid ().
1130 (update_callee_keys): Likewise.
1131 * ipa-utils.c (searchc): Likewise.
1132 (ipa_reduced_postorder): Likewise.
1133 * lto-cgraph.c (input_node): Likewise.
1134 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1135 * symbol-summary.h (symtab_insertion): Likewise.
1136 (symtab_removal): Likewise.
1137 (symtab_duplication): Likewise.
1138 * tree-pretty-print.c (dump_function_header): Likewise.
1139 * tree-sra.c (convert_callers_for_node): Likewise.
1141 2018-06-08 Martin Liska <mliska@suse.cz>
1143 * cgraph.c (symbol_table::create_edge): Always assign a new
1145 (symbol_table::free_edge): Do not recycle numbers.
1146 * cgraph.h (cgraph_edge::get): New method.
1147 * symbol-summary.h (symtab_removal): Use it.
1148 (symtab_duplication): Likewise.
1149 (call_summary::hashable_uid): Remove.
1151 2018-06-08 Martin Liska <mliska@suse.cz>
1153 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1154 (initialize_growth_caches): Remove.
1155 (free_growth_caches): Likewise.
1156 (do_estimate_edge_time): Use edge_growth_cache.
1157 (do_estimate_edge_size): Likewise.
1158 (do_estimate_edge_hints): Likewise.
1159 * ipa-inline.c (reset_edge_caches): Likewise.
1160 (recursive_inlining): Likewise.
1161 (inline_small_functions): Likewise.
1162 * ipa-inline.h (initialize_growth_caches): Remove.
1163 (estimate_edge_size): Likewise.
1164 (estimate_edge_time): Likewise.
1165 (estimate_edge_hints): Likewise.
1166 (reset_edge_growth_cache): Likewise.
1167 * symbol-summary.h (call_summary::remove): New method.
1169 2018-06-08 Martin Liska <mliska@suse.cz>
1171 * ipa-cp.c (class edge_clone_summary): New summary.
1172 (grow_edge_clone_vectors): Remove.
1173 (ipcp_edge_duplication_hook): Remove.
1174 (class edge_clone_summary_t): New call_summary class.
1175 (ipcp_edge_removal_hook): Remove.
1176 (edge_clone_summary_t::duplicate): New function.
1177 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1178 (create_specialized_node): Likewise.
1179 (ipcp_driver): Initialize edge_clone_summaries and do not
1182 2018-06-08 Martin Liska <mliska@suse.cz>
1184 * symbol-summary.h (get): New function.
1185 (call_summary::m_initialize_when_cloning): New class member.
1187 2018-06-08 Martin Liska <mliska@suse.cz>
1189 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1190 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1191 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1192 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1193 (remove_cgraph_node_from_order): Use the removed_nodes set.
1194 (do_per_function_toporder): Likwise.
1195 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1196 instead of summary_uid.
1197 (symtab_removal): Likewise.
1198 (symtab_duplication): Likewise.
1200 2018-06-08 Martin Liska <mliska@suse.cz>
1202 * ipa-cp.c (ipcp_store_bits_results): Use
1203 ipcp_transformation_sum.
1204 (ipcp_store_vr_results): Likewise.
1205 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1207 (ipcp_transformation_initialize): ... this.
1208 (ipa_set_node_agg_value_chain):
1209 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1210 (write_ipcp_transformation_info): Likewise.
1211 (read_ipcp_transformation_info): Likewise.
1212 (ipcp_update_bits): Likewise.
1213 (ipcp_update_vr): Likewise.
1214 (ipcp_transform_function): Likewise.
1215 * ipa-prop.h: Rename ipcp_transformation_summary to
1216 ipcp_transformation.
1217 (class ipcp_transformation_t): New function summary.
1218 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1219 (ipa_get_agg_replacements_for_node): Likewise.
1221 2018-06-08 Martin Liska <mliska@suse.cz>
1223 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1225 (class funct_state_summary_t): New function_summary class.
1226 (has_function_state): Remove.
1227 (get_function_state): Likewise.
1228 (set_function_state): Likewise.
1229 (add_new_function): Likewise.
1230 (funct_state_summary_t::insert): New function.
1231 (duplicate_node_data): Remove.
1232 (remove_node_data): Remove.
1233 (funct_state_summary_t::duplicate): New function.
1234 (register_hooks): Create new funct_state_summaries.
1235 (pure_const_generate_summary): Use it.
1236 (pure_const_write_summary): Likewise.
1237 (pure_const_read_summary): Likewise.
1238 (propagate_pure_const): Likewise.
1239 (propagate_nothrow): Likewise.
1240 (dump_malloc_lattice): Likewise.
1241 (propagate_malloc): Likewise.
1242 (execute): Do not register hooks, just remove summary
1244 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1247 2018-06-08 Martin Liska <mliska@suse.cz>
1249 * ipa-reference.c (remove_node_data): Remove.
1250 (duplicate_node_data): Likewise.
1251 (class ipa_ref_var_info_summary_t): New class.
1252 (class ipa_ref_opt_summary_t): Likewise.
1253 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1254 (get_reference_optimization_summary): Use
1255 ipa_ref_opt_sum_summaries.
1256 (set_reference_vars_info): Remove.
1257 (set_reference_optimization_summary): Likewise.
1258 (ipa_init): Create summaries.
1259 (init_function_info): Use function summary.
1260 (ipa_ref_opt_summary_t::duplicate): New function.
1261 (ipa_ref_opt_summary_t::remove): New function.
1262 (get_read_write_all_from_node): Fix GNU coding style.
1263 (propagate): Use function summary.
1264 (write_node_summary_p): Fix GNU coding style.
1265 (stream_out_bitmap): Likewise.
1266 (ipa_reference_read_optimization_summary): Use function summary.
1267 (ipa_reference_c_finalize): Do not release hooks.
1269 2018-06-08 Martin Liska <mliska@suse.cz>
1271 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1272 (analyze_function_body): Extract multiple calls of get_create.
1273 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1274 * ipa-inline.c (recursive_inlining): Use ::get method.
1275 * ipa-inline.h (estimate_edge_growth): Likewise.
1277 2018-06-08 Martin Liska <mliska@suse.cz>
1279 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1281 (hsa_function_summary::hsa_function_summary): Use the new enum
1283 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1284 * hsa-gen.c (hsa_get_host_function): Likewise.
1285 (get_brig_function_name): Likewise.
1286 * ipa-hsa.c (process_hsa_functions): Likewise.
1287 (ipa_hsa_write_summary): Likewise.
1288 * symbol-summary.h (symtab_duplication): Use ::get function/
1289 (get): New function.
1291 2018-06-08 Martin Liska <mliska@suse.cz>
1293 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1295 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1296 (hsa_register_kernel): Likewise.
1297 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1298 * hsa-gen.c (hsa_get_host_function): Likewise.
1299 (get_brig_function_name): Likewise.
1300 (generate_hsa): Likewise.
1301 (pass_gen_hsail::execute): Likewise.
1302 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1303 (devirtualization_time_bonus): Likewise.
1304 (ipcp_propagate_stage): Likewise.
1305 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1306 (edge_set_predicate): Likewise.
1307 (evaluate_conditions_for_known_args): Likewise.
1308 (evaluate_properties_for_edge): Likewise.
1309 (ipa_fn_summary::reset): Likewise.
1310 (ipa_fn_summary_t::duplicate): Likewise.
1311 (dump_ipa_call_summary): Likewise.
1312 (ipa_dump_fn_summary): Likewise.
1313 (analyze_function_body): Likewise.
1314 (compute_fn_summary): Likewise.
1315 (estimate_edge_devirt_benefit): Likewise.
1316 (estimate_edge_size_and_time): Likewise.
1317 (estimate_calls_size_and_time): Likewise.
1318 (estimate_node_size_and_time): Likewise.
1319 (inline_update_callee_summaries): Likewise.
1320 (remap_edge_change_prob): Likewise.
1321 (remap_edge_summaries): Likewise.
1322 (ipa_merge_fn_summary_after_inlining): Likewise.
1323 (ipa_update_overall_fn_summary): Likewise.
1324 (read_ipa_call_summary): Likewise.
1325 (inline_read_section): Likewise.
1326 (write_ipa_call_summary): Likewise.
1327 (ipa_fn_summary_write): Likewise.
1328 (ipa_free_fn_summary): Likewise.
1329 * ipa-hsa.c (process_hsa_functions): Likewise.
1330 (ipa_hsa_write_summary): Likewise.
1331 (ipa_hsa_read_section): Likewise.
1332 * ipa-icf.c (sem_function::merge): Likewise.
1333 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1334 (do_estimate_edge_time): Likewise.
1335 (estimate_size_after_inlining): Likewise.
1336 (estimate_growth): Likewise.
1337 (growth_likely_positive): Likewise.
1338 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1339 (inline_call): Likewise.
1340 * ipa-inline.c (caller_growth_limits): Likewise.
1341 (can_inline_edge_p): Likewise.
1342 (can_inline_edge_by_limits_p): Likewise.
1343 (compute_uninlined_call_time): Likewise.
1344 (compute_inlined_call_time): Likewise.
1345 (want_inline_small_function_p): Likewise.
1346 (edge_badness): Likewise.
1347 (update_caller_keys): Likewise.
1348 (update_callee_keys): Likewise.
1349 (recursive_inlining): Likewise.
1350 (inline_small_functions): Likewise.
1351 (inline_to_all_callers_1): Likewise.
1352 (dump_overall_stats): Likewise.
1353 (early_inline_small_functions): Likewise.
1354 (early_inliner): Likewise.
1355 * ipa-inline.h (estimate_edge_growth): Likewise.
1356 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1357 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1358 * ipa-prop.h (IPA_NODE_REF): Likewise.
1359 (IPA_EDGE_REF): Likewise.
1360 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1361 (propagate_malloc): Likewise.
1362 * ipa-split.c (execute_split_functions): Likewise.
1363 * symbol-summary.h: Rename get to get_create.
1365 (get_create): Likewise.
1366 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1368 2018-06-08 Martin Liska <mliska@suse.cz>
1370 * symbol-summary.h (release): Move definition out of class
1372 (symtab_removal): Likewise.
1373 (symtab_duplication): Likewise.
1375 2018-06-08 Martin Liska <mliska@suse.cz>
1377 * symbol-summary.h (function_summary): Move constructor
1378 implementation out of class declaration.
1379 (release): Likewise.
1380 (symtab_insertion): Likewise.
1381 (symtab_removal): Likewise.
1382 (symtab_duplication): Likewise.
1385 2018-06-08 Martin Liska <mliska@suse.cz>
1387 * Makefile.in: Remove support for MPX (macros, related functions,
1388 fields in cgraph_node, ...).
1389 * builtin-types.def (BT_BND): Likewise.
1390 (BT_FN_BND_CONST_PTR): Likewise.
1391 (BT_FN_CONST_PTR_BND): Likewise.
1392 (BT_FN_VOID_PTR_BND): Likewise.
1393 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1394 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1395 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1396 (expand_builtin_mempcpy_with_bounds): Likewise.
1397 (expand_builtin_memset_with_bounds): Likewise.
1398 (expand_builtin_memset_args): Likewise.
1399 (std_expand_builtin_va_start): Likewise.
1400 (expand_builtin): Likewise.
1401 (expand_builtin_with_bounds): Likewise.
1402 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1403 (DEF_LIB_BUILTIN_CHKP): Likewise.
1404 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1405 (DEF_CHKP_BUILTIN): Likewise.
1406 (BUILT_IN_MEMCPY): Likewise.
1407 (BUILT_IN_MEMMOVE): Likewise.
1408 (BUILT_IN_MEMPCPY): Likewise.
1409 (BUILT_IN_MEMSET): Likewise.
1410 (BUILT_IN_STPCPY): Likewise.
1411 (BUILT_IN_STRCAT): Likewise.
1412 (BUILT_IN_STRCHR): Likewise.
1413 (BUILT_IN_STRCPY): Likewise.
1414 (BUILT_IN_STRLEN): Likewise.
1415 (BUILT_IN_MEMCPY_CHK): Likewise.
1416 (BUILT_IN_MEMMOVE_CHK): Likewise.
1417 (BUILT_IN_MEMPCPY_CHK): Likewise.
1418 (BUILT_IN_MEMSET_CHK): Likewise.
1419 (BUILT_IN_STPCPY_CHK): Likewise.
1420 (BUILT_IN_STRCAT_CHK): Likewise.
1421 (BUILT_IN_STRCPY_CHK): Likewise.
1422 * calls.c (store_bounds): Likewise.
1423 (emit_call_1): Likewise.
1424 (special_function_p): Likewise.
1425 (maybe_warn_nonstring_arg): Likewise.
1426 (initialize_argument_information): Likewise.
1427 (finalize_must_preallocate): Likewise.
1428 (compute_argument_addresses): Likewise.
1429 (expand_call): Likewise.
1430 * cfgexpand.c (expand_call_stmt): Likewise.
1431 (expand_return): Likewise.
1432 (expand_gimple_stmt_1): Likewise.
1433 (pass_expand::execute): Likewise.
1434 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1435 (cgraph_node::remove): Likewise.
1436 (cgraph_node::dump): Likewise.
1437 (cgraph_node::verify_node): Likewise.
1438 * cgraph.h (chkp_function_instrumented_p): Likewise.
1439 (symtab_node::get_alias_target): Likewise.
1440 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1441 (cgraph_local_p): Likewise.
1442 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1443 (cgraph_edge::rebuild_references): Likewise.
1444 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1445 (walk_polymorphic_call_targets): Likewise.
1446 (cgraph_node::expand_thunk): Likewise.
1447 (symbol_table::output_weakrefs): Likewise.
1448 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1449 (ix86_handle_option): Likewise.
1450 * config/i386/constraints.md: Likewise.
1451 * config/i386/i386-builtin-types.def (BND): Likewise.
1455 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1456 (BDESC_FIRST): Likewise.
1458 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1459 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1460 * config/i386/i386.c (enum reg_class): Likewise.
1461 (ix86_target_string): Likewise.
1462 (ix86_option_override_internal): Likewise.
1463 (ix86_conditional_register_usage): Likewise.
1464 (ix86_valid_target_attribute_inner_p): Likewise.
1465 (ix86_set_indirect_branch_type): Likewise.
1466 (ix86_set_current_function): Likewise.
1467 (ix86_function_arg_regno_p): Likewise.
1468 (init_cumulative_args): Likewise.
1469 (ix86_function_arg_advance): Likewise.
1470 (ix86_function_arg): Likewise.
1471 (ix86_pass_by_reference): Likewise.
1472 (ix86_function_value_regno_p): Likewise.
1473 (ix86_function_value_1): Likewise.
1474 (ix86_function_value_bounds): Likewise.
1475 (ix86_return_in_memory): Likewise.
1476 (ix86_setup_incoming_vararg_bounds): Likewise.
1477 (ix86_va_start): Likewise.
1478 (indirect_thunk_need_prefix): Likewise.
1479 (print_reg): Likewise.
1480 (ix86_print_operand): Likewise.
1481 (ix86_expand_call): Likewise.
1482 (ix86_output_function_return): Likewise.
1483 (reg_encoded_number): Likewise.
1484 (BDESC_VERIFYS): Likewise.
1485 (ix86_init_mpx_builtins): Likewise.
1486 (ix86_init_builtins): Likewise.
1487 (ix86_emit_cmove): Likewise.
1488 (ix86_emit_move_max): Likewise.
1489 (ix86_expand_builtin): Likewise.
1490 (ix86_builtin_mpx_function): Likewise.
1491 (ix86_get_arg_address_for_bt): Likewise.
1492 (ix86_load_bounds): Likewise.
1493 (ix86_store_bounds): Likewise.
1494 (ix86_load_returned_bounds): Likewise.
1495 (ix86_store_returned_bounds): Likewise.
1496 (ix86_class_likely_spilled_p): Likewise.
1497 (ix86_hard_regno_mode_ok): Likewise.
1498 (x86_order_regs_for_local_alloc): Likewise.
1499 (ix86_mitigate_rop): Likewise.
1500 (ix86_bnd_prefixed_insn_p): Likewise.
1501 (ix86_mpx_bound_mode): Likewise.
1502 (ix86_make_bounds_constant): Likewise.
1503 (ix86_initialize_bounds): Likewise.
1504 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1505 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1506 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1507 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1508 (TARGET_CHKP_BOUND_MODE): Likewise.
1509 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1510 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1511 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1512 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1513 * config/i386/i386.h (TARGET_MPX): Likewise.
1514 (TARGET_MPX_P): Likewise.
1515 (VALID_BND_REG_MODE): Likewise.
1516 (FIRST_BND_REG): Likewise.
1517 (LAST_BND_REG): Likewise.
1518 (enum reg_class): Likewise.
1519 (BND_REG_P): Likewise.
1520 (BND_REGNO_P): Likewise.
1521 (BNDmode): Likewise.
1522 (ADJUST_INSN_LENGTH): Likewise.
1523 * config/i386/i386.md: Likewise.
1524 * config/i386/i386.opt: Likewise.
1525 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1526 (defined): Likewise.
1527 (LINK_MPX): Likewise.
1528 (MPX_SPEC): Likewise.
1529 (LIBMPX_SPEC): Likewise.
1530 (LIBMPXWRAPPERS_SPEC): Likewise.
1531 (CHKP_SPEC): Likewise.
1532 * config/i386/predicates.md: Likewise.
1533 * dbxout.c (dbxout_type): Likewise.
1534 * doc/extend.texi: Likewise.
1535 * doc/invoke.texi: Likewise.
1536 * doc/md.texi: Likewise.
1537 * doc/tm.texi: Likewise.
1538 * doc/tm.texi.in: Likewise.
1539 * dwarf2out.c (is_base_type): Likewise.
1540 (gen_formal_types_die): Likewise.
1541 (gen_subprogram_die): Likewise.
1542 (gen_type_die_with_usage): Likewise.
1543 (gen_decl_die): Likewise.
1544 (dwarf2out_late_global_decl): Likewise.
1545 * expr.c (expand_assignment): Likewise.
1546 (emit_storent_insn): Likewise.
1547 (store_expr_with_bounds): Likewise.
1548 (store_expr): Likewise.
1549 (expand_expr_real_1): Likewise.
1550 * expr.h (store_expr_with_bounds): Likewise.
1551 * function.c (use_register_for_decl): Likewise.
1552 (struct bounds_parm_data): Likewise.
1553 (assign_parms_augmented_arg_list): Likewise.
1554 (assign_parm_find_entry_rtl): Likewise.
1555 (assign_parm_is_stack_parm): Likewise.
1556 (assign_parm_load_bounds): Likewise.
1557 (assign_bounds): Likewise.
1558 (assign_parms): Likewise.
1559 (expand_function_start): Likewise.
1560 * gcc.c (CHKP_SPEC): Likewise.
1561 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1562 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1563 (wrestrict_dom_walker::check_call): Likewise.
1564 * gimple.c (gimple_build_call_from_tree): Likewise.
1565 * gimple.h (enum gf_mask): Likewise.
1566 (gimple_call_with_bounds_p): Likewise.
1567 (gimple_call_set_with_bounds): Likewise.
1568 * gimplify.c (gimplify_init_constructor): Likewise.
1569 * ipa-cp.c (initialize_node_lattices): Likewise.
1570 (propagate_constants_across_call): Likewise.
1571 (find_more_scalar_values_for_callers_subset): Likewise.
1572 * ipa-hsa.c (process_hsa_functions): Likewise.
1573 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1574 * ipa-icf.c (sem_function::merge): Likewise.
1575 * ipa-inline.c (early_inliner): Likewise.
1576 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1577 (warn_function_cold): Likewise.
1578 (propagate_pure_const): Likewise.
1579 * ipa-ref.h (enum GTY): Likewise.
1580 * ipa-split.c (find_retbnd): Likewise.
1581 (consider_split): Likewise.
1582 (split_function): Likewise.
1583 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1584 * ipa.c (walk_polymorphic_call_targets): Likewise.
1585 (symbol_table::remove_unreachable_nodes): Likewise.
1586 (process_references): Likewise.
1587 (cgraph_build_static_cdtor_1): Likewise.
1588 * lto-cgraph.c (lto_output_node): Likewise.
1589 (output_refs): Likewise.
1590 (compute_ltrans_boundary): Likewise.
1591 (input_overwrite_node): Likewise.
1592 (input_node): Likewise.
1593 (input_cgraph_1): Likewise.
1594 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1595 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1596 (class pass_chkp_instrumentation_passes): Likewise.
1597 (make_pass_chkp_instrumentation_passes): Likewise.
1598 * passes.def: Likewise.
1599 * rtl.h (struct GTY): Likewise.
1600 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1601 * stor-layout.c (layout_type): Likewise.
1602 * symtab.c: Likewise.
1603 * target.def: Likewise.
1604 * targhooks.c (default_chkp_bound_type): Likewise.
1605 (default_chkp_bound_mode): Likewise.
1606 (default_builtin_chkp_function): Likewise.
1607 (default_chkp_function_value_bounds): Likewise.
1608 (default_chkp_make_bounds_constant): Likewise.
1609 (default_chkp_initialize_bounds): Likewise.
1610 * targhooks.h (default_chkp_bound_type): Likewise.
1611 (default_chkp_bound_mode): Likewise.
1612 (default_builtin_chkp_function): Likewise.
1613 (default_chkp_function_value_bounds): Likewise.
1614 (default_chkp_make_bounds_constant): Likewise.
1615 (default_chkp_initialize_bounds): Likewise.
1616 * toplev.c (compile_file): Likewise.
1617 (process_options): Likewise.
1618 * tree-core.h (DEF_BUILTIN): Likewise.
1619 (DEF_BUILTIN_CHKP): Likewise.
1620 * tree-inline.c (declare_return_variable): Likewise.
1621 (remap_gimple_stmt): Likewise.
1622 (copy_bb): Likewise.
1623 (initialize_inlined_parameters): Likewise.
1624 (expand_call_inline): Likewise.
1625 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1626 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1627 (make_pass_ipa_chkp_produce_thunks): Likewise.
1628 (make_pass_chkp): Likewise.
1629 (make_pass_chkp_opt): Likewise.
1630 (make_pass_chkp_instrumentation_passes): Likewise.
1631 * tree-pretty-print.c (dump_generic_node): Likewise.
1632 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1633 * tree-ssa-dce.c (propagate_necessity): Likewise.
1634 (eliminate_unnecessary_stmts): Likewise.
1635 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1636 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1637 * tree-ssa-sccvn.h: Likewise.
1638 * tree-ssa-strlen.c (get_string_length): Likewise.
1639 (valid_builtin_call): Likewise.
1640 (adjust_last_stmt): Likewise.
1641 (handle_builtin_strchr): Likewise.
1642 (handle_builtin_strcpy): Likewise.
1643 (handle_builtin_stxncpy): Likewise.
1644 (handle_builtin_memcpy): Likewise.
1645 (handle_builtin_strcat): Likewise.
1646 (strlen_check_and_optimize_stmt): Likewise.
1647 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1648 * tree-streamer-in.c: Likewise.
1649 * tree-streamer.c (record_common_node): Likewise.
1650 * tree.c (tree_code_size): Likewise.
1651 (wide_int_to_tree_1): Likewise.
1652 (type_contains_placeholder_1): Likewise.
1653 (build_common_tree_nodes): Likewise.
1654 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1655 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1656 (POINTER_BOUNDS_P): Likewise.
1657 (BOUNDED_TYPE_P): Likewise.
1658 (BOUNDED_P): Likewise.
1659 (CALL_WITH_BOUNDS_P): Likewise.
1660 (pointer_bounds_type_node): Likewise.
1661 * value-prof.c (gimple_ic): Likewise.
1662 * var-tracking.c (vt_add_function_parameters): Likewise.
1663 * varasm.c (make_decl_rtl): Likewise.
1664 (assemble_start_function): Likewise.
1665 (output_constant): Likewise.
1666 (maybe_assemble_visibility): Likewise.
1667 * varpool.c (ctor_for_folding): Likewise.
1668 * chkp-builtins.def: Remove.
1669 * ipa-chkp.c: Remove.
1670 * ipa-chkp.h: Remove.
1671 * rtl-chkp.c: Remove.
1672 * rtl-chkp.h: Remove.
1673 * tree-chkp-opt.c: Remove.
1674 * tree-chkp.c: Remove.
1675 * tree-chkp.h: Remove.
1677 2018-06-07 Carl Love <cel@us.ibm.com>
1679 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1680 vextract_fp_from_shortl): Add BE support.
1682 2018-06-07 Paul Koning <ni1d@arrl.net>
1684 * compare-elim.c (try_merge_compare): Don't merge compare if
1685 address contains a side effect.
1686 (try_eliminate_compare): Likewise.
1688 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1690 * config.gcc: Support "tremont".
1691 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1692 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1694 * config/i386/i386.c (m_TREMONT): Define.
1695 (processor_target_table): Add "tremont".
1696 (PTA_TREMONT): Define.
1697 (ix86_lea_outperforms): Add TARGET_TREMONT.
1698 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1699 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1700 and M_INTEL_GOLDMONT_PLUS.
1701 (fold_builtin_cpu): Add "tremont".
1702 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1703 (ix86_option_override_internal): Add "tremont".
1704 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1705 (processor_type): Add PROCESSOR_TREMONT.
1706 * config/i386/x86-tune.def: Add m_TREMONT.
1707 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1709 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1711 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1712 symbol defined for msp430i* devices to be lower case.
1714 gcc/testsuite/gcc.target/msp430/
1715 * msp430i-device-symbol.c: New test.
1716 * msp430f-device-symbol.c: New test.
1717 * msp430.h: New test header file.
1719 2018-06-07 Richard Biener <rguenther@suse.de>
1721 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1722 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1723 Properly wrap signed arithmetic if overflow wraps.
1725 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1727 PR tree-optimization/69615
1728 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1729 of a cast from a same precision integral SSA_NAME in a bb dominated
1730 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1731 cast to utype if rhs2 has already a compatible type.
1733 2018-06-07 Richard Biener <rguenther@suse.de>
1735 PR tree-optimization/85935
1736 * graphite-scop-detection.c (find_params_in_bb): Analyze
1737 condition operands with respect to the correct loop. Assert
1738 the analysis doesn't fail.
1740 2018-06-04 Carl Love <cel@us.ibm.com>
1742 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1743 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1744 as it is slightly cheaper.
1745 (first_match_or_eos_index_<mode>):
1746 Calculate index using natural element order.
1747 (first_match_index_<mode>):
1748 Calculate index using natural element order.
1749 (first_match_or_eos_index_<mode>):
1750 Calculate index using natural order.
1751 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1752 for BE and LE modes.
1753 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1754 P9V_BUILTIN_VCLZLSBB_V16QI.
1755 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1758 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1760 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1761 indentation and line wrap for many prototypes. Add missing
1762 @smallexample directives around block of prototypes for vec_xl and
1765 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1767 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1768 track if we pass or return IEEE 128-bit floating point.
1769 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1770 C++ mangling that is compatible with GCC 8.1.
1771 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1772 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1773 floating point types.
1774 (rs6000_function_arg_advance_1): Likewise.
1775 (rs6000_mangle_type): Optionally generate mangled names that match
1776 what GCC 8.1 generated for IEEE 128-bit floating point types.
1777 (rs6000_globalize_decl_name): If we have an external function that
1778 passes or returns IEEE 128-bit types, generate a weak reference
1779 from the mangled name used in GCC 8.1 to the current mangled
1781 (rs6000_init_builtins): Make __ibm128 use the long double type if
1782 long double is IBM extended double. Make __float128 use the long
1783 double type if long double is IEEE 128-bit.
1786 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1787 macro for __ibm128 built-in functions.
1788 (PACK_IF): Add __ibm128 pack/unpack functions.
1789 (UNPACK_IF): Likewise.
1790 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1791 enable long double built-in functions if long double is IEEE
1792 128-bit floating point.
1793 (rs6000_invalid_builtin): Update long double built-in function
1795 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1796 functions, adjust the built-in function to use the long double
1797 built-in function if __ibm128 and long double are the same type.
1798 * doc/extend.texi (PowerPC builtins): Update documention for
1799 __builtin_{,un}pack_longdouble. Add documentation for
1800 __builtin_{,un}pack_ibm128.
1802 2018-06-06 Jim Wilson <jimw@sifive.com>
1804 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1805 (struct machine_function): New field interrupt_mode.
1806 (riscv_handle_type_attribute): New function. Add forward declaration.
1807 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1808 (riscv_expand_epilogue): Check interrupt_mode field.
1809 (riscv_set_current_function): Check interrupt attribute args and
1810 set interrupt_mode field.
1811 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1812 (riscv_sret, riscv_uret): New.
1813 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1814 new arguments to interrupt attribute.
1816 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1819 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1820 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1822 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1824 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1825 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1826 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1827 several redundant entries.
1829 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1831 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1832 type from "rtx" to "rtx_insn *".
1833 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1834 for local "call_insn", removing cast.
1835 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1837 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1839 PR tree-optimization/86066
1840 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1841 for BIT_INSERT_EXPR stores.
1843 2018-06-06 Richard Biener <rguenther@suse.de>
1845 PR tree-optimization/86062
1846 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1847 component refs ontop
1848 of to be offsetted base.
1850 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1852 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1853 to be static and remove check on interrupt attribute name.
1855 gcc/testsuite/gcc.target/msp430/
1856 * function-attributes-4.c: New test.
1857 * static-interrupts.c: New test.
1859 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1861 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1862 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1864 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1867 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1869 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1870 Remove second argument, change how error is called.
1871 (aarch64_layout_arg): Remove second argument from
1872 aarch64_err_no_fpadvsimd call.
1873 (aarch64_init_cumulative_args): Ditto.
1874 (aarch64_gimplify_va_arg_expr): Ditto.
1875 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1877 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1879 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1880 (*simple_return_indirect_internal<mode>): Rename from
1881 simple_return_indirect_internal. Use W mode iterator.
1882 (rstorssp): New expander.
1883 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1884 (clrssbsy): New expander.
1885 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1887 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1889 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1891 (cmse_check_pointed_object): Likewise.
1893 2018-06-05 Martin Liska <mliska@suse.cz>
1895 PR gcov-profile/47618
1896 * doc/invoke.texi: Document how -fprofile-dir format
1899 2018-06-05 Richard Biener <rguenther@suse.de>
1901 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1902 removal pretend DOM info isn't available so we do not update
1903 it and only remove edges, not dominated blocks. Actually free
1904 DOM info in case we removed something. Remove unreachable blocks.
1905 (mfb_keep_latches): Work with either DOM info or marked backedges.
1906 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1907 first. Mark backedges if DOM info isn't available.
1908 (Re-)compute DOM info after cleanup_control_flow_pre.
1910 2018-06-05 Richard Biener <rguenther@suse.de>
1912 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1913 (locus_discrim_hasher::hash): Adjust.
1914 (locus_discrim_hasher::equal): Likewise.
1915 (next_discriminator_for_locus): Work on line directly.
1916 (same_line_p): Pass in expanded locus1 as well.
1917 (assign_discriminators): Avoid redundant location expansions.
1919 2018-06-05 Richard Biener <rguenther@suse.de>
1921 PR tree-optimization/86046
1922 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1923 if required after clearing TREE_ADDRESSABLE.
1925 2018-06-05 Richard Biener <rguenther@suse.de>
1927 PR tree-optimization/86047
1928 * tree-ssa-loop.c (for_each_index): Glob handling of all
1929 decls and constants and really handle all of them.
1931 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1934 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1935 qualifier_void_pointer and qualifier_const_void_pointer.
1936 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1937 (arm_init_builtins): Handle the above.
1938 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1939 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1940 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1943 2018-06-05 Martin Liska <mliska@suse.cz>
1945 * auto-profile.c (read_autofdo_file): Do not use
1946 gcov_ctr_summary struct.
1947 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1948 * coverage.c (struct counts_entry): Likewise.
1949 (read_counts_file): Read just single summary entry.
1950 (get_coverage_counts): Use gcov_summary struct.
1951 * coverage.h (get_coverage_counts): Likewise.
1952 * gcov-dump.c (dump_working_sets): Likewise.
1953 (tag_summary): Dump just single summary.
1954 * gcov-io.c (gcov_write_summary): Write just histogram
1956 (gcov_read_summary): Read just single summary.
1957 (compute_working_sets): Use gcov_summary struct.
1958 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1959 of GCOV_COUNTERS_SUMMABLE.
1960 (GCOV_COUNTERS_SUMMABLE): Remove.
1961 (GCOV_FIRST_VALUE_COUNTER): Replace with
1962 GCOV_COUNTER_V_INTERVAL.
1963 (struct gcov_ctr_summary): Remove.
1964 (struct gcov_summary): Directly use fields of former
1966 (compute_working_sets): Use gcov_summary struct.
1967 * gcov.c (read_count_file): Do not use ctrs fields.
1968 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1970 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1972 * profile.c: Likewise.
1973 * profile.h: Likewise.
1975 2018-06-05 Martin Liska <mliska@suse.cz>
1977 PR gcov-profile/84846
1978 * gcov.c (output_lines): Print working directory only
1979 in intermediate format.
1981 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1983 * config/s390/s390-builtin-types.def: Add void function type.
1984 * config/s390/s390-builtins.def: Use the function type for the
1987 2018-06-04 Jim Wilson <jimw@sifive.com>
1989 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1991 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1992 and maybe_eh_return. Change regno to unsigned int. Use new args to
1993 handle EH_RETURN_DATA_REGNO registers properly.
1994 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1995 (riscv_expand_epilogue): Update comment. Change argument name and
1996 type. Update code to use new name and type. Pass new args to
1997 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1999 * config/riscv/riscv.md (NORMAL_RETURN): New.
2000 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2001 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2002 (eh_return): Call gen_eh_return_internal and emit barrier.
2003 (eh_return_internal): Call riscv_expand_epilogue.
2005 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2007 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2008 bit_insertion field and declare can_be_merged_into method.
2009 (merged_store_group::can_be_merged_into): New method.
2010 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2011 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2012 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2014 2018-06-04 Richard Biener <rguenther@suse.de>
2016 PR tree-optimization/85955
2017 * builtins.c (fold_builtin_sincos): Convert pointers to
2018 destination to appropriate type before dereferencing.
2020 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2022 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2024 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2026 * expr.c (expand_expr_real_1): Force the operand into memory if
2027 its TYPE_MODE is BLKmode and if there is no integer mode for
2028 the number of bits being extracted.
2030 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2034 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2035 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2037 2018-06-04 Richard Biener <rguenther@suse.de>
2039 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2040 (cleanup_tree_cfg_noloop): ... single caller. Do
2041 start_recording_case_labels later.
2043 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2045 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2046 to _IMMINTRIN_H_INCLUDED.
2047 * config/i386/pconfigintrin.h: Ditto.
2048 * config/i386/waitpkgintrin.h: Ditto.
2049 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2050 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2051 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2052 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2053 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2054 waitpkgintrin.h and cldemoteintrin.h.
2056 2018-06-04 Richard Biener <rguenther@suse.de>
2058 PR tree-optimization/86038
2059 * tracer.c (find_best_successor): Check probability for
2060 being initialized, bail out if not.
2062 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2065 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2066 of bits to ignore when comparing architectures.
2068 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2070 PR tree-optimization/69615
2071 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2072 maximum or minimum of the type, try to merge it also as if
2073 range1 is + [-, x - 1] or + [x + 1, -].
2076 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2078 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2080 PR tree-optimization/86034
2081 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2082 the unsigned bitfield type in a bit insertion sequence if it does not
2083 have a larger precision than the bitfield size.
2084 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2086 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2088 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2090 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2092 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2093 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2094 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2095 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2097 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2099 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2100 Disable -fdelete-null-pointer-checks for ELF toolchain.
2102 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2103 Kito Cheng <kito.cheng@gmail.com>
2105 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2106 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2107 (nds32*-*-*): Add glibc and uclibc conditions.
2108 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2109 (TARGET_EXCEPT_UNWIND_INFO): Define.
2110 * config/nds32/elf.h: New file.
2111 * config/nds32/linux.h: New file.
2112 * config/nds32/nds32-elf.opt: New file.
2113 * config/nds32/nds32-linux.opt: New file.
2114 * config/nds32/nds32-fp-as-gp.c
2115 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2116 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2118 (nds32_asm_file_end): Ditto.
2119 (nds32_print_operand): Ditto.
2120 (nds32_insert_attributes): Ditto.
2121 (nds32_init_libfuncs): New function.
2122 (TARGET_HAVE_TLS): Define.
2123 (TARGET_INIT_LIBFUNCS): Define.
2124 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2126 (TARGET_ELF): Apply different mcmodel setting.
2127 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2128 been migrated into elf.h and linux.h files.
2129 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2130 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2131 (mcmodel): The content has been migrated into nds32-elf.opt and
2132 nds32-linux.opt files.
2133 * config/nds32/t-elf: New file.
2134 * config/nds32/t-linux: New file.
2136 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2137 Shiva Chen <shiva0217@gmail.com>
2139 * config/nds32/constants.md (unspec_volatile_element): Add
2140 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2141 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2143 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2144 (make_pass_nds32_fp_as_gp): Declare.
2145 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2147 (nds32_asm_function_end_prologue): Remove unused asm output.
2148 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2149 (nds32_asm_file_start): Output necessary fp_as_gp information.
2150 (nds32_option_override): Adjust register usage.
2151 (nds32_expand_prologue): Consider fp_as_gp situation.
2152 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2153 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2156 (simple_return): Ditto.
2157 (omit_fp_begin): Output special directive for fp_as_gp.
2158 (omit_fp_end): Output special directive for fp_as_gp.
2159 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2160 mforbid-fp-as-gp): New options.
2162 2018-06-01 Mark Wielaard <mark@klomp.org>
2164 * dwarf2out.c (dwarf2out_finish): Remove generation of
2165 DW_AT_loclists_base.
2167 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2169 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2170 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2171 (struct merged_store_group): Add bit_insertion field.
2172 (dump_char_array): Use standard hexadecimal format.
2173 (merged_store_group::merged_store_group): Set bit_insertion to false.
2174 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2175 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2176 also print the mask in the dump file.
2177 (pass_store_merging::gate): Minor tweak.
2178 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2179 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2180 stores with INTEGER_CST stores.
2181 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2182 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2183 and use it throughout. Generate bit insertion sequences if need be.
2184 (pass_store_merging::process_store): Remove redundant condition.
2185 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2187 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2189 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2190 the 128-bit floating point types. Fix function comment.
2192 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2194 * config/aarch64/aarch64-simd.md
2195 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2197 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2200 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2202 PR tree-optimization/85989
2203 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2205 (backprop::intersect_uses): Check it when deciding whether this
2206 is a backedge reference.
2207 (backprop::process_block): Add each phi to m_visited_phis
2208 after visiting it, then clear it at the end.
2210 2018-06-01 Richard Biener <rguenther@suse.de>
2212 * tree-vectorizer.h (vect_dr_stmt): New function.
2213 (vect_get_load_cost): Adjust.
2214 (vect_get_store_cost): Likewise.
2215 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2216 Use vect_dr_stmt instead of DR_SMTT.
2217 (vect_record_base_alignments): Likewise.
2218 (vect_calculate_target_alignment): Likewise.
2219 (vect_compute_data_ref_alignment): Likewise and make static.
2220 (vect_update_misalignment_for_peel): Likewise.
2221 (vect_verify_datarefs_alignment): Likewise.
2222 (vector_alignment_reachable_p): Likewise.
2223 (vect_get_data_access_cost): Likewise. Pass down
2224 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2225 (vect_get_peeling_costs_all_drs): Likewise.
2226 (vect_peeling_hash_get_lowest_cost): Likewise.
2227 (vect_enhance_data_refs_alignment): Likewise.
2228 (vect_find_same_alignment_drs): Likewise.
2229 (vect_analyze_data_refs_alignment): Likewise.
2230 (vect_analyze_group_access_1): Likewise.
2231 (vect_analyze_group_access): Likewise.
2232 (vect_analyze_data_ref_access): Likewise.
2233 (vect_analyze_data_ref_accesses): Likewise.
2234 (vect_vfa_segment_size): Likewise.
2235 (vect_small_gap_p): Likewise.
2236 (vectorizable_with_step_bound_p): Likewise.
2237 (vect_prune_runtime_alias_test_list): Likewise.
2238 (vect_analyze_data_refs): Likewise.
2239 (vect_supportable_dr_alignment): Likewise.
2240 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2241 (vect_gen_prolog_loop_niters): Likewise.
2242 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2243 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2245 (vect_recog_mask_conversion_pattern): Likewise.
2246 (vect_try_gather_scatter_pattern): Likewise.
2247 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2248 to vect_get_store_cost.
2249 (vect_get_store_cost): Get stmt_info instead of DR.
2250 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2251 (vect_get_load_cost): Get stmt_info instead of DR.
2253 2018-06-01 Richard Biener <rguenther@suse.de>
2256 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2257 vars wrapped in MEM_REFs.
2259 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2261 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2262 Fix subreg tests so that we only return a choice between
2263 GENERAL_REGS and FP_REGS if the original classes included both.
2265 2018-06-01 Richard Biener <rguenther@suse.de>
2268 * tree-ssa-structalias.c (get_function_part_constraint):
2269 Handle NULL fi->decl.
2270 (find_func_aliases_for_call): Properly handle indirect
2271 fi from direct call.
2272 (find_func_clobbers): Likewise.
2273 (ipa_pta_execute): Likewise.
2274 (create_variable_info_for): For functions that are ifunc_resolver
2275 resolve to a varinfo that contains the result of the resolver call.
2276 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2279 2018-05-31 Michael Collison <michael.collison@arm.com>
2281 * config/aarch64/aarch64.md:
2282 (*fix_to_zero_extenddfdi2): New pattern.
2283 * gcc.target/aarch64/fix_extend1.c: New testcase.
2285 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2289 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2290 and BUILT_IN_STRNCMP_EQ.
2291 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2292 BUILT_IN_STRNCMP_EQ.
2293 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2294 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2295 (gimple_fold_builtin): Likewise.
2296 * tree-ssa-strlen.c (compute_string_length): New function.
2297 (determine_min_obsize): New function.
2298 (handle_builtin_string_cmp): New function to handle calls to
2299 string compare functions.
2300 (strlen_optimize_stmt): Add handling to builtin string compare
2302 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2303 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2304 * tree.c (build_common_builtin_nodes): Add new defines of
2305 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2307 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2310 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2311 functions with naked attribute.
2313 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2315 * config/i386/sse.md (avx_vec_concat<mode>):
2316 Substitute concat_tg_mode mode attribute with xtg_mode.
2317 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2318 (concat_tg_mode): Remove mode attribute.
2320 2018-05-31 Martin Sebor <msebor@redhat.com>
2323 * calls.c (alloc_max_size): Correct a logic error/typo.
2324 Treat excessive arguments as infinite. Warn for invalid arguments.
2325 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2327 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2330 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2331 and movx for Haswell.
2333 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2334 Cesar Philippidis <cesar@codesourcery.com>
2337 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2338 when emitting error on private/firstprivate reductions.
2339 * omp-low.c (lower_omp_target): Avoid reference-type processing
2340 on pointers for firstprivate clause.
2342 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2344 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2347 * config/aarch64/aarch64-simd.md
2348 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2349 (aarch64_ld1_x3_<mode>): Likewise
2350 (aarch64_st1x2<VALLDIF:mode>): Likewise
2351 (aarch64_st1_x2_<mode>): Likewise
2352 (aarch64_st1x3<VALLDIF:mode>): Likewise
2353 (aarch64_st1_x3_<mode>): Likewise
2354 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2355 (vld1_s8_x3): Likewise.
2356 (vld1_u16_x3): Likewise.
2357 (vld1_s16_x3): Likewise.
2358 (vld1_u32_x3): Likewise.
2359 (vld1_s32_x3): Likewise.
2360 (vld1_u64_x3): Likewise.
2361 (vld1_s64_x3): Likewise.
2362 (vld1_f16_x3): Likewise.
2363 (vld1_f32_x3): Likewise.
2364 (vld1_f64_x3): Likewise.
2365 (vld1_p8_x3): Likewise.
2366 (vld1_p16_x3): Likewise.
2367 (vld1_p64_x3): Likewise.
2368 (vld1q_u8_x3): Likewise.
2369 (vld1q_s8_x3): Likewise.
2370 (vld1q_u16_x3): Likewise.
2371 (vld1q_s16_x3): Likewise.
2372 (vld1q_u32_x3): Likewise.
2373 (vld1q_s32_x3): Likewise.
2374 (vld1q_u64_x3): Likewise.
2375 (vld1q_s64_x3): Likewise.
2376 (vld1q_f16_x3): Likewise.
2377 (vld1q_f32_x3): Likewise.
2378 (vld1q_f64_x3): Likewise.
2379 (vld1q_p8_x3): Likewise.
2380 (vld1q_p16_x3): Likewise.
2381 (vld1q_p64_x3): Likewise.
2382 (vst1_s64_x2): Likewise.
2383 (vst1_u64_x2): Likewise.
2384 (vst1_f64_x2): Likewise.
2385 (vst1_s8_x2): Likewise.
2386 (vst1_p8_x2): Likewise.
2387 (vst1_s16_x2): Likewise.
2388 (vst1_p16_x2): Likewise.
2389 (vst1_s32_x2): Likewise.
2390 (vst1_u8_x2): Likewise.
2391 (vst1_u16_x2): Likewise.
2392 (vst1_u32_x2): Likewise.
2393 (vst1_f16_x2): Likewise.
2394 (vst1_f32_x2): Likewise.
2395 (vst1_p64_x2): Likewise.
2396 (vst1q_s8_x2): Likewise.
2397 (vst1q_p8_x2): Likewise.
2398 (vst1q_s16_x2): Likewise.
2399 (vst1q_p16_x2): Likewise.
2400 (vst1q_s32_x2): Likewise.
2401 (vst1q_s64_x2): Likewise.
2402 (vst1q_u8_x2): Likewise.
2403 (vst1q_u16_x2): Likewise.
2404 (vst1q_u32_x2): Likewise.
2405 (vst1q_u64_x2): Likewise.
2406 (vst1q_f16_x2): Likewise.
2407 (vst1q_f32_x2): Likewise.
2408 (vst1q_f64_x2): Likewise.
2409 (vst1q_p64_x2): Likewise.
2410 (vst1_s64_x3): Likewise.
2411 (vst1_u64_x3): Likewise.
2412 (vst1_f64_x3): Likewise.
2413 (vst1_s8_x3): Likewise.
2414 (vst1_p8_x3): Likewise.
2415 (vst1_s16_x3): Likewise.
2416 (vst1_p16_x3): Likewise.
2417 (vst1_s32_x3): Likewise.
2418 (vst1_u8_x3): Likewise.
2419 (vst1_u16_x3): Likewise.
2420 (vst1_u32_x3): Likewise.
2421 (vst1_f16_x3): Likewise.
2422 (vst1_f32_x3): Likewise.
2423 (vst1_p64_x3): Likewise.
2424 (vst1q_s8_x3): Likewise.
2425 (vst1q_p8_x3): Likewise.
2426 (vst1q_s16_x3): Likewise.
2427 (vst1q_p16_x3): Likewise.
2428 (vst1q_s32_x3): Likewise.
2429 (vst1q_s64_x3): Likewise.
2430 (vst1q_u8_x3): Likewise.
2431 (vst1q_u16_x3): Likewise.
2432 (vst1q_u32_x3): Likewise.
2433 (vst1q_u64_x3): Likewise.
2434 (vst1q_f16_x3): Likewise.
2435 (vst1q_f32_x3): Likewise.
2436 (vst1q_f64_x3): Likewise.
2437 (vst1q_p64_x3): Likewise.
2439 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2441 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2442 user_label_prefix to name.
2444 * tree-core.h: Update comment about the format of NAME string
2445 passed to handler in attribute_spec.
2447 * config/msp430/msp430.md: Remove erroneous subreg expression from
2448 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2449 zero_extend{q,h}isi2.
2451 2018-05-30 Borislav Petkov <bp@suse.de>
2453 * doc/extend.texi: Document some architecture specific
2454 constraints and sort entries.
2456 2018-05-30 Martin Sebor <msebor@redhat.com>
2459 * builtins.c (expand_builtin_stpcpy_1): New function.
2460 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2461 only if the former succeeds.
2463 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2465 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2468 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2470 * doc/invoke.texi (-flinker-output): Document
2472 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2474 * passes.c (ipa_write_summaries): Only modify statements if body
2476 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2477 incrementally linking.
2478 (ipa_passes): Likewise.
2479 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2480 pass down resolution info.
2481 * common.opt (flag_incremental_link): Update info.
2482 * gcc.c (plugin specs): Turn flinker-output=* to
2483 -plugin-opt=-linker-output-known
2484 * toplev.c (compile_file): Also cut compilation when doing incremental
2486 * flag-types. (enum lto_partition_model): Add
2487 LTO_LINKER_OUTPUT_NOLTOREL.
2488 (invoke.texi): Add -flinker-output docs.
2489 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2490 link same way as WPA; do not stream in dead initializers.
2492 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2493 darf2out_register_external_decl): Support incremental link.
2495 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2497 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2499 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2501 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2502 it down to simple_object_copy_lto_debug_sections.
2503 (run_gcc): Determine incremental LTO link time and configure
2504 lto1 into non-wpa mode, disable renaming of debug sections.
2506 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2508 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2509 descriptions of various incorrectly documented functions.
2511 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2514 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2515 address check not strict.
2517 2018-05-30 Richard Biener <rguenther@suse.de>
2519 PR tree-optimization/85964
2520 * tracer.c (better_p): Drop initialized count check, we only
2521 call the function with initialized counts now.
2522 (find_best_successor): Do find a best edge if one
2523 has uninitialized count.
2524 (find_best_predecessor): Likewise. Do BB frequency check only
2525 if count is initialized.
2527 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2529 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2530 (aarch64_ldrstr_offset_compare): New.
2531 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2532 load/store orderings.
2533 (aarch64_gen_adjusted_ldpstp): Likewise.
2535 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2537 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2538 Check for subset of GENERAL_REGS and FP_REGS.
2539 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2542 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2544 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2545 and wi::to_poly_offset. Add the current offset and then check
2546 whether the sum fits, rather than using an unchecked addition of
2547 a checked term. Check for a shwi rather than a uhwi.
2548 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2549 (store_constructor): Use poly_int_tree_p.
2550 (expand_expr_real_1): Likewise.
2551 * function.c (assign_temp): Likewise.
2552 * fold-const.c (const_binop): Use poly_int_tree_p and
2554 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2556 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2557 to_poly_offset for MEM offsets.
2558 * ipa-icf.c (sem_variable::equals): Likewise.
2559 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2560 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2561 wi::to_poly_offset for BIT_FIELD_REF offsets.
2562 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2564 * var-tracking.c (emit_note_insn_var_location): Use
2565 tree_to_poly_uint64.
2567 2018-05-29 Jim Wilson <jimw@sifive.com>
2569 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2571 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2574 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2575 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2577 (sse4_1_round<mode>2): Use nonimmediate_operand
2578 for operand 1 predicate.
2580 2018-05-29 Martin Sebor <msebor@redhat.com>
2581 Richard Biener <rguenther@suse.de>
2584 * calls.c (get_size_range): Call determine_value_range instead
2585 of get_value_range..
2586 * tree-vrp.h (determine_value_range): Declared new function.
2587 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2589 2018-05-29 Richard Biener <rguenther@suse.de>
2591 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2592 sure to use non-pattern stmts for get_earlier_stmt arguments.
2593 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2594 called on pattern stmts.
2595 (get_later_stmt): Likewise.
2597 2018-05-29 Martin Liska <mliska@suse.cz>
2599 PR gcov-profile/85759
2600 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2603 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2605 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2607 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2611 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2612 VEC_PACK_FLOAT_EXPR): New tree codes.
2613 * tree-pretty-print.c (op_code_prio): Handle
2614 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2615 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2616 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2617 * tree-inline.c (estimate_operator_cost): Likewise.
2618 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2619 * fold-const.c (const_binop): Likewise.
2620 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2621 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2622 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2623 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2624 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2625 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2626 * expr.c (expand_expr_real_2): Likewise.
2627 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2628 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2629 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2631 * optabs.c (expand_widen_pattern_expr): For
2632 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2633 sign from result type rather than operand's type.
2634 (expand_binop_directly): For vec_packu_float_optab and
2635 vec_packs_float_optab allow result type to be different from operand's
2637 * optabs-tree.c (optab_for_tree_code): Handle
2638 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2639 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2640 * tree-vect-generic.c (expand_vector_operations_1): Handle
2641 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2642 VEC_PACK_FLOAT_EXPR.
2643 * tree-vect-stmts.c (supportable_widening_operation): Handle
2645 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2646 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2647 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2648 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2649 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2651 (vec_pack<floatprefix>_float_<mode>): New expander.
2652 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2654 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2655 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2656 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2657 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2658 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2660 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2661 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2662 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2663 VEC_PACK_FLOAT_EXPR): Document.
2665 2018-05-29 Richard Biener <rguenther@suse.de>
2667 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2669 (stmt_vec_info_vec): Make pointer.
2670 (init_stmt_vec_info_vec): Remove.
2671 (free_stmt_vec_info_vec): Likewise.
2672 (set_stmt_vec_info_vec): New function.
2673 (free_stmt_vec_infos): Likewise.
2674 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2675 (set_vinfo_for_stmt): Likewise.
2676 (get_earlier_stmt): Likewise.
2677 (get_later_stmt): Likewise.
2678 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2679 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2680 (vec_info::~vec_info): Free stmt_vec_infos.
2681 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2682 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2683 (pass_slp_vectorize::execute): Likewise.
2684 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2685 (free_stmt_vec_info_vec): Likewise.
2686 (set_stmt_vec_info_vec): New function.
2687 (free_stmt_vec_infos): Likewise.
2688 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2689 the global stmt_vec_info_vec.
2690 * tree-parloops.c (gather_scalar_reductions): Use
2691 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2694 2018-05-29 Richard Biener <rguenther@suse.de>
2696 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2698 2018-05-29 Martin Liska <mliska@suse.cz>
2699 David Malcolm <dmalcolm@redhat.com>
2701 * vec.c (test_reverse): New.
2702 (vec_c_tests): Add new test.
2703 * vec.h (vl_ptr>::reverse): New function.
2705 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2707 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2709 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2712 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2714 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2716 2018-05-28 Richard Biener <rguenther@suse.de>
2718 PR tree-optimization/85933
2719 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2720 look at stmts marked as vectorizable.
2722 2018-05-28 Richard Biener <rguenther@suse.de>
2724 PR tree-optimization/85934
2725 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2726 vector boolean check before scalar optimization.
2728 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2730 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2733 2018-05-28 Mark Wielaard <mark@klomp.org>
2735 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2736 if it is an expression containing a minus sign.
2738 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2740 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2742 2018-05-27 Paul Koning <ni1d@arrl.net>
2744 * config/pdp11/pdp11.md (truncsihi2): Remove.
2746 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2747 Chung-Ju Wu <jasonwucj@gmail.com>
2749 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2751 (unaligned_store_dw): Ditto.
2752 * config/nds32/nds32-memory-manipulation.c
2753 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2754 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2755 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2756 (emit_setmem_word_loop): Rename to ...
2757 (emit_setmem_doubleword_loop): ... this.
2758 (nds32_gen_dup_4_byte_to_word_value): New function.
2759 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2760 (nds32_expand_setmem_loop): Refine implementation.
2761 (nds32_expand_setmem_loop_v3m): Ditto.
2762 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2765 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2767 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2769 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2771 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2772 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2773 machine->attr_no_prologue_p.
2774 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2775 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2776 (nds32_expand_epilogue): Consider attr_naked_p.
2777 (nds32_expand_epilogue_v3pop): Likewise.
2778 (nds32_can_use_return_insn): Likewise.
2779 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2780 attr_no_prologue_p fields.
2781 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2783 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2786 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2788 * config/i386/sse.md
2789 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2791 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2793 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2795 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2797 (*<floatsuffix>floatv2div2sf2): Rename to ...
2798 (*float<floatunssuffix>v2div2sf2): ... this.
2799 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2800 (float<floatunssuffix>v2div2sf2_mask): ... this.
2801 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2802 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2803 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2805 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2807 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2809 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2811 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2813 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2815 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2816 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2817 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2818 gen_ufix_truncv8dfv8si2.
2819 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2820 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2821 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2822 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2823 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2824 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2825 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2826 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2828 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2832 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2834 2018-05-25 Jim Wilson <jimw@sifive.com>
2836 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2837 * config/riscv/riscv.c (struct machine_function): Add
2838 interrupt_handler_p and attribute_checked_p fields.
2839 (riscv_attribute_table): Add interrupt.
2840 (riscv_interrupt_type_p): New.
2841 (riscv_save_reg_p): Save extra regs for interrupt handler.
2842 (riscv_use_save_libcall): Return false for interrupt handler.
2843 (riscv_first_stack_step): Add forward declaration.
2844 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2845 for interrupt handler with large frame. Use it for saved reg list.
2846 (riscv_expand_prologue): Move flag_stack_usage_info support to
2847 eliminate duplication.
2848 (riscv_expand_epilogue): Generate mret for interrupt handler.
2849 (riscv_epilogue_uses): New.
2850 (riscv_can_use_return_insn): Return false for interrupt handler.
2851 (riscv_function_ok_for_sibcall): Likewise.
2852 (riscv_set_current_function): Add interrupt handler support.
2853 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2854 * config/riscv/riscv.md (UNSPECV_MRET): New.
2856 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2858 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2860 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2862 PR tree-optimization/85712
2863 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2864 this candidate has already been replaced in-situ by a copy.
2866 2018-05-25 Jason Merrill <jason@redhat.com>
2868 PR c++/80485 - inline function non-zero address.
2869 * symtab.c (nonzero_address): Check DECL_COMDAT.
2871 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2874 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2875 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2876 extension of SImode operation. Use const123_operand predicate.
2877 (*saddsi_1): Remove.
2878 (*saddl_se_1): Ditto.
2880 (*ssubl_se_1): Ditto.
2881 * config/alpha/predicates.md (const123_operand): New predicate.
2882 * config/alpha/constraints.md (P): Use IN_RANGE.
2884 2018-05-25 Richard Biener <rguenther@suse.de>
2886 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2888 (ref_maybe_used_by_stmt_p): Likewise.
2889 (stmt_may_clobber_ref_p): Likewise.
2890 (stmt_may_clobber_ref_p_1): Likewise.
2891 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2893 (ref_maybe_used_by_stmt_p): Likewise.
2894 (stmt_may_clobber_ref_p): Likewise.
2895 (stmt_may_clobber_ref_p_1): Likewise.
2896 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2897 the alias oracle to disambiguate DRs with stmts DR analysis
2899 (vect_analyze_data_refs): Do not give up on not analyzable
2900 DRs for BB vectorization. Remove code truncating the dataref
2903 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2906 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2907 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2908 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2910 2018-05-25 Richard Biener <rguenther@suse.de>
2912 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2913 function, combining stmt data ref gathering and fatal analysis
2915 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2916 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2917 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2918 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2919 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2921 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2923 PR tree-optimization/85720
2924 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2925 SCC if all partitions are builtins.
2926 (version_loop_by_alias_check): New parameter. Generate cancelable
2927 runtime alias check if all partitions are builtins.
2928 (distribute_loop): Update call to above function.
2930 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2932 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2933 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2934 (parm_default_def_partition_arg): Ditto.
2935 (set_parm_default_def_partition): Ditto.
2936 (get_parm_default_def_partitions): Ditto and make it static.
2937 (get_undefined_value_partitions): Ditto and make it static.
2938 (remove_ssa_form): Refactor call to init_var_map here.
2939 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2940 computation for loop region.
2941 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2942 (register_default_def): Delete.
2943 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2944 (parm_default_def_partition_arg): Ditto.
2945 (set_parm_default_def_partition): Ditto.
2946 (get_parm_default_def_partitions): Ditto and make it static.
2947 (get_undefined_value_partitions): Ditto and make it static.
2948 (coalesce_with_default, coalesce_with_default): Update comment.
2949 (create_coalesce_list_for_region): New func factored out from
2950 create_outofssa_var_map.
2951 (populate_coalesce_list_for_outofssa): New func factored out from
2952 create_outofssa_var_map and coalesce_ssa_name.
2953 (create_outofssa_var_map): Delete.
2954 (coalesce_ssa_name): Refactor to support live range computation.
2955 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2956 (get_parm_default_def_partitions): Delete.
2957 (get_undefined_value_partitions): Ditto.
2958 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2959 computation for loop region.
2960 (new_tree_live_info, loe_visit_block): Ditto.
2961 (live_worklist, set_var_live_on_entry): Ditto.
2962 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2963 * tree-ssa-live.h (struct _var_map): New fields.
2964 (init_var_map): Change decl.
2965 (region_contains_p): New.
2967 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2969 * tree-ssa-live.h (live_merge_and_clear): Delete.
2971 2018-05-25 Richard Biener <rguenther@suse.de>
2974 * tree-dump.c (dequeue_and_dump): Remove access to removed
2975 operand 2 of a SWITCH_EXPR.
2977 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2979 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2980 multiplication and division.
2981 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2982 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2983 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2984 (cond_udiv_optab, cond_umod_optab): New optabs.
2985 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2986 (IFN_COND_RDIV): New internal functions.
2987 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2988 TRUNC_MOD_EXPR and RDIV_EXPR.
2989 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2990 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2992 (SVE_INT_BINARY): Include mult.
2993 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2994 (optab, sve_int_op): Handle mult.
2995 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2997 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2998 for SVE_INT_BINARY_SD.
3000 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3002 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3003 (optab, sve_int_op): Handle div and udiv.
3004 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3005 for SVE_INT_BINARY_SD.
3006 (*<optab><mode>3): New insn for the same.
3008 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3010 * tree-vect-patterns.c: Include predict.h.
3011 (vect_recog_divmod_pattern): Restrict check for division support
3012 to when optimizing for size.
3014 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3016 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3017 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3018 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3019 (gimple_match_op::set_op): Likewise.
3020 (gimple_resimplify4): Declare.
3021 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3022 (expr::gen_transform): Likewise.
3023 (decision_tree::gen): Generate a simplification routine for 4 operands.
3024 * gimple-match-head.c (gimple_simplify): Add an overload for
3025 4 operands. In the top-level function, handle up to 4 call
3026 arguments and call gimple_resimplify4.
3027 (gimple_resimplify4): New function.
3028 (build_call_internal): Pass a fourth operand.
3029 (maybe_push_to_seq): Likewise.
3030 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3031 Fold VEC_COND_EXPRs of an operation and a default value into
3032 an IFN_COND_* function if possible.
3033 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3035 (SVE_COND_FP_BINARY): Include them.
3036 (optab, sve_fp_op): Handle them.
3037 (SVE_INT_BINARY_REV): New code iterator.
3038 (SVE_COND_FP_BINARY_REV): New int iterator.
3039 (commutative): New int attribute.
3040 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3042 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3044 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3045 (*cond_<optab><mode>): New patterns for reversed operands.
3047 2018-05-25 Richard Biener <rguenther@suse.de>
3049 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3050 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3051 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3052 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3053 * tree-vect-data-refs.c (everywhere): Adjust users.
3054 * tree-vect-loop.c (everywhere): Likewise.
3055 * tree-vect-slp.c (everywhere): Likewise.
3056 * tree-vect-stmts.c (everywhere): Likewise.
3057 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3059 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3061 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3063 (gcc_cv_as_section_exclude): ... this.
3064 Try Solaris as #exclude syntax.
3065 * configure: Regenerate.
3066 * config.in: Regenerate.
3067 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3069 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3070 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3072 * varasm.c (default_elf_asm_named_section): Don't check if
3073 HAVE_GAS_SECTION_EXCLUDE is defined.
3075 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3077 * doc/md.texi: Update the documentation of the cond_* optabs
3078 to mention the new final operand. Fix GET_MODE_NUNITS call.
3079 Describe the scalar case too.
3080 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3081 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3083 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3084 (get_conditional_internal_fn): Update comment.
3085 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3086 accumulator value as a final argument to conditional functions.
3087 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3088 a define_expand and add an "else" operand. Assert for now that
3089 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3090 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3091 (*cond_<optab><mode>): New patterns.
3092 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3093 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3094 (UNSPEC_COND_EOR): Delete.
3095 (optab): Remove associated mappings.
3096 (SVE_INT_BINARY): New code iterator.
3097 (sve_int_op): Remove int attribute and add "minus" to the code
3099 (SVE_COND_INT_OP): Delete.
3100 (SVE_COND_FP_OP): Rename to...
3101 (SVE_COND_FP_BINARY): ...this.
3103 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3105 * optabs.c (can_reuse_operands_p): New function.
3106 (maybe_legitimize_operands): Try to reuse the results for
3109 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3111 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3112 Add {q} suffix to insn mnemonic.
3114 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3116 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3117 (msp430_warn_func_return): New.
3119 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3121 * fold-const.c (tree_nonzero_bits): New function.
3122 * fold-const.h (tree_nonzero_bits): Likewise.
3123 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3124 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3126 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3130 * varasm.c (assemble_alias): Check ifunc_resolver only on
3133 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3136 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3137 when memory input operand is handled.
3139 2018-05-24 Luis Machado <luis.machado@linaro.org>
3141 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3143 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3145 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3147 * match.pd: Delay FMA folds until after vectorization.
3149 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3152 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3153 address check not strict.
3155 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3157 * gimple-match.h (gimple_match_op): New class.
3158 (mprts_hook): Replace parameters with a gimple_match_op *.
3159 (maybe_build_generic_op): Likewise.
3160 (gimple_simplified_result_is_gimple_val): Replace parameters with
3161 a const gimple_match_op *.
3162 (gimple_simplify): Replace code_helper * and tree * parameters with
3163 a gimple_match_op * parameter.
3164 (gimple_resimplify1): Replace code_helper *, tree and tree *
3165 parameters with a gimple_match_op * parameter.
3166 (gimple_resimplify2): Likewise.
3167 (gimple_resimplify3): Likewise.
3168 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3169 parameters with a gimple_match_op * parameter.
3170 * gimple-match-head.c (gimple_simplify): Change prototypes of
3171 auto-generated functions to take a gimple_match_op * instead of
3172 separate code_helper * and tree * parameters. Make the same
3173 change in the top-level overload and update calls to the
3174 gimple_resimplify routines. Update calls to the auto-generated
3175 functions and to maybe_push_res_to_seq in the publicly-facing
3176 operation-specific gimple_simplify overloads.
3177 (gimple_match_op::MAX_NUM_OPS): Define.
3178 (gimple_resimplify1): Replace rcode and ops with a single res_op
3179 parameter. Update call to gimple_simplify.
3180 (gimple_resimplify2): Likewise.
3181 (gimple_resimplify3): Likewise.
3182 (mprts_hook): Replace parameters with a gimple_match_op *.
3183 (maybe_build_generic_op): Likewise.
3184 (build_call_internal): Replace type, nargs and ops with
3185 a gimple_match_op *.
3186 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3187 with a single gimple_match_op *. Update calls to mprts_hook,
3188 build_call_internal and gimple_simplified_result_is_gimple_val.
3189 Factor out code that is common to the tree_code and combined_fn cases.
3190 * genmatch.c (expr::gen_transform): Replace tem_code and
3191 tem_ops with a gimple_match_op called tem_op. Update calls
3192 to the gimple_resimplify functions and maybe_push_res_to_seq.
3193 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3194 res_ops. Update call to the gimple_resimplify functions.
3195 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3196 (decision_tree::gen): Make the functions take a gimple_match_op *
3197 called res_op instead of separate res_code and res_ops parameters.
3198 Update call accordingly.
3199 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3200 and ops with a single res_op parameter. Update calls to
3201 maybe_build_generic_op and maybe_push_res_to_seq.
3202 (fold_stmt_1): Update calls to gimple_simplify and
3203 replace_stmt_with_simplification.
3204 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3205 and gimple_simplified_result_is_gimple_val.
3206 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3208 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3209 with a gimple_match_op *.
3210 (vn_nary_build_or_lookup): Likewise. Update call to
3211 vn_nary_build_or_lookup_1.
3212 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3213 gimple_match_op *. Update calls to the gimple_resimplify routines
3214 and to gimple_simplified_result_is_gimple_val.
3215 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3216 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3217 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3218 (visit_nary_op): Likewise.
3219 (visit_reference_op_load): Likewise.
3221 2018-05-23 Luis Machado <luis.machado@linaro.org>
3223 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3224 modifier for printing the step amount.
3226 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3229 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3232 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3234 * doc/sourcebuild.texi (Endianness): New subsubsection.
3236 2018-05-23 Luis Machado <luis.machado@linaro.org>
3238 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3239 <prefetch_dynamic_strides>: New const bool field.
3240 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3241 prefetch_dynamic_strides.
3242 (exynosm1_prefetch_tune): Likewise.
3243 (thunderxt88_prefetch_tune): Likewise.
3244 (thunderx_prefetch_tune): Likewise.
3245 (thunderx2t99_prefetch_tune): Likewise.
3246 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3248 (aarch64_override_options_internal): Update to set
3249 PARAM_PREFETCH_DYNAMIC_STRIDES.
3250 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3251 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3252 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3253 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3254 prefetch-dynamic-strides setting.
3256 2018-05-23 Luis Machado <luis.machado@linaro.org>
3258 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3259 <minimum_stride>: New const int field.
3260 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3261 minimum_stride field defaulting to -1.
3262 (exynosm1_prefetch_tune): Likewise.
3263 (thunderxt88_prefetch_tune): Likewise.
3264 (thunderx_prefetch_tune): Likewise.
3265 (thunderx2t99_prefetch_tune): Likewise.
3266 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3267 <default_opt_level>: Set to 3.
3268 (aarch64_override_options_internal): Update to set
3269 PARAM_PREFETCH_MINIMUM_STRIDE.
3270 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3271 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3272 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3273 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3274 stride is constant and is below the minimum stride threshold.
3276 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3278 * config/arm/arm-cpus.in (mode26): Delete.
3279 (armv4): Delete mode26 reference.
3280 * config/arm/arm.c (arm_configure_build_target): Delete use of
3283 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3285 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3287 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3288 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3290 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3291 (floatunsdidf2): Ditto.
3293 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3295 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3296 (fixuns_trunc<mode>si2_avx512f): Ditto.
3297 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3298 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3299 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3301 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3303 PR rtl-optimization/79985
3304 * df-scan.c (df_insn_refs_collect): Remove special case for
3305 global registers and asm statements.
3307 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3309 * extend.texi (Global Register Variables): Rewrite the bullet list.
3310 Note that the register is available for allocation. Note that access
3311 via inline asm must use constraints. Add note about async-signal
3312 handlers. Remove paragraph about automagic register selection.
3314 2018-05-23 Richard Biener <rguenther@suse.de>
3316 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3317 of fixed offset from memset VN.
3319 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3321 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3323 (alloc_cand_and_find_basis): Initialize first_interp field.
3324 (slsr_process_mul): Modify first_interp field.
3325 (slsr_process_add): Likewise.
3326 (slsr_process_cast): Modify first_interp field for each new
3328 (slsr_process_copy): Likewise.
3329 (dump_candidate): Dump first_interp field.
3330 (replace_mult_candidate): Process all interpretations, not just
3332 (replace_rhs_if_not_dup): Likewise.
3333 (replace_one_candidate): Likewise.
3335 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3337 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3339 (aarch64_needs_frame_chain): New function.
3340 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3342 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3345 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3346 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3347 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3348 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3349 as true for strict-align.
3350 (aarch64_can_inline_p): Perform checks even when callee has no
3351 attributes to check for strict alignment.
3352 * doc/extend.texi (AArch64 Function Attributes): Document
3354 * doc/invoke.texi: (AArch64 Options): Likewise.
3356 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3358 PR tree-optimization/85853
3359 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3360 the handling of the root of the node to...
3361 (vect_slp_analyze_node_operations_1): ...this new function,
3362 and run the whole thing with the child nodes' def types
3363 set according to their SLP node's def type.
3365 2018-05-23 Richard Biener <rguenther@suse.de>
3368 * tree-data-ref.c (create_runtime_alias_checks): Defer
3369 and ignore overflow warnings.
3371 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3373 PR tree-optimization/85822
3374 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3377 2018-05-23 Richard Biener <rguenther@suse.de>
3379 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3380 memset constants via native_interpret_expr.
3382 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3385 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3387 (cgraph_node::create_alias): Likewise.
3388 (cgraph_node::get_availability): Check ifunc_resolver instead
3389 of looking up ifunc attribute.
3390 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3391 * varasm.c (do_assemble_alias): Likewise.
3392 (assemble_alias): Likewise.
3393 (default_binds_local_p_3): Likewise.
3394 * cgraph.h (cgraph_node): Add ifunc_resolver.
3395 (cgraph_node::only_called_directly_or_aliased_p): Return false
3397 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3399 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3400 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3401 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3402 instead of looking up ifunc attribute.
3404 2018-05-22 Luis Machado <luis.machado@linaro.org>
3406 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3408 2018-05-22 Martin Sebor <msebor@redhat.com>
3411 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3412 only when expasion succeeds.
3413 (expand_builtin_strcmp): Same.
3414 (expand_builtin_strncmp): Same.
3416 2018-05-22 Martin Sebor <msebor@redhat.com>
3418 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3420 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3421 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3423 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3424 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3426 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3427 Add check for legitimate_address.
3428 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3429 (aarch64_swap_ldrstr_operands): New.
3430 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3433 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3434 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3436 * config/aarch64/aarch64.md: New patterns to generate stp
3438 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3439 single words and double words.
3440 (load_pair_sw, load_pair_dw): Likewise.
3441 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3443 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3445 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3446 for different mode ldpstp and add peephole for merged zero stores.
3448 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3450 (aarch64_gen_store_pair): Rename calls to match new patterns.
3451 (aarch64_gen_load_pair): Rename calls to match new patterns.
3452 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3453 (load_pair<DREG:mode><DREG2:mode>): ... This.
3454 (store_pair<mode>): Rename to...
3455 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3456 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3458 (V_INT_EQUIV): Handle SImode.
3459 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3462 2018-05-22 Martin Sebor <msebor@redhat.com>
3465 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3466 or ajust the presumed bound on an operation to avoid unnecessary
3469 2018-05-22 Martin Sebor <msebor@redhat.com>
3471 PR tree-optimization/85826
3472 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3473 assuming that a DECL necesarily has a constant size.
3475 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3478 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3480 2018-05-22 Richard Biener <rguenther@suse.de>
3482 PR tree-optimization/85834
3483 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3484 non-constant and non-zero memset arguments.
3486 2018-05-22 Martin Liska <mliska@suse.cz>
3489 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3491 2018-05-22 Richard Biener <rguenther@suse.de>
3493 PR tree-optimization/85863
3494 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3495 comparisons when vectype is specified.
3496 (vectorizable_condition): Do not specify vectype for
3497 vect_is_simple_cond when SLP vectorizing.
3499 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3502 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3503 define __ibm128 as long double.
3504 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3505 as a distinct type when IEEE 128-bit support is enabled.
3506 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3507 128-bit types to use the correct functions.
3508 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3509 convert between 128-bit floating point types that have different
3510 modes but the same representation, instead of using gen_lowpart to
3512 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3514 (IFKF_reg): New attributes to give the register constraints for
3516 (extend<mode>tf2_internal): New insns to mark an explicit
3517 conversion between 128-bit floating point types that have a
3518 different mode but share the same representation.
3520 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3522 PR tree-optimization/85814
3523 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3524 a null return from get_strinfo when unsharing the next
3525 strinfo in the chain.
3527 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3530 * varasm.c (weak_finish): Clean up weak_decls.
3532 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3534 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3535 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3536 UNSPEC_UADALP values.
3537 * config/aarch64/iterators.md (ABAL): New int iterator.
3540 (sur): Add mappings for the above.
3541 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3543 (aarch64_<sur>abal<mode>_4): Likewise.
3544 (aarch64_<sur>adalp<mode>_3): Likewise.
3545 (<sur>sadv16qi): New define_expand.
3547 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3549 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3550 (*movdf_internal): Ditto.
3551 (*rcpsf2_sse): Ditto.
3552 (*rsqrtsf2_sse): Ditto.
3553 (*sqrt<mode>2_sse): Ditto.
3555 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3557 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3559 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3560 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3561 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3562 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3564 * config/aarch64/arm_neon.h: Likewise.
3565 * config/aarch64/iterators.md (VQ_I): New.
3567 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3569 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3571 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3574 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3576 * config/nds32/constraints.md (S): New constraint.
3577 * config/nds32/nds32.md (call_internal): Use constraint S.
3578 (call_value_internal): Likewise.
3579 (sibcall_internal): Likewise.
3580 (sibcall_value_internal): Likewise.
3582 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3583 Chung-Ju Wu <jasonwucj@gmail.com>
3585 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3588 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3589 Chung-Ju Wu <jasonwucj@gmail.com>
3591 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3592 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3593 (nds32_rtx_costs_impl): Simplify.
3594 (nds32_address_cost_impl): Simplify.
3595 (nds32_init_rtx_costs): New function.
3596 (nds32_rtx_costs_speed_prefer): Likewise.
3597 (nds32_rtx_costs_size_prefer): Likewise.
3598 (nds32_address_cost_speed_prefer): Likewise.
3599 (nds32_address_cost_speed_fwprop): Likewise.
3600 (nds32_address_cost_size_prefer): Likewise.
3601 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3602 * config/nds32/nds32.c (nds32_option_override): Use
3603 nds32_init_rtx_costs function.
3605 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3607 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3608 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3609 (TARGET_PIPELINE_N8): Likewise.
3610 (TARGET_PIPELINE_N10): Likewise.
3611 (TARGET_PIPELINE_N13): Likewise.
3612 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3614 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3616 * config/nds32/nds32-fpu.md: Update copyright year.
3618 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3620 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3622 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3624 * config/nds32/nds32.c
3625 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3626 * config/nds32/nds32.opt (minline-asm-r15): New option.
3628 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3630 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3632 * config/nds32/nds32.md (abssi2): New pattern.
3634 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3636 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3637 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3638 Merge insn pattern from sse_cvtsi2ss<round_name> and
3639 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3640 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3641 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3642 using SWI48 mode iterator.
3643 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3644 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3645 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3646 pattern from sse_cvttss2si<round_saeonly_name>
3647 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3648 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3649 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3650 using SWI48 mode iterator.
3651 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3652 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3653 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3654 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3655 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3656 using SWI48 mode iterator.
3657 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3658 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3659 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3660 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3661 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3662 SWI48 mode iterator.
3663 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3664 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3665 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3666 pattern from sse_cvttsd2si<round_saeonly_name>
3667 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3669 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3671 * config/nds32/nds32-md-auxiliary.c
3672 (nds32_valid_smw_lwm_base_p): Refine.
3673 (nds32_output_smw_single_word): Refine.
3674 (nds32_output_smw_double_word): New.
3675 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3677 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3679 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3680 (nds32_output_stack_pop): Refine.
3681 (nds32_expand_unaligned_load): Refine.
3682 (nds32_expand_unaligned_store): Refine.
3684 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3685 Chung-Ju Wu <jasonwucj@gmail.com>
3687 * config/nds32/constants.md: Add TP_REGNUM constant.
3688 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3689 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3691 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3692 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3693 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3694 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3695 and PIC code generation.
3696 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3698 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3700 * config/nds32/nds32.md: Support TLS and PIC.
3701 * config/nds32/nds32.c: Support TLS and PIC.
3702 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3703 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3706 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3708 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3709 mode with E_ prefix.
3711 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3712 Chung-Ju Wu <jasonwucj@gmail.com>
3714 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3715 * config/nds32/nds32-md-auxiliary.c
3716 (symbolic_reference_mentioned_p): New.
3717 (nds32_legitimize_ict_address): New.
3718 (nds32_expand_ict_move): New.
3719 (nds32_indirect_call_referenced_p): New.
3720 (nds32_symbol_binds_local_p): Delete.
3721 (nds32_long_call_p): Modify.
3722 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3723 * config/nds32/nds32-protos.h
3724 (symbolic_reference_mentioned_p): Declare.
3725 (nds32_legitimize_ict_address): Declare.
3726 (nds32_expand_ict_move): Declare.
3727 (nds32_indirect_call_referenced_p): Declare.
3728 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3729 (nds32_relax_group): Use nds32_ict_const_p as condition.
3730 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3731 (nds32_asm_file_start): Output ict_model directive in asm code.
3732 (nds32_legitimate_address_p): Consider indirect call.
3733 (nds32_print_operand): Consider indirect call.
3734 (nds32_print_operand_address): Consider indirect call.
3735 (nds32_insert_attributes): Handle "indirect_call" attribute.
3736 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3737 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3738 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3739 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3740 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3741 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3742 (TARGET_ICT_MODEL_SMALL): Define.
3743 (TARGET_ICT_MODEL_LARGE): Define.
3744 * config/nds32/nds32.md (movsi): Consider ict model.
3745 (call, call_value): Consider ict model.
3746 (sibcall, sibcall_value): Consider ict model.
3747 * config/nds32/nds32.opt (mict-model): New option.
3748 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3751 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3752 Monk Chiang <sh.chiang04@gmail.com>
3753 Jim Wilson <jimw@sifive.com>
3755 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3756 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3757 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3758 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3759 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3760 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3761 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3762 compute save_libcall_adjustment properly.
3763 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3764 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3765 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3766 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3767 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3768 (ABI_SPEC): Handle mabi=ilp32e.
3769 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3770 (RVE): Add RVE mask.
3771 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3772 <-march>: Add rv32e as an example.
3774 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3777 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3778 (intra_create_variable_infos): Handle C++ constructors.
3780 2018-05-18 Martin Liska <mliska@suse.cz>
3782 * passes.def: Remove a redundant pass.
3784 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3787 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3789 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3791 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3793 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3795 (armv2, armv2a, armv3, armv3m): Delete architectures.
3796 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3797 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3798 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3800 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3801 (*mulsidi3adddi): Likewise.
3802 (mulsidi3): Likewise.
3803 (*mulsidi3_nov6): Likewise.
3804 (umulsidi3): Likewise.
3805 (umulsidi3_nov6): Likewise.
3806 (umaddsidi4): Likewise.
3807 (*umulsidi3adddi): Likewise.
3808 (smulsi3_highpart): Likewise.
3809 (*smulsi3_highpart_nov6): Likewise.
3810 (umulsi3_highpart): Likewise.
3811 (*umulsi3_highpart_nov6): Likewise.
3812 * config/arm/arm.h (arm_arch3m): Delete.
3813 * config/arm/arm.c (arm_arch3m): Delete.
3814 (arm_option_override_internal): Update armv3-related comment.
3815 (arm_configure_build_target): Delete use of isa_bit_mode32.
3816 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3817 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3818 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3820 (mulusa3): Likewise.
3821 * config/arm/arm-protos.h (arm_arch3m): Delete.
3822 * config/arm/arm-tables.opt: Regenerate.
3823 * config/arm/arm-tune.md: Likewise.
3824 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3825 deleted architectures.
3827 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3829 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3830 (armv5t, armv5te): New features.
3831 (ARMv5, ARMv5e): Delete fgroups.
3832 (ARMv5t, ARMv5te): Adjust for above changes.
3834 (armv5, armv5e): Delete arches.
3835 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3837 (*call_reg_arm): Likewise.
3838 (*call_value_reg_armv5): Likewise.
3839 (*call_value_reg_arm): Likewise.
3840 (*call_symbol): Likewise.
3841 (*call_value_symbol): Likewise.
3842 (*sibcall_insn): Likewise.
3843 (*sibcall_value_insn): Likewise.
3845 (prefetch): Likewise.
3846 (define_split and define_peephole2 dependent on arm_arch5):
3848 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3850 (TARGET_ARM_QBIT): Likewise.
3851 (TARGET_DSP_MULTIPLY): Likewise.
3852 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3853 (arm_arch5, arm_arch5e): Delete.
3854 (arm_arch5t, arm_arch5te): Declare.
3855 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3856 (arm_arch5t): Declare.
3857 (arm_option_reconfigure_globals): Update for the above.
3858 (arm_options_perform_arch_sanity_checks): Update comment, replace
3859 use of arm_arch5 with arm_arch5t.
3860 (use_return_insn): Likewise.
3861 (arm_emit_call_insn): Likewise.
3862 (output_return_instruction): Likewise.
3863 (arm_final_prescan_insn): Likewise.
3864 (arm_coproc_builtin_available): Likewise.
3865 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3866 arm_arch5e with arm_arch5t and arm_arch5te.
3867 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3868 (arm_arch5t, arm_arch5te): Declare.
3869 * config/arm/arm-tables.opt: Regenerate.
3870 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3871 * config/arm/t-multilib: Likewise.
3872 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3873 instead of arm_arch5.
3874 (*call_reg_thumb1): Likewise.
3875 (*call_value_reg_thumb1_v5): Likewise.
3876 (*call_value_reg_thumb1): Likewise.
3877 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3879 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3881 2018-05-18 Martin Liska <mliska@suse.cz>
3883 PR gcov-profile/84846
3884 * doc/gcov.texi: Document -t option of gcov tool.
3886 2018-05-18 Martin Liska <mliska@suse.cz>
3888 PR gcov-profile/84846
3889 * gcov.c (print_usage): Add new -t option.
3890 (process_args): Handle the option.
3891 (generate_results): Use stdout as output when requested by
3894 2018-05-18 Martin Liska <mliska@suse.cz>
3896 PR gcov-profile/84846
3897 * coverage.c (coverage_init): Write PWD to .gcno file.
3898 * doc/gcov.texi: Document how working directory is printed.
3899 * gcov-dump.c (dump_gcov_file): Print PWD.
3900 * gcov.c (output_intermediate_file): Likewise.
3901 (read_graph_file): Read PWD string.
3902 (output_lines): Print PWD.
3904 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3907 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3908 for retval and return false if all args to phi are zero.
3910 2018-05-18 Richard Biener <rguenther@suse.de>
3912 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3914 (evrp_dom_walker::before_dom_children): Call it.
3916 2018-05-18 Richard Biener <rguenther@suse.de>
3918 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3919 results when processing array refs with variable index.
3921 2018-05-18 Toon Moene <toon@moene.org>
3923 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3924 directly after that of -floop-interchange. Indicate that both
3925 options are enabled by default when specifying -O3.
3927 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3929 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3930 iterator. Delete separate integer-mode vec_set<mode> expander.
3931 (aarch64_simd_vec_setv2di): Delete.
3932 (vec_setv2di): Delete.
3933 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3934 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3935 the "w, r" alternative.
3937 2018-05-18 Martin Liska <mliska@suse.cz>
3939 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3940 * tree-pass.h (make_pass_lower_switch_O0): New function.
3941 * tree-switch-conversion.c (node_has_low_bound): Remove.
3942 (node_has_high_bound): Likewise.
3943 (node_is_bounded): Likewise.
3944 (class pass_lower_switch): Make it a template type and create
3946 (pass_lower_switch::execute): Add template argument.
3947 (make_pass_lower_switch): New function.
3948 (make_pass_lower_switch_O0): New function.
3949 (do_jump_if_equal): Remove.
3950 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3951 all the hard work to tree optimization passes.
3953 2018-05-18 Martin Liska <mliska@suse.cz>
3955 * dbgcnt.c (limit_low): Renamed from limit.
3956 (limit_high): New variable.
3957 (dbg_cnt_is_enabled): Check for upper limit.
3958 (dbg_cnt): Adjust dumping.
3959 (dbg_cnt_set_limit_by_index): Add new argument for high
3961 (dbg_cnt_set_limit_by_name): Likewise.
3962 (dbg_cnt_process_single_pair): Parse new format.
3963 (dbg_cnt_process_opt): Use strtok.
3964 (dbg_cnt_list_all_counters): Remove 'value' and add
3966 * doc/invoke.texi: Document changes.
3968 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3970 * doc/sourcebuild.texi (scalar_all_fma): Document.
3971 * tree.def (FMA_EXPR): Delete.
3972 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3973 * internal-fn.c (ternary_direct): New macro.
3974 (expand_ternary_optab_fn): Likewise.
3975 (direct_ternary_optab_supported_p): Likewise.
3976 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3977 * builtins.c (fold_builtin_fma): Delete.
3978 (fold_builtin_3): Don't call it.
3979 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3980 * expr.c (expand_expr_real_2): Likewise.
3981 * fold-const.c (operand_equal_p): Likewise.
3982 (fold_ternary_loc): Likewise.
3983 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3984 * gimple.c (DEFTREECODE): Likewise.
3985 * gimplify.c (gimplify_expr): Likewise.
3986 * optabs-tree.c (optab_for_tree_code): Likewise.
3987 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3988 * tree-eh.c (operation_could_trap_p): Likewise.
3989 (stmt_could_throw_1_p): Likewise.
3990 * tree-inline.c (estimate_operator_cost): Likewise.
3991 * tree-pretty-print.c (dump_generic_node): Likewise.
3992 (op_code_prio): Likewise.
3993 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3994 * tree-ssa-operands.c (get_expr_operands): Likewise.
3995 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3996 * fold-const-call.h (fold_fma): Delete.
3997 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3998 CFN_FNMA and CFN_FNMS.
4000 * genmatch.c (combined_fn): New enum.
4001 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4002 (commutative_op): New function.
4003 (commutate): Use it. Handle more than 2 operands.
4004 (dt_operand::gen_gimple_expr): Use commutative_op.
4005 (parser::parse_expr): Allow :c to be used with non-binary
4006 operators if the commutative operand is known.
4007 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4008 CFN_FMS, CFN_FNMA and CFN_FNMS.
4009 (backprop::process_assign_use): Remove FMA_EXPR handling.
4010 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4011 (gen_hsa_fma): New function.
4012 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4013 IFN_FNMA and IFN_FNMS.
4014 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4015 * gimple-fold.h (follow_all_ssa_edges): Declare.
4016 * gimple-fold.c (follow_all_ssa_edges): New function.
4017 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4018 gimple_build interface and use follow_all_ssa_edges to fold the result.
4019 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4020 instead of checking for optabs directly.
4021 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4022 rather than FMA_EXPRs.
4023 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4024 call to IFN_FMA instead of an FMA_EXPR.
4026 2018-05-17 Jim Wilson <jimw@sifive.com>
4028 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4029 sign extended promoted subreg, and the range does not have the sign bit
4030 set, then do a sign extend.
4032 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4033 test, check for sign extended subreg and/or constant operands, and
4034 do a sign extend in that case.
4036 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4038 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4039 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4041 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4042 Change logics_shift_reg to logics_shift_imm.
4043 (thunderx2t99_fp_loadpair_basic): Delete.
4044 (thunderx2t99_fp_storepair_basic): Delete.
4045 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4046 (thunderx2t99_asimd_polynomial): Delete.
4047 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4048 and neon_fp_mul_d_scalar_q.
4049 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4050 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4051 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4052 (thunderx2t99_asimd_lut): Add missing tbl types.
4053 (thunderx2t99_asimd_ext): Delete.
4054 (thunderx2t99_asimd_load1_1_mult): Delete.
4055 (thunderx2t99_asimd_load1_2_mult): Delete.
4056 (thunderx2t99_asimd_load1_ldp): New.
4057 (thunderx2t99_asimd_load1): New.
4058 (thunderx2t99_asimd_load2): Add missing *load2* types.
4059 (thunderx2t99_asimd_load3): New.
4060 (thunderx2t99_asimd_load4): New.
4061 (thunderx2t99_asimd_store1_1_mult): Delete.
4062 (thunderx2t99_asimd_store1_2_mult): Delete.
4063 (thunderx2t99_asimd_store2_mult): Delete.
4064 (thunderx2t99_asimd_store2_onelane): Delete.
4065 (thunderx2t99_asimd_store_stp): New.
4066 (thunderx2t99_asimd_store1): New.
4067 (thunderx2t99_asimd_store2): New.
4068 (thunderx2t99_asimd_store3): New.
4069 (thunderx2t99_asimd_store4): New.
4071 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4073 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4074 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4076 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4077 Segher Boessenkool <segher@kernel.crashing.org>
4080 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4083 2018-05-17 Richard Biener <rguenther@suse.de>
4085 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4086 for pruning loop and prune defs feeding only already visited PHIs.
4088 2018-05-17 Richard Biener <rguenther@suse.de>
4090 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4092 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4093 Richard Biener <rguenther@suse.de>
4095 PR tree-optimization/85793
4096 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4097 for VMAT_ELEMENTWISE.
4099 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4101 * internal-fn.h (lookup_internal_fn): Declare
4102 * internal-fn.c (lookup_internal_fn): New function.
4103 * gimple.c (gimple_build_call_from_tree): Handle calls to
4105 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4106 internal function names.
4107 * tree-pretty-print.c (dump_generic_node): Likewise.
4108 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4110 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4112 * gimple-fold.h (gimple_build): Make the function forms take
4113 combined_fn rather than built_in_function.
4114 (gimple_simplify): Likewise.
4115 * gimple-match-head.c (gimple_simplify): Likewise.
4116 * gimple-fold.c (gimple_build): Likewise.
4117 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4118 rather than gimple_build_call_internal.
4119 (get_initial_defs_for_reduction): Likewise.
4120 (vect_create_epilog_for_reduction): Likewise.
4121 (vectorizable_live_operation): Likewise.
4123 2018-05-17 Martin Liska <mliska@suse.cz>
4125 * gimple-ssa-sprintf.c (format_directive): Do not use
4126 space in between 'G_' and '('.
4128 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4131 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4132 even if the mask is not all ones.
4135 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4137 (ix86_gimple_fold_builtin): Likewise.
4140 * config/i386/i386.c: Include tree-vector-builder.h.
4141 (ix86_vector_shift_count): New function.
4142 (ix86_fold_builtin): Fold shift builtins by scalar count.
4143 (ix86_gimple_fold_builtin): Likewise.
4145 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4146 _mm512_setzero): New intrinsics.
4148 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4149 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4151 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4152 code generation for cases where splatting a value is not useful.
4153 * simplify-rtx.c (simplify_ternary_operation): Simplify
4154 vec_merge across a vec_duplicate and a paradoxical subreg forming
4155 a vector mode to a vec_concat.
4157 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4159 * config.gcc: Support "goldmont-plus".
4160 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4162 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4163 PROCESSOR_GOLDMONT_PLUS.
4164 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4165 (processor_target_table): Add "goldmont-plus".
4166 (PTA_GOLDMONT_PLUS): Define.
4167 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4168 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4169 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4170 (fold_builtin_cpu): Add "goldmont-plus".
4171 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4172 (ix86_option_override_internal): Add "goldmont-plus".
4173 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4174 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4175 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4176 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4178 2018-05-17 Richard Biener <rguenther@suse.de>
4180 PR tree-optimization/85757
4181 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4182 remove defs that only feed that PHI from further processing.
4184 2018-05-16 Jim Wilson <jimw@sifive.com>
4186 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4188 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4190 2018-05-16 Mark Wielaard <mark@klomp.org>
4192 * dwarf2out.c (count_index_strings): New function.
4193 (output_indirect_strings): Call count_index_strings and generate
4194 header for dwarf_version >= 5.
4196 2018-05-16 Mark Wielaard <mark@klomp.org>
4198 * dwarf2out.c (dwarf_FORM): New function.
4199 (set_indirect_string): Use dwarf_FORM.
4200 (reset_indirect_string): Likewise.
4201 (size_of_die): Likewise.
4202 (value_format): Likewise.
4203 (output_die): Likewise.
4204 (add_skeleton_AT_string): Likewise.
4205 (output_macinfo_op): Likewise.
4206 (index_string): Likewise.
4207 (output_index_string_offset): Likewise.
4208 (output_index_string): Likewise.
4209 (count_index_strings): Likewise.
4211 2018-05-16 Carl Love <cel@us.ibm.com>
4213 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4214 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4216 2018-05-16 Martin Jambor <mjambor@suse.cz>
4218 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4219 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4221 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4223 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4224 (fnma<mode>4): Likewise.
4225 (fms<mode>4): Likewise.
4226 (fnms<mode>4): Likewise.
4227 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4228 (aarch64_fnma<mode>4): Likewise.
4229 (aarch64_fms<mode>4): Likewise.
4230 (aarch64_fnms<mode>4): Likewise.
4231 (aarch64_fnmadd<mode>4): Likewise.
4233 2018-05-16 Jason Merrill <jason@redhat.com>
4235 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4237 2018-05-16 Richard Biener <rguenther@suse.de>
4239 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4240 (dump_stmt_cost): Declare.
4241 (add_stmt_cost): Dump cost we add.
4242 (add_stmt_costs): New function.
4243 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4245 (vect_analyze_stmt): Adjust prototype.
4246 (vectorizable_condition): Likewise.
4247 (vectorizable_live_operation): Likewise.
4248 (vectorizable_reduction): Likewise.
4249 (vectorizable_induction): Likewise.
4250 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4251 cost vector to pass to vectorizable_ and record afterwards.
4252 (vect_model_reduction_cost): Take cost vector argument and adjust.
4253 (vect_model_induction_cost): Likewise.
4254 (vectorizable_reduction): Likewise.
4255 (vectorizable_induction): Likewise.
4256 (vectorizable_live_operation): Likewise.
4257 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4258 SLP_TREE_NUMBER_OF_VEC_STMTS.
4259 (vect_analyze_slp_cost_1): Remove.
4260 (vect_analyze_slp_cost): Likewise.
4261 (vect_slp_analyze_node_operations): Take visited args and
4262 a target cost vector. Avoid processing already visited stmt sets.
4263 (vect_slp_analyze_operations): Use a local cost vector to gather
4264 costs and register those of non-discarded instances.
4265 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4266 (vect_schedule_slp_instance): Remove copying of
4267 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4269 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4270 adding cost. Record cost entry location.
4271 (vect_prologue_cost_for_slp_op): Function to compute cost of
4272 a constant or invariant generated for SLP vect in the prologue,
4273 split out from vect_analyze_slp_cost_1.
4274 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4275 (vect_model_promotion_demotion_cost): Likewise.
4276 (vect_model_store_cost): Likewise, make static.
4277 (vect_model_load_cost): Likewise.
4278 (vectorizable_bswap): Add cost vector arg and adjust.
4279 (vectorizable_call): Likewise.
4280 (vectorizable_simd_clone_call): Likewise.
4281 (vectorizable_conversion): Likewise.
4282 (vectorizable_assignment): Likewise.
4283 (vectorizable_shift): Likewise.
4284 (vectorizable_operation): Likewise.
4285 (vectorizable_store): Likewise.
4286 (vectorizable_load): Likewise.
4287 (vectorizable_condition): Likewise.
4288 (vectorizable_comparison): Likewise.
4289 (can_vectorize_live_stmts): Likewise.
4290 (vect_analyze_stmt): Likewise.
4291 (vect_transform_stmt): Adjust calls to vectorizable_*.
4292 * tree-vectorizer.c: Include gimple-pretty-print.h.
4293 (dump_stmt_cost): New function.
4295 2018-05-16 Richard Biener <rguenther@suse.de>
4297 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4298 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4299 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4300 (check_name): New callback.
4301 (dse_classify_store): Track cycles via a visited bitmap of PHI
4302 defs and simplify handling of in-loop and across loop dead stores
4303 and properly fail for loop-variant refs. Handle byte-tracking with
4304 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4307 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4309 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4310 (vect_get_mask_type_for_stmt): Likewise.
4311 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4313 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4314 to determine the statement's vector type and the vector type that
4315 should be used for calculating nunits. Deal with cases in which
4316 the type has to be deferred.
4317 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4318 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4319 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4320 (vect_determine_vf_for_stmt): New functions, split out from...
4321 (vect_determine_vectorization_factor): ...here.
4322 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4323 (vect_get_mask_type_for_stmt): New functions, split out from
4324 vect_determine_vectorization_factor.
4326 2018-05-16 Richard Biener <rguenther@suse.de>
4328 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4329 verify the [VEC_]COND_EXPR embedded comparison.
4331 2018-05-15 Martin Sebor <msebor@redhat.com>
4333 PR tree-optimization/85753
4334 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4335 RECORD_TYPE in addition to ARRAY_TYPE.
4337 2018-05-15 Martin Sebor <msebor@redhat.com>
4340 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4342 2018-05-15 Richard Biener <rguenther@suse.de>
4344 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4345 add by_clobber_p one. Change algorithm to collect all defs
4346 representing uses we need to walk and try reducing them to
4347 a single one before failing.
4348 (dse_dom_walker::dse_optimize_stmt): Adjust.
4350 2018-05-13 Mark Wielaard <mark@klomp.org>
4352 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4353 (size_of_loc_descr): Likewise.
4354 (output_loc_operands): Likewise.
4355 (output_loc_operands_raw): Likewise.
4356 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4357 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4358 (hash_loc_operands): Likewise.
4359 (compare_loc_operands): Likewise.
4361 2018-05-14 Mark Wielaard <mark@klomp.org>
4363 * dwarf2out.c (count_index_addrs): New function.
4364 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4366 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4368 PR tree-optimization/83648
4369 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4370 return value as malloc candidate.
4372 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4375 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4376 param as true in call to suggest_attribute.
4378 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4380 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4381 -mreadonly-in-sdata.
4383 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4385 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4387 (aarch64_crypto_aesd_fused): Likewise.
4389 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4391 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4392 (movsi_aarch64): Likewise.
4393 (load_pairsi): Likewise.
4394 (load_pairdi): Likewise.
4395 (store_pairsi): Likewise.
4396 (store_pairdi): Likewise.
4397 (load_pairsf): Likewise.
4398 (load_pairdf): Likewise.
4399 (store_pairsf): Likewise.
4400 (store_pairdf): Likewise.
4401 (zero_extend): Likewise.
4402 (trunc): Swap alternatives.
4403 (fcvt_target): Add '?' to prefer w over r.
4405 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4408 * config/i386/i386.md: Disallow non-commutative arithmetics in
4409 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4410 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4411 in the peephole2 before it.
4413 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4415 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4416 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4417 (ix86_handle_option): Handle -mcldemote.
4418 * config.gcc: New header.
4419 * config/i386/cldemoteintrin.h: New file.
4420 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4421 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4423 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4424 OPTION_MASK_ISA_CLDEMOTE.
4425 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4426 (ix86_valid_target_attribute_inner_p): Ditto.
4427 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4428 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4429 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4430 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4431 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4433 * config/i386/i386.opt: Add -mcldemote.
4434 * config/i386/x86intrin.h: New header.
4435 * doc/invoke.texi: Add -mcldemote.
4437 2018-05-14 Richard Biener <rguenther@suse.de>
4439 * doc/match-and-simplify.texi: Adjust :s documentation.
4441 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4443 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4444 intended memcpy size.
4445 (REORDER_45): Likewise.
4447 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4449 * sort.cc: New file.
4450 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4451 * vec.c (qsort_chk): Use gcc_qsort.
4452 * Makefile.in (OBJS-libcommon): Add sort.o.
4453 (build/sort.o): New target. Use it...
4454 (BUILD_RTL): ... here, and...
4455 (build/gencfn-macros): ... here, and...
4456 (build/genmatch): ... here.
4458 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4459 Chung-Ju Wu <jasonwucj@gmail.com>
4461 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4462 * config/nds32/nds32-graywolf.md: New file.
4463 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4464 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4466 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4467 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4468 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4469 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4470 * config/nds32/pipelines.md: Include n15 settings.
4472 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4473 Chung-Ju Wu <jasonwucj@gmail.com>
4475 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4476 * config/nds32/nds32-n13.md: New file.
4477 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4478 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4480 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4481 * config/nds32/nds32.md (pipeline_model): Add n13.
4482 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4483 * config/nds32/pipelines.md: Include n13 settings.
4485 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4486 Chung-Ju Wu <jasonwucj@gmail.com>
4488 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4489 * config/nds32/nds32-n10.md: New file.
4490 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4491 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4493 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4494 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4495 * config/nds32/nds32.md (pipeline_model): Add n10.
4496 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4497 * config/nds32/pipelines.md: Include n10 settings.
4499 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4500 Kito Cheng <kito.cheng@gmail.com>
4501 Chung-Ju Wu <jasonwucj@gmail.com>
4503 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4504 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4505 Add enum values for DSP extension instructions.
4506 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4508 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4509 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4511 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4512 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4513 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4514 * config/nds32/nds32-intrinsic.md: Likewise.
4515 * config/nds32/nds32_intrinsic.h: Likewise.
4516 * config/nds32/nds32-md-auxiliary.c: Likewise.
4517 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4518 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4519 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4520 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4521 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4522 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4523 TYPE_DMAC in switch statement.
4524 * config/nds32/nds32.c: New checking and implementation for DSP
4525 extension instructions.
4526 * config/nds32/nds32.h: Likewise.
4527 * config/nds32/nds32.md: Likewise.
4528 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4529 * config/nds32/predicates.md: Implement new predicates for DSP
4532 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4534 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4535 Reformat alternatives and attributes so it is easier to identify
4536 which constraints/attributes go with which instruction.
4537 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4538 (mov<mode>_softfloat32, FMOVE64): Likewise.
4539 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4540 (mov<mode>_softfloat64, FMOVE64): Likewise.
4542 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4544 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4546 (Basic PowerPC Built-in Functions): The new name of the
4547 subsection previously known as "PowerPC Built-in Functions".
4548 (Basic PowerPC Built-in Functions Available on all Configurations):
4550 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4551 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4552 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4553 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4555 2018-05-11 Martin Jambor <mjambor@suse.cz>
4558 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4561 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4564 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4566 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4568 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4569 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4570 (ix86_handle_option): Handle -mwaitpkg.
4571 * config.gcc: New header.
4572 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4573 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4574 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4576 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4577 OPTION_MASK_ISA_WAITPKG.
4578 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4579 (ix86_option_override_internal): Add PTA_WAITPKG.
4580 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4581 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4582 IX86_BUILTIN_TPAUSE.
4583 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4584 __builtin_ia32_umwait and __builtin_ia32_tpause.
4585 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4586 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4587 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4588 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4589 UNSPECV_TPAUSE): New.
4590 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4591 * config/i386/i386.opt: Add -mwaitpkg.
4592 * config/i386/waitpkgintrin.h: New file.
4593 * config/i386/x86intrin.h: New header.
4594 * doc/invoke.texi: Add -mwaitpkg.
4596 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4599 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4601 (cortex-m0): Use armv6s-m isa.
4602 (cortex-m0plus): Likewise.
4603 (cortex-m1): Likewise.
4604 (cortex-m0.small-multiply): Likewise.
4605 (cortex-m0plus.small-multiply): Likewise.
4606 (cortex-m1.small-multiply): Likewise.
4608 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4609 Jakub Jelinek <jakub@redhat.com>
4611 PR tree-optimization/85692
4612 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4613 source permute as well.
4615 2018-05-11 Martin Liska <mliska@suse.cz>
4618 * doc/extend.texi: Document LLVM style format for no_sanitize
4621 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4623 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4624 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4625 (mode_supports_vsx_dform_quad): Likewise.
4626 (mode_supports_vmx_dform): Move these functions to be next to the
4627 other mode_supports functions.
4628 (mode_supports_dq_form): Likewise.
4629 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4630 mode_supports_dq_form.
4631 (reg_offset_addressing_ok_p): Likewise.
4632 (offsettable_ok_by_alignment): Likewise.
4633 (rs6000_legitimate_offset_address_p): Likewise.
4634 (legitimate_lo_sum_address_p): Likewise.
4635 (rs6000_legitimize_address): Likewise.
4636 (rs6000_legitimize_reload_address): Likewise.
4637 (rs6000_secondary_reload_inner): Likewise.
4638 (rs6000_preferred_reload_class): Likewise.
4639 (rs6000_output_move_128bit): Likewise.
4641 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4643 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4644 Generate SImode target register for null target.
4645 <case IX86_BUILTIN_XGETBV>: Ditto.
4646 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4647 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4649 2018-05-10 Carl Love <cel@us.ibm.com>
4651 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4652 dcbtt and dcbtstt if operands[2] is 0.
4654 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4657 * config/i386/sse.md (usadv64qi): New expander.
4659 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4661 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4662 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4663 -maltivec=be support.
4664 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4665 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4666 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4667 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4668 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4669 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4670 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4671 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4672 altivec_vsumsws): Adjust.
4673 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4674 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4675 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4677 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4678 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4679 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4680 (altivec_lve<VI_char>x): Delete expand.
4681 (*altivec_lve<VI_char>x_internal): Rename to...
4682 (altivec_lve<VI_char>x): ... this.
4683 (altivec_lvxl_<mode>): Delete expand.
4684 (*altivec_lvxl_<mode>_internal): Rename to ...
4685 (altivec_lvxl_<mode>): ... this.
4686 (altivec_stvxl_<mode>): Delete expand.
4687 (*altivec_stvxl_<mode>_internal): Rename to ...
4688 (altivec_stvxl_<mode>): ... this.
4689 (altivec_stve<VI_char>x): Delete expand.
4690 (*altivec_stve<VI_char>x_internal): Rename to ...
4691 (altivec_stve<VI_char>x): ... this.
4692 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4693 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4694 reduc_plus_scal_<mode>): Adjust.
4695 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4697 (rs6000_cpu_cpp_builtins): Adjust.
4698 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4699 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4700 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4701 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4702 -maltivec=be support.
4703 (rs6000_split_vec_extract_var): Adjust.
4704 (rs6000_split_v4si_init): Adjust.
4705 (swap_selector_for_mode): Delete.
4706 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4707 altivec_expand_stvex_be): Delete.
4708 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4709 -maltivec=be support.
4710 (rs6000_gimple_fold_builtin): Ditto.
4711 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4713 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4714 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4715 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4716 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4717 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4718 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4719 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4720 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4721 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4722 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4723 anonymous split): Adjust.
4724 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4725 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4727 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4729 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4730 when --with-gxx-include-dir is also specified.
4731 * configure: Regenerate.
4733 2018-05-09 Jim Wilson <jimw@sifive.com>
4736 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4737 * config/riscv/t-withmultilib: New.
4738 * config/riscv/withmultilib.h: New.
4739 * doc/install.texi: Document RISC-V --with-multilib-list support.
4741 2018-05-09 Richard Biener <rguenther@suse.de>
4743 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4745 (vect_bb_vectorization_profitable_p): Adjust. Compute
4746 actual scalar cost using the cost vector and the add_stmt_cost
4749 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4751 PR rtl-optimization/85645
4752 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4753 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4755 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4757 PR rtl-optimization/85645
4758 * shrink-wrap.c (spread_components): Return a boolean saying if
4759 anything was changed.
4760 (try_shrink_wrapping_separate): Iterate spread_components until
4761 nothing changes anymore.
4763 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4765 PR rtl-optimization/85645
4766 * regrename.c (build_def_use): Also kill the chains that include the
4767 destination of a REG_CFA_REGISTER note.
4769 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4771 PR rtl-optimization/85645
4772 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4773 insn that has a REG_CFA_REGISTER note.
4775 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4777 * cfgexpand.c (expand_clobber): New function.
4778 (expand_gimple_stmt_1): Use it.
4779 * tree-vect-stmts.c (vect_clobber_variable): New function,
4781 (vectorizable_simd_clone_call): ...here.
4782 (vectorizable_store): Emit a clobber either side of an
4783 IFN_STORE_LANES sequence.
4784 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4786 2018-05-09 Tom de Vries <tom@codesourcery.com>
4789 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4790 (define_insn "trap_if_false"): Add exit after trap.
4792 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4794 PR rtl-optimization/85638
4795 * bb-reorder.c: Include common/common-target.h.
4796 (create_forwarder_block): New function extracted from...
4797 (fix_up_crossing_landing_pad): ...here. Rename into...
4798 (dw2_fix_up_crossing_landing_pad): ...this.
4799 (sjlj_fix_up_crossing_landing_pad): New function.
4800 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4801 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4802 from both partitions and exit the loop after one iteration.
4804 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4807 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4809 (Basic PowerPC Built-in Functions): The new name of the
4810 subsection previously known as "PowerPC Built-in Functions".
4811 (Basic PowerPC Built-in Functions Available on all Configurations):
4813 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4815 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4817 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4819 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4822 2018-05-08 Jim Wilson <jimw@sifive.com>
4824 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4825 (LD_EMUL_SUFFIX): New.
4826 (LINK_SPEC): Use it.
4828 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4830 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4832 (Basic PowerPC Built-in Functions): The new name of the
4833 subsection previously known as "PowerPC Built-in Functions".
4834 (Basic PowerPC Built-in Functions Available on all Configurations):
4836 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4838 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4840 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4842 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4845 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4848 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4849 after cmpelim optimization.
4851 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4853 * config.gcc: Support "goldmont".
4854 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4855 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4857 * config/i386/i386.c (m_GOLDMONT): Define.
4858 (processor_target_table): Add "goldmont".
4859 (PTA_GOLDMONT): Define.
4860 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4861 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4862 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4863 (fold_builtin_cpu): Add "goldmont".
4864 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4865 (ix86_option_override_internal): Add "goldmont".
4866 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4867 (processor_type): Add PROCESSOR_GOLDMONT.
4868 * config/i386/i386.md: Add CPU "glm".
4869 * config/i386/glm.md: New file.
4870 * config/i386/x86-tune.def: Add m_GOLDMONT.
4871 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4873 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4876 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4878 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4879 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4880 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4883 * config/i386/i386.c (ix86_fold_builtin): Handle
4884 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4887 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4888 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4890 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4893 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4894 (check_arch): Likewise.
4895 (check_fpu): Return the result rather than printing it.
4896 (end arch): Fix operator precedence.
4897 (end cpu): Likewise.
4898 (END): Print the result from check_fpu.
4900 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4901 Alan Hayward <alan.hayward@arm.com>
4902 David Sherwood <david.sherwood@arm.com>
4904 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4905 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4906 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4907 (*fcmuo<mode>_and): New patterns.
4909 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4911 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4912 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4913 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4914 (cmp_op, sve_imm_con): New code attributes.
4915 (SVE_COND_INT_CMP, imm_con): Delete.
4916 (cmp_op): Remove above unspecs from int attribute.
4917 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4919 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4920 comparison-specific unspecs.
4921 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4922 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4923 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4924 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4925 (*vec_fcm<cmp_op><mode>): Rename to...
4926 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4927 (*vec_fcmuo<mode>): Rename to...
4928 (*fcmuo<mode>): ...this and adjust likewise.
4929 (*pred_fcm<cmp_op><mode>): New pattern.
4930 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4931 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4933 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4935 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4936 (aarch64_emit_sve_predicated_cond): New function.
4937 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4938 (aarch64_emit_unspec_cond_or): Replace with...
4939 (aarch64_emit_sve_or_conds): ...this new function. Use
4940 aarch64_emit_sve_ptrue_op for the individual comparisons and
4941 aarch64_emit_binop to OR them together.
4942 (aarch64_emit_inverted_unspec_cond): Replace with...
4943 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4944 aarch64_emit_sve_ptrue_op for the comparison and
4945 aarch64_emit_unop to invert the result.
4946 (aarch64_expand_sve_vec_cmp_float): Update after the above
4947 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4949 2018-05-07 Nathan Sidwell <nathan@acm.org>
4951 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4952 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4953 (Backwards Compatibility): Likewise.
4955 2018-05-07 Luis Machado <luis.machado@linaro.org>
4959 2018-05-07 Luis Machado <luis.machado@linaro.org>
4961 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4962 <prefetch_dynamic_strides>: New const bool field.
4963 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4964 prefetch_dynamic_strides.
4965 (exynosm1_prefetch_tune): Likewise.
4966 (thunderxt88_prefetch_tune): Likewise.
4967 (thunderx_prefetch_tune): Likewise.
4968 (thunderx2t99_prefetch_tune): Likewise.
4969 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4971 (aarch64_override_options_internal): Update to set
4972 PARAM_PREFETCH_DYNAMIC_STRIDES.
4973 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4974 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4975 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4976 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4977 prefetch-dynamic-strides setting.
4979 2018-05-07 Luis Machado <luis.machado@linaro.org>
4981 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4982 <minimum_stride>: New const int field.
4983 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4984 minimum_stride field.
4985 (exynosm1_prefetch_tune): Likewise.
4986 (thunderxt88_prefetch_tune): Likewise.
4987 (thunderx_prefetch_tune): Likewise.
4988 (thunderx2t99_prefetch_tune): Likewise.
4989 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4990 (aarch64_override_options_internal): Update to set
4991 PARAM_PREFETCH_MINIMUM_STRIDE.
4992 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4993 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4994 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4995 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4996 stride is constant and is below the minimum stride threshold.
4998 2018-05-07 Luis Machado <luis.machado@linaro.org>
5000 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5003 2018-05-07 Luis Machado <luis.machado@linaro.org>
5005 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5006 <prefetch_dynamic_strides>: New const bool field.
5007 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5008 prefetch_dynamic_strides.
5009 (exynosm1_prefetch_tune): Likewise.
5010 (thunderxt88_prefetch_tune): Likewise.
5011 (thunderx_prefetch_tune): Likewise.
5012 (thunderx2t99_prefetch_tune): Likewise.
5013 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5015 (aarch64_override_options_internal): Update to set
5016 PARAM_PREFETCH_DYNAMIC_STRIDES.
5017 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5018 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5019 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5020 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5021 prefetch-dynamic-strides setting.
5023 2018-05-07 Luis Machado <luis.machado@linaro.org>
5025 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5026 <minimum_stride>: New const int field.
5027 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5028 minimum_stride field.
5029 (exynosm1_prefetch_tune): Likewise.
5030 (thunderxt88_prefetch_tune): Likewise.
5031 (thunderx_prefetch_tune): Likewise.
5032 (thunderx2t99_prefetch_tune): Likewise.
5033 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5034 (aarch64_override_options_internal): Update to set
5035 PARAM_PREFETCH_MINIMUM_STRIDE.
5036 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5037 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5038 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5039 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5040 stride is constant and is below the minimum stride threshold.
5042 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5045 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5046 the type is addressable. Don't force op into register if it has
5049 2018-05-05 Roland McGrath <mcgrathr@google.com>
5052 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5053 any section for which we don't know a specific type it should have,
5054 regardless of name. Previously this was done only for the exact
5055 names ".init_array", ".fini_array", and ".preinit_array".
5056 (default_elf_asm_named_section): Add comment about
5057 relationship with default_section_type_flags and SECTION_NOTYPE.
5058 (get_section): Don't consider it a type conflict if one side has
5059 SECTION_NOTYPE and the other doesn't, as long as neither has the
5060 SECTION_BSS et al used in the default_section_type_flags logic.
5062 2018-05-05 Tom de Vries <tom@codesourcery.com>
5065 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5066 (workaround_barsyncs): New function.
5067 (nvptx_reorg): Use workaround_barsyncs.
5068 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5069 (define_expand "nvptx_membar_cta"): New define_expand.
5070 (define_insn "*nvptx_membar_cta"): New insn.
5072 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5074 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5075 To improve optimization opportunities.
5076 * builtin-types.def: The new needed builtin types for the above.
5078 2018-05-04 Richard Biener <rguenther@suse.de>
5080 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5081 * gimple-ssa-store-merging.c
5082 (imm_store_chain_info::output_merged_store): Remove redundant create,
5083 release split_store vector contents on failure.
5084 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5085 scalar stmt vector on cache hit.
5087 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5089 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5091 * config.gcc (powerpc-xilinx-eabi*): Remove.
5092 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5094 (fusion_addis_mem_combo_load): Ditto.
5095 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5097 (rs6000_cpu_cpp_builtins): Ditto.
5098 * config/rs6000/rs6000-linux.c
5099 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5100 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5101 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5103 (rs6000_setup_reg_addr_masks): Ditto.
5104 (rs6000_init_hard_regno_mode_ok): Ditto.
5105 (rs6000_option_override_internal): Ditto.
5106 (legitimate_lo_sum_address_p): Ditto.
5107 (rs6000_legitimize_address): Ditto.
5108 (rs6000_legitimize_reload_address): Ditto.
5109 (rs6000_legitimate_address_p): Ditto.
5110 (abi_v4_pass_in_fpr): Ditto.
5111 (setup_incoming_varargs): Ditto.
5112 (rs6000_gimplify_va_arg): Ditto.
5113 (rs6000_split_multireg_move): Ditto.
5114 (rs6000_savres_strategy): Ditto.
5115 (rs6000_emit_prologue_components): Ditto.
5116 (rs6000_emit_epilogue_components): Ditto.
5117 (rs6000_emit_prologue): Ditto.
5118 (rs6000_emit_epilogue): Ditto.
5119 (rs6000_elf_file_end): Ditto.
5120 (rs6000_function_value): Ditto.
5121 (rs6000_libcall_value): Ditto.
5122 * config/rs6000/rs6000.h: Ditto.
5123 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5124 (TARGET_MINMAX): ... this. New.
5125 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5126 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5127 (*movsi_internal1_single): Delete.
5128 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5129 mfpu=, mxilinx-fpu): Delete.
5130 * config/rs6000/singlefp.h: Delete.
5131 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5132 * config/rs6000/t-rs6000: Ditto.
5133 * config/rs6000/t-xilinx: Delete.
5134 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5135 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5136 (VStype_simple): Delete.
5137 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5138 * config/rs6000/xfpu.h: Delete.
5139 * config/rs6000/xfpu.md: Delete.
5140 * config/rs6000/xilinx.h: Delete.
5141 * config/rs6000/xilinx.opt: Delete.
5142 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5143 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5145 2018-05-04 Tom de Vries <tom@codesourcery.com>
5148 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5151 2018-05-04 Richard Biener <rguenther@suse.de>
5154 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5155 (expand_complex_div_wide): Likewise.
5156 (expand_complex_operations_1): Likewise.
5157 (expand_complex_libcall): Preserve EH info of the original stmt.
5158 (tree_lower_complex): Handle removed blocks.
5159 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5160 on complex multiplication and division libcall builtins.
5162 2018-05-04 Richard Biener <rguenther@suse.de>
5165 * fold-const.c (negate_expr_p): Restrict negation of operand
5166 zero of a division to when we know that can happen without
5168 (fold_negate_expr_1): Likewise.
5170 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5173 * real.h (real_nextafter): Declare.
5174 * real.c (real_nextafter): New function.
5175 * fold-const-call.c (fold_const_nextafter): New function.
5176 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5177 CASE_CFN_NEXTTOWARD.
5178 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5179 even when arg1_mode is different from arg0_mode.
5181 2018-05-03 Nathan Sidwell <nathan@acm.org>
5183 * doc/extend.texi (Deprecated Features): Remove
5185 (Backwards Compatibility): Likewise.
5186 * doc/invoke.texi (C++ Language Options): Likewise.
5187 (C++ Dialect Options): Likewise.
5189 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5192 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5193 _mm512_mask_mullox_epi64): New intrinsics.
5195 2018-05-03 Tom de Vries <tom@codesourcery.com>
5198 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5199 dump files): Add offload-tree.
5201 2018-05-03 Richard Biener <rguenther@suse.de>
5203 PR tree-optimization/85615
5204 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5205 to loops not nested in BBs loop father to avoid creating multi-entry
5208 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5210 PR tree-optimization/70291
5211 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5212 arguments. Change return type to tree. Emit libcall as a new
5213 statement rather than replacing existing one when inplace_p is true.
5214 (expand_complex_multiplication_components): New function.
5215 (expand_complex_multiplication): Expand floating-point complex
5216 multiplication using the above.
5217 (expand_complex_division): Rename inner_type parameter to type.
5218 Update expand_complex_libcall call-site.
5219 (expand_complex_operations_1): Update expand_complex_multiplication
5220 and expand_complex_division call-sites.
5222 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5225 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5226 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5227 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5228 the highest significant bit of the shift count mask is clear. In
5229 check whether and[sq]i3 is needed verify that all significant bits
5230 of the shift count other than the highest are set.
5232 2018-05-02 Tom de Vries <tom@codesourcery.com>
5235 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5236 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5237 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5238 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5239 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5240 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5241 __builtin_goacc_parlevel_size.
5243 2018-05-02 Richard Biener <rguenther@suse.de>
5245 PR tree-optimization/85597
5246 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5247 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5250 2018-05-02 Tom de Vries <tom@codesourcery.com>
5253 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5254 dump files): Add ltrans-tree.
5256 2018-05-02 Tom de Vries <tom@codesourcery.com>
5259 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5260 dump files): Add wpa-ipa.
5262 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5264 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5265 powerpc*-*-linux*paired* target.
5266 * config/rs6000/750cl.h: Delete.
5267 * config/rs6000/paired.h: Delete.
5268 * config/rs6000/paired.md: Delete.
5269 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5271 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5272 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5273 comment. Remove paired float support.
5274 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5275 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5277 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5278 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5280 * config/rs6000/rs6000.c: Remove paired float support.
5281 (paired_expand_vector_init, paired_expand_vector_move,
5282 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5283 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5284 paired_expand_builtin, paired_expand_predicate_builtin,
5285 paired_init_builtins): Delete.
5286 * config/rs6000/rs6000.h: Remove paired float support.
5287 * config/rs6000/rs6000.md: Remove paired float support.
5288 (move_from_CR_ov_bit): Delete.
5289 * config/rs6000/rs6000.opt (mpaired): Delete.
5290 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5291 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5293 2018-05-02 Richard Biener <rguenther@suse.de>
5296 * gimplify.c (gimplify_save_expr): When in SSA form allow
5297 SAVE_EXPRs to compute to SSA vars.
5299 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5302 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5303 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5304 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5305 clobber operands[2], instead use a new pseudo. Formatting fixes.
5307 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5309 PR tree-optimization/85586
5310 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5311 exit early for statements in the same group if the accesses are
5314 2018-05-02 Tom de Vries <tom@codesourcery.com>
5317 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5320 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5322 PR tree-optimization/85143
5323 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5325 2018-05-01 Tom de Vries <tom@codesourcery.com>
5328 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5329 not found" error message.
5331 2018-05-01 Tom de Vries <tom@codesourcery.com>
5334 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5335 * vec.c (test_ordered_remove_if): New function.
5336 (vec_c_tests): Call test_ordered_remove_if.
5337 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5338 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5339 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5340 VEC_ORDERED_REMOVE_IF.
5342 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5344 PR tree-optimization/82665
5345 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5346 pointer subtraction where arguments come from a memchr call.
5348 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5350 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5351 --push-state --as-needed and --pop-state instead of --as-needed and
5352 --no-as-needed if ld supports it.
5353 * configure: Regenerated.
5356 * doc/install.texi2html: Replace _002d with - and _002a with * in
5357 generated html files using sed.
5359 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5362 * gcc-rich-location.c (blank_line_before_p): New function.
5363 (use_new_line): New function.
5364 (gcc_rich_location::add_fixit_insert_formatted): New function.
5365 * gcc-rich-location.h
5366 (gcc_rich_location::add_fixit_insert_formatted): New function.
5368 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5370 * selftest.c (assert_streq): Rename "expected" and "actual" to
5371 "val1" and "val2". Extend NULL-handling to cover both inputs
5372 symmetrically, while still requiring both to be non-NULL for a pass.
5373 * selftest.h (assert_streq): Rename "expected" and "actual" to
5375 (ASSERT_EQ): Likewise.
5376 (ASSERT_EQ_AT): Likewise.
5377 (ASSERT_KNOWN_EQ): Likewise.
5378 (ASSERT_KNOWN_EQ_AT): Likewise.
5379 (ASSERT_NE): Likewise.
5380 (ASSERT_MAYBE_NE): Likewise.
5381 (ASSERT_MAYBE_NE_AT): Likewise.
5382 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5383 the assertion to pass.
5384 (ASSERT_STREQ_AT): Likewise.
5386 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5388 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5389 interaction with -pie.
5391 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5393 * selftest.h: Fix alphabetization of per-source-file selftest
5396 2018-04-30 Jason Merrill <jason@redhat.com>
5398 PR c++/61982 - dead stores to destroyed objects.
5399 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5402 2018-04-30 Jason Merrill <jason@redhat.com>
5404 * tree.c (build_clobber): New.
5405 * tree.h: Declare it.
5406 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5408 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5410 * diagnostic-show-locus.c (layout::layout): Update for
5411 location_get_source_line returning a char_span.
5412 (struct char_span): Move to input.h.
5413 (struct correction): Update for fields in char_span becoming
5415 (struct source_line): Update for location_get_source_line
5416 returning a char_span.
5417 (layout::print_line): Likewise.
5418 * edit-context.c (edited_file::print_content): Likewise.
5419 (edited_file::print_diff_hunk): Likewise.
5420 (edited_file::print_run_of_changed_lines): Likewise.
5421 (edited_file::get_num_lines): Likewise.
5422 (edited_line::edited_line): Likewise.
5423 * final.c (asm_show_source): Likewise.
5424 * input.c (location_get_source_line): Convert return type
5425 from const char * to char_span, losing the final "line_len"
5427 (dump_location_info): Update for the above.
5428 (get_substring_ranges_for_loc): Likewise. Use a char_span
5429 when handling the literal within the line.
5430 (test_reading_source_line): Update for location_get_source_line
5431 returning a char_span.
5432 * input.h (class char_span): Move here from
5433 diagnostic-show-locus.c, converting from a struct to a class.
5434 Make data members private.
5435 (char_span::operator bool): New.
5436 (char_span::length): New.
5437 (char_span::get_buffer): New.
5438 (char_span::operator[]): New.
5439 (char_span::subspan): Make const.
5440 (char_span::xstrdup): New.
5441 (location_get_source_line): Convert return type from const char *
5442 to char_span, losing the final "line_size" param.
5444 2018-04-30 Jan Hubicka <jh@suse.cz>
5446 * lto-wrapper.c (ltrans_priorities): New static var.
5447 (cmp_priority): New.
5448 (run_gcc): Read priorities and if doing parallel build order
5449 the Makefile by them.
5451 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5453 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5455 2018-04-30 Richard Biener <rguenther@suse.de>
5457 * tree-cfg.c (verify_address): Remove base argument, add
5458 flag whether to check TREE_ADDRESSABLE and do that.
5459 (verify_expr): Remove.
5460 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5461 (verify_gimple_assign_single): Likewise.
5462 (verify_gimple_switch): Likewise.
5463 (verify_expr_location_1): Dereference tp once. Add (disabled)
5464 piece from verify_expr.
5465 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5467 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5469 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5471 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5473 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5474 (small_data_pattern): Likewise.
5475 (arc_rewrite_small_data): Likewise.
5476 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5477 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5478 (get_symbol_alignment): New function.
5479 (legitimate_small_data_address_p): Likewise.
5480 (legitimate_scaled_address): Update, call
5481 legitimate_small_data_address_p.
5482 (output_sdata): New static variable.
5483 (arc_print_operand): Update how we handle small data operands.
5484 (arc_print_operand_address): Likewise.
5485 (arc_legitimate_address_p): Update, use
5486 legitimate_small_data_address_p.
5487 (arc_rewrite_small_data_p): Remove.
5488 (arc_rewrite_small_data_1): Likewise.
5489 (arc_rewrite_small_data): Likewise.
5490 (small_data_pattern): Likewise.
5491 (compact_sda_memory_operand): Update to use
5492 legitimate_small_data_address_p and get_symbol_alignment.
5493 (prepare_move_operands): Don't rewite sdata pattern.
5494 (prepare_extend_operands): Remove.
5495 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5497 (zero_extendqisi2): Likewise.
5498 (zero_extendhisi2): Likewise.
5499 (extendqihi2): Likewise.
5500 (extendqisi2): Likewise.
5501 (extendhisi2): Likewise.
5505 * config/arc/constraints.md (Usd): Change it to memory constraint.
5507 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5509 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5510 as source of std instructions.
5511 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5512 allow 6-bit constants as source for store instructions.
5513 (movdi_insn): Update instruction pattern to allow 6-bit constants
5514 as source for store instructions.
5516 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5518 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5520 2018-04-30 Nathan Sidwell <nathan@acm.org>
5521 Sandra Loosemore <sandra@codesourcery.com>
5523 * dumpfile.c (dump_open): Allow '-' for stdout.
5524 * doc/invoke.texi (Developer Options): Document dump filename
5525 determination early. Document stdin/stdout selection.
5527 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5529 Microblaze Target: PIC data text relative
5531 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5532 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5534 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5535 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5536 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5537 New addressing mode for data-text relative position indepenedent code.
5538 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5539 'ADDRESS_SYMBOLIC_TXT_REL'.
5540 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5541 (microblaze_legitimate_pic_operand): Exclude function calls from
5542 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5543 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5545 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5546 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5547 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5548 for 'address + offset'.
5549 (microblaze_expand_prologue): Add new function prologue call for
5551 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5552 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5553 table in case of TARGET_PIC_DATA_TEXT_REL.
5554 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5555 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5556 Add new macros 'UNSPEC_TEXT',
5557 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5558 + exclude function calls from 'UNSPEC_PLT' in case of data text
5560 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5561 new target hook for generating address diff vector tables in case of
5563 * doc/tm.texi : Regenerate.
5564 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5565 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5566 of addr diff vector generation.
5567 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5568 target hook definition.
5569 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5570 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5571 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5572 option for data text relative.
5574 2018-04-30 Richard Biener <rguenther@suse.de>
5576 * tree-chrec.h (evolution_function_is_constant_p): Remove
5578 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5580 2018-04-30 Richard Biener <rguenther@suse.de>
5583 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5585 2018-04-30 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/28364
5588 PR tree-optimization/85275
5589 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5590 copying first exit test.
5592 2018-04-28 Mark Wielaard <mark@klomp.org>
5594 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5596 (dwarf_AT): Handle DW_AT_addr_base.
5597 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5599 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5602 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5603 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5604 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5605 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5607 2018-04-28 Richard Biener <rguenther@suse.de>
5609 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5610 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5611 to reflect use. Only add interesting stmts.
5613 2018-04-27 Martin Jambor <mjambor@suse.cz>
5616 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5617 the jump function allows for passing through aggregate values.
5619 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5621 * input.h (in_system_header_at): Convert from macro to inline
5623 (from_macro_expansion_at): Likewise.
5624 (from_macro_definition_at): Likewise.
5626 2018-04-27 Jeff Law <law@redhat.com>
5628 * config.gcc: Mark tile* targets as deprecated/obsolete.
5630 2018-04-27 Richard Biener <rguenther@suse.de>
5632 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5635 2018-04-27 Richard Biener <rguenther@suse.de>
5637 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5639 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5641 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5642 with Yd constraint. Set "preferred_for_speed" attribute from
5643 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5645 (*movdi_internal): Ditto.
5646 (movti_interunit splitters): Remove
5647 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5648 (movdi_interunit splitters): Ditto.
5649 * config/i386/constraints.md (Ye): Remove.
5650 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5652 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5655 * config/aarch64/constraints.md (Usg): Limit to 31.
5658 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5660 PR tree-optimization/85529
5661 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5662 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5663 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5664 zero extension or masking of the MSB bit.
5665 (optimize_range_tests): Add FIRST_BB argument, pass it through
5666 to optimize_range_tests_var_bound.
5667 (maybe_optimize_range_tests, reassociate_bb): Adjust
5668 optimize_range_tests callers.
5670 2018-04-26 Richard Biener <rguenther@suse.de>
5671 Jakub Jelinek <jakub@redhat.com>
5673 * cgraph.h (symbol_table): Just declare debug method here.
5674 * symtab.c (symbol_table::debug): Define.
5676 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5678 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5680 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5682 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5683 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5684 (*movdi_internal): Substitute Yi and Yj constraint with x
5685 and Ym and Yn constraint with y constraint. Update "isa"
5686 attribute and set "preferred_for_speed" attribute from
5687 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5688 (*movsi_internal): Ditto.
5689 (*movdf_internal): Ditto.
5690 (*movsf_internal): Ditto.
5691 (*zero_extendsidi2): Ditto.
5692 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5693 (sse2_loadld): Ditto.
5694 (*vec_extract<ssevecmodelower>_0): Ditto.
5695 (*vec_extractv4si_0_zext_sse4): Ditto.
5696 (vec_concatv2di): Ditto.
5697 (*vec_dup<mode>): Ditto.
5698 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5699 * config/i386/constraints.md (Yi): Remove.
5704 2018-04-26 Nathan Sidwell <nathan@acm.org>
5706 * dumpfile.c (dump_open): New.
5707 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5708 (dump_finish): Detect stdio/stderr by value not name.
5710 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5712 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5714 2018-04-26 Tom de Vries <tom@codesourcery.com>
5717 * config/nvptx/nvptx.c (verify_neutering_jumps)
5718 (verify_neutering_labels): New function
5719 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5721 2018-04-26 Tom de Vries <tom@codesourcery.com>
5724 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5725 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5728 2018-04-26 Richard Biener <rguenther@suse.de>
5729 Tom de Vries <tom@codesourcery.com>
5732 * lto-streamer-out.c (output_function): Fixup loops if required to match
5733 discovery done in the reader.
5735 2018-04-26 Richard Biener <rguenther@suse.de>
5737 PR tree-optimization/85116
5738 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5739 have a loop exit from the single latch predecessor. Remove
5740 case of header with just condition.
5741 (ch_base::copy_headers): Exclude infinite loops from any
5743 (pass_ch::execute): Record exits.
5745 2018-04-26 Richard Biener <rguenther@suse.de>
5747 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5748 prologue cost vector and pass it to vect_get_load_cost.
5749 (vect_get_peeling_costs_all_drs): Likewise.
5750 (vect_peeling_hash_get_lowest_cost): Likewise.
5751 (vect_enhance_data_refs_alignment): Likewise.
5753 2018-04-26 Richard Biener <rguenther@suse.de>
5756 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5757 checking of integer<->pointer conversions.
5758 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5759 sign-/zero-extending pointer types.
5760 (expand_omp_for_static_chunk): Likewise.
5762 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5763 Jean Lee <xiaoyur347@gmail.com>
5765 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5766 (TARGET_ASAN_SHADOW_OFFSET): Define.
5767 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5768 true for -fsanitize=address.
5770 2018-04-25 Mark Wielaard <mark@klomp.org>
5772 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5775 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5777 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5778 than "alu", remove explicit "memory" and "imm_disp" attributes.
5779 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5782 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5783 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5784 gen_lowpart_no_emit.
5786 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5789 * config/i386/i386.c (ix86_expand_builtin): Change memory
5790 operand to XI, extend p0 to Pmode.
5791 * config/i386/i386.md: Change unspec volatile and operand
5792 1 mode to XI, change operand 0 mode to P.
5794 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5796 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5797 GET_MODE_MASK before any checking.
5798 (nds32_can_use_bset_p): Likewise.
5799 (nds32_can_use_btgl_p): Likewise.
5801 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5803 * config/nds32/nds32-doubleword.md: New define_split pattern for
5804 illegal register number.
5806 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5808 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5810 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5812 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5814 2018-04-25 Richard Biener <rguenther@suse.de>
5816 * lto-streamer.h (LTO_major_version): Bump to 8.
5818 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5820 * BASE-VER: Set to 9.0.0.
5822 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5824 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5825 in __abskf2 and __powikf2.
5827 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5830 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5831 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5832 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5833 Use the above on operand 2. Reindent.
5834 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5836 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5839 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5841 * config/i386/i386.opt (mcet): Removed.
5842 * doc/install.texi: Remove -mcet documentation.
5843 * doc/invoke.texi: Likewise.
5845 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5848 * doc/install.texi: Remove -mcet from bootstrap-cet.
5850 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5853 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5854 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5858 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5859 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5860 containing a CONST_VECTOR.
5862 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5864 * doc/install.texi: Update newlib dependency for nvptx.
5866 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5869 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5870 instead of INTVAL when shifting x left.
5872 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5874 PR tree-optimization/85478
5875 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5876 vect_grouped_store_supported for single element vectors.
5878 2018-04-24 Richard Biener <rguenther@suse.de>
5881 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5882 load cost increase to the case of non-constant step.
5884 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5887 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5888 destination if any_malformed_asm.
5890 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5893 * expr.c (store_field): In the bitfield case, if the value comes from
5894 a function call and is returned in registers by means of a PARALLEL,
5895 do not change the mode of the temporary unless BLKmode and VOIDmode.
5897 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5899 PR rtl-optimization/85423
5900 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5901 dependencies to debug insns when the previous insn is non-debug.
5903 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5905 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5906 enums into a single definition.
5907 (fls): Fix predicates and printing.
5910 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5912 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5913 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5914 and short u6 immediate.
5915 (check_if_valid_sleep_operand): Remove.
5916 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5918 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5920 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5921 flag_always_save_lp condition.
5922 * config/nds32/nds32.opt (malways-save-lp): New option.
5924 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5926 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5927 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5928 * config/nds32/nds32.h
5929 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5930 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5932 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5934 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5935 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5937 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5938 Chung-Ju Wu <jasonwucj@gmail.com>
5940 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5942 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5943 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5945 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5947 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5949 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5951 * config/nds32/nds32-protos.h (nds32_data_alignment,
5952 nds32_local_alignment): Declare.
5953 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5954 nds32_local_alignment): New functions.
5955 (TARGET_CONSTANT_ALIGNMENT): Define.
5956 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5958 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5960 * config/nds32/nds32.c
5961 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5962 (TARGET_MODES_TIEABLE_P): Likewise.
5964 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5966 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5969 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5970 Chung-Ju Wu <jasonwucj@gmail.com>
5972 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5973 for unaligned access.
5974 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5976 * config/nds32/nds32-intrinsic.md: Likewise.
5977 * config/nds32/nds32_intrinsic.h: Likewise.
5978 * config/nds32/nds32.h (nds32_builtins): Likewise.
5979 * config/nds32/nds32.opt (munaligned-access): New option.
5980 * config/nds32/nds32.c (nds32_asm_file_start): Display
5981 flag_unaligned_access status.
5983 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5985 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5986 -mno-relax is present.
5987 * config/riscv/linux.h (LINK_SPEC): Ditto.
5989 2018-04-20 Martin Sebor <msebor@redhat.com>
5992 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5994 (gimple_fold_builtin_stxcpy_chk): Same.
5995 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5997 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6000 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6001 __powikf2 when long double is IEEE 128-bit.
6003 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6005 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6006 step to make sure stack always aligned.
6008 2018-04-20 Carl Love <cel@us.ibm.com>
6011 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6012 size check for arg0.
6014 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6015 Tom de Vries <tom@codesourcery.com>
6018 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6019 Emit insns for calls too.
6020 (nvptx_find_par): Always look for worker-level predecessor insn.
6021 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6023 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6024 (nvptx_process_pars): Propagate frames for calls.
6026 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6029 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6031 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6032 (ix86_handle_option): Don't handle OPT_mibt.
6033 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6035 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6037 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6038 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6039 (ix86_target_macros): Define __CET__ with flag_cf_protection
6040 for -fcf-protection.
6041 * config/i386/i386.c (isa2_opts): Remove -mibt.
6042 * config/i386/i386.h (TARGET_IBT): Removed.
6043 (TARGET_IBT_P): Likewise.
6044 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6045 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6046 * config/i386/i386.opt (mcet): Update help message.
6049 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6050 -mcet as an alias for -mshstk.
6052 2018-04-20 Richard Biener <rguenther@suse.de>
6055 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6056 complexity by forcing a single use of the multiply operand.
6058 2018-04-20 Martin Jambor <mjambor@suse.cz>
6061 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6062 recursion dependency to only apply to non-clones.
6064 2018-04-20 Martin Jambor <mjambor@suse.cz>
6067 * ipa-cp.c (create_specialized_node): Check that clones of
6068 self-recursive edges exist during IPA-CP.
6070 2018-04-19 Toon Moene <toon@moene.org>
6072 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6075 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6077 PR tree-optimization/85467
6078 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6079 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6080 VECTOR_CST element to type.
6082 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6085 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6086 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6087 (builtin_longjmp): Likewise.
6088 (save_stack_nonlocal): New pattern.
6089 (restore_stack_nonlocal): Likewise.
6091 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6094 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6095 Replace ASM_OUTPUT_LABEL with fprintf.
6097 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6100 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6101 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6102 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6103 define __IBT__ and __SHSTK__ for -fcf-protection.
6104 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6106 (ix86_trampoline_init): Likewise.
6107 (x86_output_mi_thunk): Likewise.
6108 (ix86_notrack_prefixed_insn_p): Likewise.
6109 (ix86_option_override_internal): Don't disallow -fcf-protection.
6110 * config/i386/i386.md (rdssp<mode>): Also enable for
6112 (incssp<mode>): Likewise.
6113 (nop_endbr): Likewise.
6114 * config/i386/i386.opt (mcet): Change help message to built-in
6118 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6119 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6120 enable CET built-in functions.
6122 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6124 * common/config/i386/i386-common.c
6125 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6126 OPTION_MASK_ISA_MOVDIRI_UNSET,
6127 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6128 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6129 * config.gcc (movdirintrin.h): New header.
6130 * config/i386/cpuid.h (bit_MOVDIRI,
6131 bit_MOVDIR64B): New bits.
6132 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6134 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6135 (VOID, PVOID, PCVOID)): New function types.
6136 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6137 __builtin_ia32_directstoreu_u64,
6138 __builtin_ia32_movdir64b): New builtins.
6139 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6140 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6142 (ix86_valid_target_attribute_inner_p): Ditto.
6143 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6144 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6145 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6146 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6147 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6148 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6149 (movdiri<mode>, movdir64b_<mode>): New.
6150 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6151 * config/i386/immintrin.h: Include movdirintrin.h.
6152 * config/i386/movdirintrin.h: New file.
6153 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6155 2018-04-19 Richard Biener <rguenther@suse.de>
6158 * cfg.c (clear_bb_flags): When loop state says we have
6159 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6161 2018-04-19 Richard Biener <rguenther@suse.de>
6163 PR tree-optimization/84737
6164 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6165 copying restrict info.
6166 (vect_setup_realignment): Use it.
6167 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6168 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6169 the first DR to all generated stores.
6170 (vectorizable_load): Likewise for loads.
6172 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6174 PR tree-optimization/85446
6175 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6176 the integral and pointer types to have the same precision.
6178 * doc/install.texi: Document --disable-cet being the default and
6181 2018-04-18 Martin Liska <mliska@suse.cz>
6183 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6186 2018-04-18 Martin Liska <mliska@suse.cz>
6189 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6192 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6193 arguments if they are comparable.
6195 2018-04-18 Martin Liska <mliska@suse.cz>
6198 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6201 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6204 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6207 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6208 ENDBR after calling __morestack.
6210 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6213 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6214 by using gcc_base_ver to generate a gcc_driver_version, and use
6215 it when generating GCC_DRIVER_NAME.
6216 * configure: Regenerate.
6218 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6221 * config.gcc: Obsolete powerpc*-*-*spe*.
6223 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6226 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6227 (stabstr_D): Change type of unum from unsigned int to
6228 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6231 2018-04-17 Jim Wilson <jimw@sifive.com>
6234 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6235 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6236 Set arg_pointer_offset after using pretend_args_size.
6238 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6240 PR rtl-optimization/85431
6241 * dse.c (record_store): Ignore zero width stores.
6244 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6245 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6246 __builtin_stack_restore rather than after it.
6247 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6248 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6249 argument instead of virtual_dynamic_stack_rtx.
6251 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6253 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6255 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6256 Add note to error message to explain internal mapping of overloaded
6257 built-in function name to non-overloaded built-in function name.
6258 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6261 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6264 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6265 where the inputs overlap with the output.
6267 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6269 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6270 (=v, v) alternative and explicit "memory" attribute.
6271 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6272 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6274 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6275 "sselog1" type instead of "sselog".
6276 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6277 "sselog". Remove explicit "memory" attribute.
6278 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6279 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6281 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6282 "sselog1" type instead of "sselog". Remove explicit "memory"
6284 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6285 use "sselog1" type instead of "sselog". Remove explicit "memory"
6287 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6288 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6290 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6291 "sselog1" type instead of "sselog". Remove explicit "memory"
6293 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6294 use "sselog1" type instead of "sselog". Remove explicit "memory"
6298 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6301 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6304 2018-04-17 Martin Jambor <mjambor@suse.cz>
6307 * ipa-cp.c (create_specialized_node): Call
6308 expand_all_artificial_thunks if necessary.
6310 2018-04-17 Martin Liska <mliska@suse.cz>
6313 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6314 in message, remote space in between '_G' and '('.
6316 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6319 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6320 avx512f_vmcmp<mode>3<round_saeonly_name>,
6321 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6322 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6323 avx512f_rndscale<mode><round_saeonly_name>,
6324 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6325 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6326 Use %<iptr>2 instead of %2 for -masm=intel.
6327 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6328 avx512f_vcvttss2usi<round_saeonly_name>,
6329 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6331 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6332 avx512f_vcvttsd2usi<round_saeonly_name>,
6333 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6334 Use %q1 instead of %1 for -masm=intel.
6335 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6336 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6337 of %3 for -masm=intel.
6338 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6340 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6342 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6344 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6345 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6347 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6349 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6350 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6351 %g1 and one with %0 and %1.
6352 (avx512er_vmrcp28<mode><round_saeonly_name>,
6353 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6355 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6356 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6357 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6358 of %0 and %{%4%} for -masm=intel.
6359 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6360 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6361 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6362 order of %0 and %{%5%}%{z%} for -masm=intel.
6364 2018-04-17 Jan Hubicka <jh@suse.cz>
6367 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6369 2018-04-17 Martin Liska <mliska@suse.cz>
6372 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6373 for target_clone error message. Make default implementation
6374 clone to be a local declaration.
6375 (separate_attrs): Add new argument and check for an empty
6377 (expand_target_clones): Handle it.
6378 (ipa_target_clone): Make redirection just for target_clones
6381 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6382 Tom de Vries <tom@codesourcery.com>
6385 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6386 tiled basic blocks without omp continue statements.
6388 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6391 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6392 vec_extract expression as having side effects to make sure it gets
6395 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6398 * config/i386/i386.c (get_builtin_code_for_version): Check
6401 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6404 * gcc/config.gcc: Support "skylake".
6405 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6407 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6408 (processor_target_table): Add "skylake".
6409 (ix86_option_override_internal): Add "skylake".
6410 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6411 PROCESSOR_CANNONLAKE.
6412 (get_builtin_code_for_version): Fix priority for
6413 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6414 PROCESSOR_SKYLAKE-AVX512.
6415 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6416 (processor_type): Add PROCESSOR_SKYLAKE.
6418 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6419 Jason Merrill <jason@redhat.com>
6422 * convert.c (convert_to_integer_1): Use direct recursion for
6423 enumeral types and types with a precision less than the number
6424 of bits in their mode.
6426 2018-04-16 Julia Koval <julia.koval@intel.com>
6429 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6430 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6432 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6435 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6436 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6437 and -mno-direct-move.
6439 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6442 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6443 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6444 Ensure negative shifts result in {0}.
6446 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6448 PR rtl-optimization/79916
6449 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6450 regs (if any) to define how to gnerate SD moves when LRA is in
6453 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6455 PR rtl-optimization/85393
6456 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6457 * except.c (expand_dw2_landing_pad_for_region): Make static.
6458 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6459 a label and unconditional jump to old_bb, rather than
6460 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6463 PR rtl-optimization/85376
6464 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6465 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6466 instead of a specific value.
6468 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6469 Bin Cheng <bin.cheng@arm.com>
6471 PR tree-optimization/82965
6472 PR tree-optimization/83991
6473 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6474 by_profile_only parameter.
6475 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6476 information if the loop was predicted to iterate too many times.
6477 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6479 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6482 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6485 2018-04-13 Martin Liska <mliska@suse.cz>
6486 Jakub Jelinek <jakub@redhat.com>
6489 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6490 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6491 * builtins.c (expand_builtin_memory_copy_args): Use
6492 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6493 handle dest_addr == pc_rtx.
6495 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6498 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6499 asked to not generate direct moves.
6500 (fix_trunc<mode>si2_stfiwx): Similar.
6501 (fix_trunc<mode>si2_internal): Similar.
6503 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6506 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6507 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6508 lookup if dest in some wider mode is known to be const0_rtx and
6509 if so, record permanent equivalence for it to be ZERO_EXTEND of
6510 the narrower mode destination.
6512 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6514 * lto-streamer-out.c (output_function): Revert 259346.
6515 * omp-expand.c (expand_oacc_for): Likewise.
6517 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6519 PR rtl-optimization/85354
6520 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6521 * sel-sched.c (sel_global_init): ... here.
6523 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6526 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6527 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6528 mode for PE-COFF targets.
6529 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6530 (i386_pe_asm_lto_end): Likewise.
6531 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6532 (TARGET_ASM_LTO_END): Likewise.
6533 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6534 (i386_pe_asm_lto_start): New function.
6535 (i386_pe_asm_lto_end): Likewise.
6537 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6538 Richard Biener <rguenther@suse.de>
6541 * lto-streamer-out.c (output_function): Fix CFG loop state before
6543 * omp-expand.c (expand_oacc_for): Handle calls to internal
6544 functions like regular functions.
6546 2018-04-12 Richard Biener <rguenther@suse.de>
6549 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6550 for the early LTO debug to properly generate references to it
6551 during DIE emission. Do not re-use that for the skeleton for
6553 (dwarf2out_early_finish): Likewise.
6555 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6558 * config/i386/sse.md
6559 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6560 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6561 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6562 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6563 and output is a reg, avoid creating invalid lowpart subreg, but
6564 instead split into a 512-bit move. Don't split if not AVX512VL,
6565 input is xmm16+ reg and output is a mem.
6566 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6567 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6568 xmm16+ reg and output is a mem.
6570 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6572 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6573 also for flag_dwarf2_cfi_asm.
6575 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6577 PR rtl-optimization/85342
6578 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6579 a bool scalar var inside of the loop instead. Don't try to update
6580 recog_data.operand after failed apply_change_group.
6582 2018-04-12 Tom de Vries <tom@codesourcery.com>
6585 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6586 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6587 array with flexible array member as array without given dimension.
6588 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6589 argument for undefined param to true.
6591 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6594 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6595 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6596 from PowerPC section.
6597 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6598 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6599 help text that is too long.
6600 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6601 help text that is too long.
6602 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6603 help text that is too long.
6605 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6607 * config/alpha/alpha.md (stack_probe_internal): Rename
6608 from "probe_stack". Update all callers.
6610 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6612 PR rtl-optimization/84566
6613 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6614 sched_macro_fuse_insns.
6616 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6619 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6620 (compute_block_dependences): ... from here.
6622 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6624 PR tree-optimization/85331
6625 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6626 from int to HOST_WIDE_INT.
6628 2018-04-11 Martin Jambor <mjambor@suse.cz>
6631 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6632 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6633 not the same as the source val.
6634 (cgraph_edge_brings_value_p): New parameter.
6635 (gather_edges_for_value): Pass destination value to
6636 cgraph_edge_brings_value_p.
6637 (perhaps_add_new_callers): Likewise.
6638 (get_info_about_necessary_edges): Likewise and exclude values brought
6639 only by self-recursive edges.
6640 (create_specialized_node): Redirect only clones of self-calling edges.
6641 (+self_recursive_pass_through_p): New function.
6642 (find_more_scalar_values_for_callers_subset): Use it.
6643 (find_aggregate_values_for_callers_subset): Likewise.
6644 (known_aggs_to_agg_replacement_list): Removed.
6645 (decide_whether_version_node): Re-calculate known constants for all
6646 remaining context clones.
6648 2018-04-11 Richard Biener <rguenther@suse.de>
6651 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6652 from early DWARF output.
6653 (dwarf2out_early_finish): Output line info unconditionally into
6654 early DWARF and add reference to it.
6656 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6659 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6660 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6661 other than V2DFmode using iptr mode attribute.
6662 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6664 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6666 PR rtl-optimization/84659
6667 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6669 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6672 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6674 (output_loc_list): Pass address of a dummy size variable even in the
6675 locview handling loop.
6676 (index_location_lists): Add comment on why skip_loc_list_entry can't
6679 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6682 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6685 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6688 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6689 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6690 and -mstring-compare-inline-limit.
6692 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6695 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6696 for stack clash protection in a register whenever we need it to be in
6699 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6701 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6702 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6704 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6707 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6709 (mlong-double-): Ditto.
6710 * config/rs6000/sysv4.opt (msdata=): Ditto.
6711 (mtls-size=): Ditto.
6713 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6715 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6716 erroneous entries for
6717 "vector int vec_ldl (int, long int *)", and
6718 "vector unsigned int vec_ldl (int, unsigned long int *)".
6719 Add comments and entries for
6720 "vector bool char vec_ldl (int, bool char *)",
6721 "vector bool short vec_ldl (int, bool short *)",
6722 "vector bool int vec_ldl (int, bool int *)",
6723 "vector bool long long vec_ldl (int, bool long long *)",
6724 "vector pixel vec_ldl (int, pixel *)",
6725 "vector long long vec_ldl (int, long long *)",
6726 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6727 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6728 type tree bool_long_long_type_node and correct definition of
6729 bool_V2DI_type_node to make reference to this new type tree.
6730 (rs6000_mangle_type): Replace erroneous reference to
6731 bool_long_type_node with bool_long_long_type_node.
6732 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6733 comments to emphasize sign distinctions for char and int types and
6734 replace RS6000_BTI_bool_long constant with
6735 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6736 use of RS6000_BTI_pixel.
6737 (bool_long_type_node): Remove this macro definition.
6738 (bool_long_long_type_node): New macro definition
6740 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6742 PR rtl-optimization/85300
6743 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6744 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6745 simplify_unary_operation fails.
6747 2018-04-10 Martin Liska <mliska@suse.cz>
6749 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6750 cgraph_edge and ipa_ref.
6752 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6756 * config/i386/sse.md
6757 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6758 computation of the VEC_MERGE selector from mask.
6759 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6760 Fix decoding of the VEC_MERGE selector into mask.
6762 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6764 PR tree-optimization/85286
6765 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6767 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6769 * final.c (final_1): Set insn_last_address as well as
6770 insn_current_address.
6772 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6775 * explow.c (emit_stack_probe): Call validize_mem on memory location
6776 before passing it to gen_probe_stack. Create address operand and
6777 legitimize it for the probe_stack_address case.
6779 2018-04-09 Jan Hubicka <jh@suse.cz>
6782 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6783 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6784 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6786 (free_lang_data): Rebuild type inheritance graph.
6788 2018-04-09 Martin Sebor <msebor@redhat.com>
6790 * invoke.texi (-finline-small-functions): Mention other optimization
6792 (-findirect-inlining, -fpartial-inlining): Same.
6793 (-finline-functions-called-once): Same.
6794 (-freorder-blocks-and-partition): Same.
6796 2018-04-09 Jan Hubicka <jh@suse.cz>
6799 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6800 jumps; choose last target that matches the criteria (i.e.
6801 no partition changes for non-crossing jumps).
6802 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6803 support for redirecting crossing jumps to non-crossing.
6805 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6807 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6808 also for naked functions.
6810 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6812 * config/arc/arc.md (add_shift): New pattern.
6813 (add_shift2): Likewise.
6814 (sub_shift): Likewise.
6815 (sub_shift_cmp0_noout): Likewise.
6816 (compare_si_ashiftsi): Likewise.
6817 (xbfu_cmp0_noout): New combine pattern.
6818 (xbfu_cmp0"): Likewise.
6819 (movsi_set_cc_insn): Place the predicable variant first.
6820 (commutative_binary_cmp0_noout): Remove clobber.
6821 (commutative_binary_cmp0): New pattern.
6822 (noncommutative_binary_cmp0): Likewise.
6823 (noncommutative_binary_cmp0_noout): Likewise.
6824 (noncommutative_binary_comparison_result_used): Removed.
6825 (rsub_cmp0): New pattern.
6826 (rsub_cmp0_noout): Likewise.
6827 (extzvsi): Changed, keep only meaningful variants.
6828 (SQH, SEZ): New iterators.
6829 (SQH_postfix): New mode attribute.
6830 (SEZ_prefix): New code attribute.
6831 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6832 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6833 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6835 (noncommutative_operator): Check the availability of barrel
6838 2018-04-09 Richard Biener <rguenther@suse.de>
6840 PR tree-optimization/85284
6841 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6842 Only use the niter constraining form of simple_iv when the exit
6845 2018-04-09 Tom de Vries <tom@codesourcery.com>
6848 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6849 (define_expand "*memory_barrier"): New define_expand.
6850 (define_insn "memory_barrier"): New insn.
6852 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6854 PR rtl-optimization/80463
6855 PR rtl-optimization/83972
6856 PR rtl-optimization/83480
6858 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6859 correct producer for the insn.
6860 (tidy_control_flow): Fixup seqnos in case of debug insns.
6862 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6864 PR rtl-optimization/83913
6866 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6867 different sched-times when merging exprs.
6869 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6871 PR rtl-optimization/83962
6873 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6874 tidy_fallthru_edge and tidy_control_flow.
6876 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6878 PR rtl-optimization/83530
6880 * sel-sched.c (force_next_insn): New global variable.
6881 (remove_insn_for_debug): When force_next_insn is true, also leave only
6882 next insn in the ready list.
6883 (sel_sched_region): When the region wasn't scheduled, make another pass
6884 over it with force_next_insn set to 1.
6886 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6888 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6890 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6891 for interrupt control.
6892 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6893 functions for interrupt control.
6894 * config/nds32/nds32-intrinsic.md: Likewise.
6895 * config/nds32/nds32_intrinsic.h: Likewise.
6896 * config/nds32/nds32.h (nds32_builtins): Likewise.
6898 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6900 * config/nds32/nds32.c (nds32_init_machine_status,
6901 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6902 strict_aligned_p field.
6903 (nds32_expand_to_rtl_hook): New function.
6904 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6905 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6907 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6908 Chung-Ju Wu <jasonwucj@gmail.com>
6910 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6911 * config/nds32/nds32-n7.md: New file.
6912 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6913 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6915 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6916 * config/nds32/nds32.md (pipeline_model): Add n7.
6917 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6918 * config/nds32/pipelines.md: Include n7 settings.
6920 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6921 Chung-Ju Wu <jasonwucj@gmail.com>
6923 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6924 * config/nds32/nds32-e8.md: New file.
6925 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6926 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6928 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6929 * config/nds32/nds32.md (pipeline_model): Add e8.
6930 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6931 * config/nds32/pipelines.md: Include e8 settings.
6933 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6934 Chung-Ju Wu <jasonwucj@gmail.com>
6936 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6937 * config/nds32/nds32-n8.md: New file.
6938 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6939 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6941 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6942 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6943 * config/nds32/nds32.md (pipeline_model): Add n8.
6944 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6945 * config/nds32/pipelines.md: Include n8 settings.
6947 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6948 Chung-Ju Wu <jasonwucj@gmail.com>
6950 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6951 * config/nds32/nds32-n9-2r1w.md: New file.
6952 * config/nds32/nds32-n9-3r2w.md: New file.
6953 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6954 nds32_register_ports): New or modify for cpu n9.
6955 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6957 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6958 * config/nds32/nds32-utils.c: New file.
6959 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6960 TARGET_MUL_SLOW): Define.
6961 * config/nds32/nds32.md (pipeline_model): New attribute.
6962 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6963 New options that support cpu n9.
6964 * config/nds32/pipelines.md: Include n9 settings.
6965 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6967 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6969 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6970 information if necessary.
6971 (output_cond_branch_compare_zero): Likewise.
6972 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6973 (nds32_target_alignment): Refine for alignment.
6974 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6975 (FUNCTION_BOUNDARY): Modify.
6976 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6978 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6980 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6982 * config/nds32/constants.md (unspec_volatile_element): Add values for
6983 TLB operation and data prefetch.
6984 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6985 functions for TLB operation and data prefetch.
6986 * config/nds32/nds32-intrinsic.md: Likewise.
6987 * config/nds32/nds32_intrinsic.h: Likewise.
6988 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6989 (nds32_print_operand): Likewise.
6990 * config/nds32/nds32.h (nds32_builtins): Likewise.
6992 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6993 Andrew Pinski <pinsika@gcc.gnu.org>
6996 * match.pd: Use constant_boolean_node of correct type instead of
6997 boolean_true_node or boolean_false_node for simplifying
6998 pointer comparisons to zero.
7000 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7002 PR tree-optimization/80021
7003 * tree.c (verify_type_variant): Make error call in verify_variant_match
7004 translatable and remove final full stop.
7006 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7008 * config/nds32/constants.md (unspec_volatile_element): Add
7009 UNSPEC_VOLATILE_EH_RETURN.
7010 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7011 nds32_output_stack_pop): Support dwarf exception handling process.
7012 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7013 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7014 exception handling process.
7015 (nds32_compute_stack_frame): Likewise.
7016 (nds32_return_addr_rtx): Likewise.
7017 (nds32_initial_elimination_offset): Likewise.
7018 (nds32_expand_prologue): Likewise.
7019 (nds32_expand_epilogue): Likewise.
7020 (nds32_dynamic_chain_address): New function.
7021 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7023 (DYNAMIC_CHAIN_ADDRESS): Define.
7024 (EH_RETURN_DATA_REGNO): Define.
7025 (EH_RETURN_STACKADJ_RTX): Define.
7026 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7027 patterns for dwarf exception handling.
7029 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7031 * config/nds32/nds32.h: Clean up obsolete macros.
7033 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7035 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7036 Add enum values for particular instructions.
7037 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7038 particular intrinsic functions.
7039 * config/nds32/nds32-intrinsic.md: Likewise.
7040 * config/nds32/nds32_intrinsic.h: Likewise.
7041 * config/nds32/nds32.h (nds32_builtins): Likewise.
7042 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7043 (btst, ave): New patterns for particular instructions.
7045 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7047 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7048 Add enum values for atomic load/store and memory sync.
7049 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7051 * config/nds32/nds32-intrinsic.md: Likewise.
7052 * config/nds32/nds32_intrinsic.h: Likewise.
7053 * config/nds32/nds32.h (nds32_builtins): Likewise.
7055 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7057 PR tree-optimization/85257
7058 * fold-const.c (native_encode_vector): If not all elts could fit
7059 and off is -1, return 0 rather than offset.
7060 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7061 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7062 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7063 adjust buffer in native_interpret_expr call.
7065 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7067 * config/nds32/constants.md (unspec_volatile_element): Add cache
7068 control enum values.
7069 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7070 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7071 * config/nds32/nds32.c (nds32_cctl_names): New.
7072 (nds32_print_operand): Handle cache control register names.
7073 * config/nds32/nds32.h (nds32_builtins): New enum values.
7074 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7076 * config/nds32/nds32.md (type): Add mmu.
7077 * config/nds32/pipelines.md (simple_insn): Add mmu.
7079 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7081 * config/nds32/nds32.md (type): Remove call.
7082 * config/nds32/pipelines.md (simple_insn): Likewise.
7084 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7086 * config/nds32/constants.md (unspec_volatile_element): Add
7087 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7088 UNSPEC_VOLATILE_FMFCFG.
7089 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7090 description for fmfcfg and fmfcsr.
7091 (bdesc_1arg): Add fmtcsr.
7092 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7093 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7094 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7095 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7096 unspec_fmfcfg): New patterns.
7097 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7098 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7099 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7100 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7101 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7102 __nds32__fmfcfg): Define.
7104 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7106 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7107 intrinsic register names.
7108 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7109 intrinsic register enum values and macros.
7111 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7113 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7114 for load/store addressing form.
7115 (nds32_print_operand_address): Likewise.
7117 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7120 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7121 based on LABEL_REF. Remove useless assertion.
7122 (pic_address_needs_scratch): Fix formatting.
7123 (sparc_legitimize_pic_address): Minor tweaks.
7124 (sparc_delegitimize_address): Adjust assertion accordingly.
7125 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7126 into symbolic_operand.
7127 (movsi_high_pic_label_ref): Likewise.
7128 (movsi_lo_sum_pic_label_ref): Likewise.
7129 (movdi_pic_label_ref): Likewise.
7130 (movdi_high_pic_label_ref): Likewise.
7131 (movdi_lo_sum_pic_label_ref): Likewise.
7133 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7135 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7136 custom LIB_SPEC setup.
7138 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7139 Kito Cheng <kito.cheng@gmail.com>
7141 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7142 * config/riscv/freebsd.h: New.
7144 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7146 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7147 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7150 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7151 Kito Cheng <kito.cheng@gmail.com>
7153 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7154 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7155 * config/nds32/nds32-protos.h (nds32_output_call,
7156 nds32_output_return): Declare.
7157 * config/nds32/nds32.md: Refine all the call and return patterns.
7159 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7162 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7163 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7165 PR rtl-optimization/84872
7166 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7167 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7170 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7172 * expr.c (copy_blkmode_to_reg): Revert 254862.
7173 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7175 2018-04-06 Richard Biener <rguenther@suse.de>
7178 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7179 after seeing a component reference with an adjacent field. Treat
7180 refs to arrays at struct end of external decls similar to
7181 refs to unconstrained commons.
7183 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7186 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7187 look through SAVE_EXPRs with non-side-effects argument. Adjust
7189 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7192 2018-04-06 Richard Biener <rguenther@suse.de>
7195 * alias.c (find_base_term): New wrapper around find_base_term
7196 unwinding CSELIB_VAL_PTR changes.
7197 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7200 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7202 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7204 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7205 constant definitions.
7206 ("nop"): lr 0,0 -> nopr r0
7207 ("nop_lr0", "nop_lr1"): New insn definitions.
7209 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7211 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7212 NDS32_V3PUSH_AVAILABLE_P macro.
7214 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7215 Chung-Ju Wu <jasonwucj@gmail.com>
7217 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7218 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7219 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7220 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7221 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7222 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7223 * config/nds32/constraints.md: New constraints and checking for hard
7224 float configuration.
7225 * config/nds32/iterators.md: New mode iterator and attribute for hard
7226 float configuration.
7227 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7229 * config/nds32/nds32-fpu.md: New file.
7230 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7231 deal with hard float code generation.
7232 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7234 (abi_type, float_reg_number): New enum type.
7235 * config/nds32/nds32-predicates.c: New predicates for hard float.
7236 * config/nds32/nds32-protos.h: Declare functions for hard float.
7237 * config/nds32/nds32.c: Implementation for hard float configuration.
7238 * config/nds32/nds32.h: Definitions for hard float configuration.
7239 * config/nds32/nds32.md: Include hard float machine description and
7240 modify patterns for hard float configuration.
7241 * config/nds32/nds32.opt: New options for hard float configuration.
7242 * config/nds32/predicates.md: New predicates for hard float
7245 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7247 * common/config/nds32/nds32-common.c
7248 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7250 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7253 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7254 CONSTRUCTOR_ELT (ctor, ...)->value.
7256 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7259 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7261 2018-04-05 Tom de Vries <tom@codesourcery.com>
7264 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7267 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7268 Kito Cheng <kito.cheng@gmail.com>
7270 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7271 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7272 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7273 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7275 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7276 Kito Cheng <kito.cheng@gmail.com>
7278 * config/nds32/constraints.md (Ufe): New memory constraint.
7279 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7280 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7281 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7283 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7284 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7286 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7288 * config/nds32/nds32.md: Use optimize_size in the condition for
7289 alu-shift instructions.
7291 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7293 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7295 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7297 * config/nds32/nds32.md (negsi2): Refine pattern.
7299 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7300 Chung-Ju Wu <jasonwucj@gmail.com>
7302 * config/nds32/iterators.md (shift_rotate): New code iterator.
7303 (shift): New code attribute.
7304 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7305 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7306 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7307 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7308 bit-wise operations.
7309 (andsi3, *andsi3): Ditto.
7310 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7311 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7312 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7313 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7314 nds32_ior_operand, nds32_xor_operand): New predicates.
7316 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7318 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7319 (addsi3, subsi3): ... this.
7321 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7323 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7325 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7327 * config/nds32/nds32.md: Adjust indention.
7329 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7331 * config/nds32/nds32.md (feature): New attribute.
7333 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7335 * config/nds32/nds32.md (subtype): New attribute.
7337 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7340 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7341 expansion to perform a bitwise AND of the argument followed by a
7342 boolean negation of the result.
7344 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7346 PR rtl-optimization/84878
7347 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7348 the basic block. Assert the use reference is not artificial and that
7349 it has an associated insn.
7351 2018-04-04 Michael Matz <matz@suse.de>
7353 * builtins.c (compute_objsize): Pass correct operand
7354 to array_at_struct_end_p.
7356 2018-04-04 Richard Biener <rguenther@suse.de>
7359 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7360 from contexts for DINFO_LEVEL_TERSE and below.
7362 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7364 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7365 resiter_operand condition.
7366 * config/nds32/nds32.md (*move<mode>): Ditto.
7368 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7369 Monk Chiang <sh.chiang04@gmail.com>
7371 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7373 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7375 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7377 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7378 Kito Cheng <kito.cheng@gmail.com>
7380 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7381 nds32_cond_code_str, output_cond_branch,
7382 output_cond_branch_compare_zero, nds32_expand_cbranch,
7383 nds32_expand_cstore, nds32_expand_movcc,
7384 nds32_output_cbranchsi4_equality_zero,
7385 nds32_output_cbranchsi4_equality_reg,
7386 nds32_output_cbranchsi4_equality_reg_or_const_int,
7387 nds32_output_cbranchsi4_greater_less_zero: New functions.
7388 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7389 nds32_expand_cstore, nds32_expand_movcc,
7390 nds32_output_cbranchsi4_equality_zero,
7391 nds32_output_cbranchsi4_equality_reg,
7392 nds32_output_cbranchsi4_equality_reg_or_const_int,
7393 nds32_output_cbranchsi4_greater_less_zero): Declare.
7394 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7395 nds32_rimm11s_operand): New predicates.
7396 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7397 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7400 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7402 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7403 * config/nds32/nds32.md: Ditto.
7404 * config/nds32/pipelines.md: Ditto.
7406 2018-04-04 Richard Biener <rguenther@suse.de>
7408 PR tree-optimization/85168
7409 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7410 propagating abnormals.
7412 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7414 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7416 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7417 Kito Cheng <kito.cheng@gmail.com>
7419 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7420 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7421 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7422 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7423 * config/nds32/nds32.md (sibcall_internal): New.
7424 (sibcall_register): Remove.
7425 (sibcall_immediate): Remove.
7426 (sibcall_value_internal): New.
7427 (sibcall_value_register): Remove.
7428 (sibcall_value_immediate): Remove.
7429 * config/nds32/predicates.md (nds32_general_register_operand): New.
7430 (nds32_call_address_operand): New.
7432 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7434 PR rtl-optimization/85167
7435 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7436 bb_defs if *split_p, instead preinitialize it to NULL.
7438 PR tree-optimization/85156
7439 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7440 evaluating the argument multiple times.
7442 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7444 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7446 (_mm_cvtpd_ps): Likewise.
7447 (_mm_cvttpd_epi32): Likewise.
7448 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7449 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7450 vector, pixel, and bool following altivec.h include.
7452 2018-04-03 Martin Sebor <msebor@redhat.com>
7454 * doc/extend.texi (Common Function Attributes): Clarify.
7455 (const attribute): Likewise.
7456 (pure attribute): Likewise.
7458 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7461 * config/i386/i386.c (ix86_expand_vector_set): Use
7462 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7464 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7466 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7467 instructions when changing rounding bits to preserve precision bits
7468 in the x87 control word.
7470 2018-04-03 Martin Liska <mliska@suse.cz>
7472 PR tree-optimization/82491
7473 * rtl.h (strip_offset_and_add): Replace += suboffset with
7474 poly_uint64 () + suboffset.
7476 2018-03-29 Martin Liska <mliska@suse.cz>
7477 Martin Jambor <mjambor@suse.cz>
7480 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7481 param_type is not an integral or pointer type.
7483 2018-04-03 Richard Biener <rguenther@suse.de>
7485 * sese.h (recompute_all_dominators): Remove.
7487 2018-04-02 Martin Sebor <msebor@redhat.com>
7489 * doc/invoke.texi (-Wrestrict): Fix typos.
7491 2018-04-02 Jim Wilson <jimw@sifive.com>
7493 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7494 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7495 (<optab>di3, <optab>si3_extend): Likewise.
7496 (<optab>si3_mask, <optab>si3_mask_1): New.
7497 (<optab>di3_mask, <optab>di3_mask_1): New.
7498 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7499 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7500 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7502 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7504 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7507 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7509 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7510 (nds32_canonicalize_comparison): New function.
7512 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7513 Kito Cheng <kito.cheng@gmail.com>
7514 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7516 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7517 * config/nds32/constants.md (unspec_volatile_element): Add
7518 UNSPEC_VOLATILE_RELAX_GROUP.
7519 * config/nds32/nds32-relax-opt.c: New file.
7520 * config/nds32/nds32-predicates.c
7521 (nds32_symbol_load_store_p): New function.
7522 * config/nds32/nds32-protos.h
7523 (nds32_symbol_load_store_p): Declare function.
7524 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7525 * config/nds32/nds32.c
7526 (nds32_register_pass): New function to register pass.
7527 (nds32_register_passes): New function to register passes.
7528 * config/nds32/nds32.md (relax_group): New pattern.
7529 * config/nds32/nds32.opt (mrelax-hint): New option.
7530 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7532 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7534 * config/nds32/t-nds32: Modify files dependency.
7536 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7538 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7539 (PROFILE_HOOK): Define its implementation.
7541 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7543 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7544 type and 32-bit size.
7546 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7549 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7550 (V_128_256): New mode iterator.
7551 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7552 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7553 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7555 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7556 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7558 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7561 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7562 NaN inputs correctly.
7564 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7567 * config/rs6000/vsx.md (??r): New mode attribute.
7568 (*vsx_mov<mode>_64bit): Use it.
7569 (*vsx_mov<mode>_32bit): Likewise.
7571 2018-03-30 Martin Sebor <msebor@redhat.com>
7573 PR tree-optimization/84818
7574 * builtins.c (check_access): Use warning_n.
7576 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7579 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7581 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7584 2018-03-30 Julia Koval <julia.koval@intel.com>
7587 * x86-tune.def (movx, partial_reg_dependency): Enable for
7590 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7593 * lra-constraints.c (process_alt_operands): Move setting
7594 this_alternative_matches below.
7596 2018-03-29 Martin Liska <mliska@suse.cz>
7599 * doc/invoke.texi: Document how LTO works with debug info.
7600 Describe auto-load support of binutils. Mention 'x86-64'
7601 as valid option value of -march option.
7603 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7605 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7608 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7609 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7610 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7613 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7616 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7617 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7618 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7619 for RS6000_BTM_POWERPC64.
7620 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7621 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7622 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7627 2018-03-28 Carl Love <cel@us.ibm.com>
7630 2017-09-27 Carl Love <cel@us.ibm.com>
7632 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7633 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7634 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7637 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7639 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7640 instead of __vector bool.
7641 (_mm_max_pu8): Likewise.
7642 (_mm_min_pi16): Likewise.
7644 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7647 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7648 (DIVWEUO): Likewise.
7650 (DIVDEUO): Likewise.
7651 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7652 DIVWEUO and DIVDEUO.
7653 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7654 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7655 (div_extend): Likewise.
7656 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7658 (__builtin_divweuo): Likewise.
7659 (__builtin_divdeo): Likewise.
7660 (__builtin_divdeuo): Likewise.
7662 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7665 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7666 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7668 PR tree-optimization/82004
7669 * gimple-match-head.c (optimize_pow_to_exp): New function.
7670 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7671 Don't fold to exp if optimize_pow_to_exp is false.
7673 2018-03-28 Martin Liska <mliska@suse.cz>
7676 * calls.c (initialize_argument_information): Fix trailing space.
7677 * common.opt: Fix typo and provide better explanation for
7678 -fsanitize-coverage option.
7679 * config/i386/i386.opt: Fix typo.
7681 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7682 Martin Liska <mliska@suse.cz>
7685 * gimplify.c (asan_poison_variable): Don't do the check for
7686 gimplify_omp_ctxp here.
7687 (gimplify_decl_expr): Do it here.
7688 (gimplify_target_expr): Likewise.
7690 2018-03-28 Martin Liska <mliska@suse.cz>
7693 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7694 chkp_type_bounds_count if MPX is not enabled.
7696 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7698 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7700 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7703 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7704 function to create the function decl for complex long double
7705 multiply and divide for -mabi=ieeelongdouble.
7706 (init_float128_ieee): Call it.
7708 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7711 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7712 -fcf-protection=branch -mibt.
7713 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7715 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7718 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7719 arm_word_relocations.
7721 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7724 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7725 extern array declarations.
7727 2018-03-27 Richard Biener <rguenther@suse.de>
7730 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7731 explicit single_use checks.
7733 2018-03-27 Richard Biener <rguenther@suse.de>
7735 PR tree-optimization/85082
7736 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7739 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7741 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7742 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7743 Turn on fasynchronous-unwind-tables and funwind-tables.
7745 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7748 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7749 (*bmi_blsr_<mode>_ccz): Ditto.
7751 2018-03-26 Tom de Vries <tom@codesourcery.com>
7753 PR tree-optimization/85063
7754 * omp-general.c (offloading_function_p): New function. Factor out
7756 * omp-offload.c (pass_omp_target_link::gate): ... here.
7757 * omp-general.h (offloading_function_p): Declare.
7758 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7759 with attribute omp declare target for offloading functions.
7761 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7763 PR tree-optimization/84005
7764 * tree-data-ref.h (get_base_for_alignment): Declare.
7765 * tree-data-ref.c (get_base_for_alignment_1): New function.
7766 (get_base_for_alignment): Likewise.
7767 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7768 get_base_for_alignment to find a suitable base object, instead
7769 of always using drb->base_address.
7771 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7774 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7775 known size by default.
7777 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7780 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7781 and non BLKmode operands.
7783 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7786 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7787 Clean up attributes.
7789 2018-03-23 Richard Biener <rguenther@suse.de>
7792 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7793 we are going to emit early debug for LTO.
7795 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7798 * function.c (match_asm_constraints_1): Don't optimize if input
7799 doesn't satisfy general_operand predicate for output's mode.
7802 * alias.c (write_dependence_p): Don't require for x_canonicalized
7803 non-VOIDmode if x has VOIDmode.
7806 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7807 just don't try to optimize it rather than assert it never happens.
7809 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7811 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7812 macro expansions for definition of ST_INTERNAL_<mode> and
7813 LD_INTERNAL_<mode> builtins.
7814 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7816 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7818 (altivec_expand_st_builtin): Likewise.
7819 (altivec_expand_builtin): Remove calls to deleted functions.
7820 (rs6000_address_for_altivec): Delete this function.
7821 * config/rs6000/vector.md: Remove expands for
7822 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7824 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7827 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7828 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7829 re-computing once computed.
7830 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7831 (arm_init_machine_status): Initialize
7832 machine->static_chain_stack_bytes.
7834 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7837 * doc/extend.texi: Add four new prototypes for vec_ld.
7838 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7839 definitions for more logical presentation.
7840 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7841 entries for V1TI variants of __builtin_altivec_ld builtin.
7842 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7843 handling of V1TI variant of LVX icode pattern.
7844 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7845 (rs6000_gimple_fold_builtin): Likewise.
7846 (altivec_init_builtins): Add code to define
7847 __builtin_altivec_lvx_v1ti function.
7849 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7852 * function.c (match_asm_constraints_1): Don't do the optimization
7853 if input isn't a REG, SUBREG, MEM or constant.
7855 2018-03-22 Tom de Vries <tom@codesourcery.com>
7857 PR tree-optimization/84956
7858 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7859 bb_has_abnormal_pred.
7861 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7864 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7865 DECL_INITIAL (decl) to decl at the end.
7866 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7869 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7871 * doc/extend.texi (__builtin_tgmath): Document when complex
7872 integer types are treated as _Complex _Float64.
7874 2018-03-21 Tom de Vries <tom@codesourcery.com>
7876 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7878 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7880 PR tree-optimization/84960
7881 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7882 if it is ENTRY block, move them into single succ of ENTRY in that case.
7884 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7886 PR tree-optimization/84811
7887 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7888 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7889 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7890 (known_subrange_p): Update accordingly. Cast each value involved
7891 in the size comparison, rather than casting the result of the
7894 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7896 PR tree-optimization/84982
7897 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7898 by flipping the least significant bit rather than all bits from
7899 bitpos to bitpos + bitsize - 1.
7901 2018-03-21 Nathan Sidwell <nathan@acm.org>
7903 * doc/extend.texi (Deprecated Features): Remove mention of
7904 long-deleted deprecations.
7906 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7909 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7910 * configure: Regenerate.
7912 2018-03-21 Tom de Vries <tom@codesourcery.com>
7914 PR tree-optimization/83126
7915 * tree-parloops.c (num_phis): New function.
7916 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7918 2018-03-21 Nathan Sidwell <nathan@acm.org>
7920 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7921 mention anon-struct/union members and trailing attributes.
7923 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7925 PR tree-optimization/84969
7926 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7927 builtin memset partitions if they set different rhs values.
7929 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7931 PR rtl-optimization/84989
7932 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7933 VEC_DUPLICATE with scalar result mode.
7935 2018-03-21 Martin Liska <mliska@suse.cz>
7938 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7939 not intended return statement.
7941 2018-03-21 Martin Liska <mliska@suse.cz>
7944 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7945 (chkp_find_bound_slots_1): Limit number of iterations.
7947 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7950 * Minor grammar fixes for x86 options.
7952 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7955 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7956 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7958 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7961 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7962 (altivec_lvx_<mode>_1op): Likewise.
7963 (altivec_stvx_<mode>_2op): Likewise.
7964 (altivec_stvx_<mode>_1op): Likewise.
7965 (altivec_lvx_<VM2:mode>): New define_expand.
7966 (altivec_stvx_<VM2:mode>): Likewise.
7967 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7968 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7969 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7970 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7971 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7972 (rs6000_gen_lvx): Likewise.
7973 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7974 (altivec_expand_stv_builtin): Likewise.
7975 (altivec_expand_builtin): Likewise.
7976 * config/rs6000/vector.md: Likewise.
7978 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7981 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7984 2018-03-20 Richard Biener <rguenther@suse.de>
7987 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7988 sign-conversions as zero, fall back to standard scalar_stmt
7991 2018-03-20 Martin Liska <mliska@suse.cz>
7994 * predict.c (rebuild_frequencies): Handle case when we have
7995 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7997 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8000 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8001 flag_section_anchors.
8002 * varasm.c (use_blocks_for_decl_p): Remove hack for
8003 dw2_force_const_mem.
8006 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8008 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8009 be created, use lowpart_subreg of operands[0] rather than operands[0]
8011 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8012 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8013 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8014 and n constraint instead of aarch64_shift_imm_di and Usd.
8015 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8016 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8018 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8021 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8022 to favor GPR over NEON registers.
8023 (<shift>di3_neon): Likewise.
8025 2018-03-20 Tom de Vries <tom@codesourcery.com>
8028 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8029 (nvptx_process_pars): Emit bar.sync asap and alap.
8031 2018-03-20 Tom de Vries <tom@codesourcery.com>
8034 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8035 seen_label if seen_label is already set.
8037 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8040 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8041 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8042 Use 1U instead of 1. Formatting fixes.
8045 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8046 instead of TREE_TYPE (s1) for the return value.
8048 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8050 PR tree-optimization/84946
8051 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8052 bitsize + bitsize in poly_uint64 rather than poly_int64.
8055 * dwarf2asm.c: Include fold-const.c.
8056 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8057 of decl rather than decl itself.
8059 PR rtl-optimization/84643
8060 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8062 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8065 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8066 calling assemble_variable.
8068 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8071 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8072 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8074 2018-03-19 Jim Wilson <jimw@sifive.com>
8077 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8078 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8079 (riscv_first_stack_step): Likewise.
8080 (riscv_option_override): Use STACK_BOUNDARY instead of
8082 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8084 (BIGGEST_ALIGNMENT): Set to 128.
8085 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8086 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8089 2018-03-19 Richard Biener <rguenther@suse.de>
8091 PR tree-optimization/84933
8092 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8093 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8095 2018-03-19 Richard Biener <rguenther@suse.de>
8097 PR tree-optimization/84859
8098 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8099 (cond_if_else_store_replacement): Perform sinking operation on
8100 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8101 Generalize what a BB with a single eligible store is.
8103 2018-03-19 Richard Biener <rguenther@suse.de>
8105 PR tree-optimization/84929
8106 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8107 chrec_is_positive against non-chrec arg.
8109 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8112 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8114 2018-03-18 Martin Liska <mliska@suse.cz>
8116 PR rtl-optimization/84635
8117 * regrename.c (build_def_use): Use matches_mode only when
8120 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8122 PR tree-optimization/84913
8123 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8124 vectorize chains of COND_EXPRs.
8126 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8128 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8130 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8132 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8134 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8136 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8138 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8139 Kito Cheng <kito.cheng@gmail.com>
8141 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8142 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8143 (nds32_adjust_reg_alloc_order): New function.
8144 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8146 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8148 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8149 nds32_print_operand, nds32_print_operand_address): Use
8150 HOST_WIDE_INT_PRINT_DEC instead.
8152 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8154 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8156 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8159 * config/i386/i386.c (initial_ix86_tune_features,
8160 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8162 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8163 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8164 rather than 1u << ix86_tune. Formatting fix.
8165 (ix86_option_override_internal): Change ix86_arch_mask from
8166 unsigned int to unsigned HOST_WIDE_INT, initialize to
8167 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8168 (ix86_function_specific_restore): Likewise.
8170 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8173 * postreload.c (reload_combine_recognize_pattern): Perform
8174 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8175 truncate_int_for_mode the result for the destination's mode.
8178 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8179 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8181 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8184 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8187 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8188 regno_allocno_class_array and sorted_pseudos.
8189 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8190 insns where regno is used.
8192 2018-03-16 Martin Liska <mliska@suse.cz>
8195 * multiple_target.c (create_dispatcher_calls): Redirect
8196 reference in the symbol table.
8198 2018-03-16 Martin Liska <mliska@suse.cz>
8201 * multiple_target.c (create_dispatcher_calls): Redirect also
8204 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8208 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8209 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8210 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8212 2018-03-16 Julia Koval <julia.koval@intel.com>
8214 * doc/invoke.texi (Skylake Server): Add CLWB.
8215 Cannonlake): Remove CLWB.
8217 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8219 PR tree-optimization/84841
8220 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8222 (FLOAT_ONE_CONST_TYPE): Define.
8223 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8224 (sort_by_operand_rank): Put entries with higher constant_type last
8225 rather than first to match comments.
8227 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8229 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8232 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8235 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8236 check and use address of target always.
8238 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8241 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8242 (indirect_thunk_bnd_needed): Likewise.
8243 (indirect_thunks_used): Likewise.
8244 (indirect_thunks_bnd_used): Likewise.
8245 (indirect_return_needed): New.
8246 (indirect_return_bnd_needed): Likewise.
8247 (output_indirect_thunk_function): Add a bool argument for
8249 (output_indirect_thunk_function): Don't generate alias for
8250 function return thunk.
8251 (ix86_code_end): Call output_indirect_thunk_function to generate
8252 function return thunks.
8253 (ix86_output_function_return): Set indirect_return_bnd_needed
8254 and indirect_return_needed instead of indirect_thunk_bnd_needed
8255 and indirect_thunk_needed.
8257 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8259 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8260 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8261 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8263 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8264 Paul Hua <paul.hua.gm@gmail.com>
8267 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8269 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8271 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8272 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8275 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8278 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8279 instead of GET_MODE_SIZE when comparing Units.
8281 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8284 * varasm.c (hash_section): Return an unchangeble hash value
8285 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8286 Return !aarch64_can_use_per_function_literal_pools_p ().
8288 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8291 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8292 rather than address of cmode as last argument to prepare_cmp_insn.
8294 2018-03-15 Julia Koval <julia.koval@intel.com>
8296 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8297 F_AVX512VNNI, F_AVX512BITALG): New.
8299 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8302 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8303 insn for floating-point loads and stores.
8305 2018-03-14 Carl Love <cel@us.ibm.com>
8307 * config/rs6000/rs6000-c.c: Add macro definitions for
8308 ALTIVEC_BUILTIN_VEC_PERMXOR.
8309 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8310 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8311 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8312 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8314 * config/doc/extend.texi: Add prototypes for vec_permxor.
8316 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8319 * diagnostic-show-locus.c (class layout_point): Convert m_line
8320 from int to linenum_type.
8321 (line_span::comparator): Use linenum "compare" function when
8322 comparing line numbers.
8323 (test_line_span): New function.
8324 (layout_range::contains_point): Convert param "row" from int to
8326 (layout_range::intersects_line_p): Likewise.
8327 (layout::will_show_line_p): Likewise.
8328 (layout::print_source_line): Likewise.
8329 (layout::should_print_annotation_line_p): Likewise.
8330 (layout::print_annotation_line): Likewise.
8331 (layout::print_leading_fixits): Likewise.
8332 (layout::annotation_line_showed_range_p): Likewise.
8333 (struct line_corrections): Likewise for field m_row.
8334 (line_corrections::line_corrections): Likewise for param "row".
8335 (layout::print_trailing_fixits): Likewise.
8336 (layout::get_state_at_point): Likewise.
8337 (layout::get_x_bound_for_row): Likewise.
8338 (layout::print_line): Likewise.
8339 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8340 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8341 * input.c (selftest::test_linenum_comparisons): New function.
8342 (selftest::input_c_tests): Call it.
8343 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8344 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8345 * selftest.h (ASSERT_GT): New macro.
8346 (ASSERT_GT_AT): New macro.
8347 (ASSERT_LT): New macro.
8348 (ASSERT_LT_AT): New macro.
8350 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8352 PR rtl-optimization/84780
8353 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8355 2018-03-14 Martin Liska <mliska@suse.cz>
8357 * tree.c (record_node_allocation_statistics): Use
8358 get_stats_node_kind.
8359 (get_stats_node_kind): New function extracted from
8360 record_node_allocation_statistics.
8361 (free_node): Use get_stats_node_kind.
8363 2018-03-14 Richard Biener <rguenther@suse.de>
8365 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8366 that the value-set of ANTIC_IN doesn't grow.
8369 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8371 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8372 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8374 2018-03-14 Julia Koval <julia.koval@intel.com>
8376 * config.gcc (icelake-client, icelake-server): New.
8378 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8379 (initial_ix86_arch_features): Ditto.
8380 (PTA_SKYLAKE): Add SGX.
8381 (PTA_ICELAKE): Remove.
8382 (PTA_ICELAKE_CLIENT): New.
8383 (PTA_ICELAKE_SERVER): New.
8384 (ix86_option_override_internal): Split up icelake on icelake client and
8386 (get_builtin_code_for_version): Ditto.
8387 (fold_builtin_cpu): Ditto.
8388 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8389 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8390 * config/i386/i386.h (processor_type): Ditto.
8391 * doc/invoke.texi: Ditto.
8393 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8396 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8397 INTEGER_CST offset, add it together with bitpos / 8 and
8398 sign extend based on POINTER_SIZE.
8402 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8405 * config/i386/constraints.md (Yc): New register constraint.
8406 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8407 Use Yc constraint for alternative 2 of operand 0. Remove
8408 preferred_for_speed attribute.
8410 2018-03-14 Richard Biener <rguenther@suse.de>
8412 PR tree-optimization/84830
8413 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8414 with the old one to avoid oscillations.
8416 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8419 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8421 (assign_by_spills): Return a flag of reload assignment failure.
8422 Do not process the reload assignment failures. Do not spill other
8423 reload pseudos if they has the same reg class. Update n if
8425 (lra_assign): Add a return arg. Set up from the result of
8426 assign_by_spills call.
8427 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8428 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8429 usage_insns if it is not NULL.
8430 (spill_hard_reg_in_range): New function.
8431 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8432 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8433 function prototypes.
8434 (lra_assign): Change prototype.
8435 * lra.c (lra): Add code to deal with fails by splitting hard reg
8438 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8440 * config/riscv/riscv.opt (mrelax): New option.
8441 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8442 "norelax" when riscv_mrelax is disabled.
8443 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8445 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8448 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8449 reassociation for int modes.
8451 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8453 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8454 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8456 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8457 * config/aarch64/aarch64-sve.md
8458 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8459 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8460 (*extend<mode><Vwide>2): Rename to...
8461 (aarch64_sve_extend<mode><Vwide>2): ...this.
8462 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8463 renaming the old pattern to...
8464 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8466 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8467 define_expand, renaming the old pattern to...
8468 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8469 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8470 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8471 account when deciding which SVE instruction the optab should use.
8472 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8474 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8476 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8477 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8478 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8479 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8480 (tlsdesc_small_<mode>): Turn a define_expand and use
8481 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8482 (tlsdesc_small_advsimd_<mode>): ...this.
8483 (tlsdesc_small_sve_<mode>): New pattern.
8485 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8487 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8488 (UNSPEC_UMUL_HIGHPART): New constants.
8489 (MUL_HIGHPART): New int iteraor.
8490 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8491 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8493 (*<su>mul<mode>3_highpart): New define_insn.
8495 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8498 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8501 2018-03-13 Martin Liska <mliska@suse.cz>
8504 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8506 (sem_item_optimizer::~sem_item_optimizer): Release it.
8507 (sem_item_optimizer::merge_classes): Register variable aliases.
8508 (sem_item_optimizer::fixup_pt_set): New function.
8509 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8510 * ipa-icf.h: Declare new variables and functions.
8512 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8515 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8516 integer_pow2p@2 and test integer_pow2p in condition.
8517 (A < 0 ? C : 0): Similarly for @1.
8520 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8521 characters starting at p contain '\0' character, don't look beyond
8525 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8526 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8529 * reg-stack.c (change_stack): Change update_end var from int to
8530 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8531 also call set_block_for_insn on the newly added insns and rescan.
8534 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8535 on the last operand.
8538 * tree.c (stabilize_reference_1): Return save_expr (e) for
8539 STATEMENT_LIST even if it doesn't have side-effects.
8541 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8543 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8545 2018-03-12 Renlin Li <renlin.li@arm.com>
8547 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8548 aarch64_output_scalar_simd_mov_immediate.
8550 2018-03-12 Martin Sebor <msebor@redhat.com>
8552 PR tree-optimization/83456
8553 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8554 for perfectly overlapping calls to memcpy.
8555 (gimple_fold_builtin_memory_chk): Same.
8556 (gimple_fold_builtin_strcpy): Handle no-warning.
8557 (gimple_fold_builtin_stxcpy_chk): Same.
8558 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8560 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8562 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8563 parameter. Use it for SFmode.
8564 (rs6000_function_arg_advance_1): Adjust.
8565 (rs6000_function_arg): Adjust.
8566 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8568 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8570 PR rtl-optimization/84169
8571 PR rtl-optimization/84780
8572 * combine.c (can_combine_p): Check for a 2-insn combination whether
8573 the destination register is used between the two insns, too.
8575 2018-03-12 Richard Biener <rguenther@suse.de>
8577 PR tree-optimization/84803
8578 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8579 for refs DR analysis didn't process.
8581 2018-03-12 Richard Biener <rguenther@suse.de>
8583 PR tree-optimization/84777
8584 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8585 force-vectorize loops ignore whether we are optimizing for size.
8587 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8589 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8590 (TARGET_MD_ASM_ADJUST): Define.
8592 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8593 Kito Cheng <kito.cheng@gmail.com>
8594 Chung-Ju Wu <jasonwucj@gmail.com>
8596 * config/nds32/nds32.c (nds32_compute_stack_frame,
8597 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8598 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8599 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8600 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8601 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8602 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8603 * config/nds32/nds32.md (prologue, epilogue): Use macro
8604 NDS32_V3PUSH_AVAILABLE_P to do checking.
8606 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8609 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8610 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8611 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8612 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8613 addition of most attributes on !orig_type_die or the attribute not
8614 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8616 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8617 Chung-Ju Wu <jasonwucj@gmail.com>
8619 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8621 * config/nds32/nds32.opt (mvh): New option.
8623 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8624 Chung-Ju Wu <jasonwucj@gmail.com>
8626 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8628 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8629 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8632 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8633 Chung-Ju Wu <jasonwucj@gmail.com>
8635 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8637 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8638 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8640 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8641 Kito Cheng <kito.cheng@gmail.com>
8642 Chung-Ju Wu <jasonwucj@gmail.com>
8644 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8645 UNSPEC_FFMISM and UNSPEC_FLMISM.
8646 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8647 for ffb, ffmism and flmism.
8648 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8649 (unspec_ffmism): Ditto.
8650 (unspec_flmism): Ditto.
8651 (nds32_expand_builtin_impl): Check if string extension is available.
8652 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8653 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8655 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8658 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8661 * lra-assigns.c (assign_by_spills): Return a flag of reload
8662 assignment failure. Do not process the reload assignment
8663 failures. Do not spill other reload pseudos if they has the same
8665 (lra_assign): Add a return arg. Set up from the result of
8666 assign_by_spills call.
8667 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8668 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8669 usage_insns if it is not NULL.
8670 (spill_hard_reg_in_range): New function.
8671 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8672 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8673 function prototypes.
8674 (lra_assign): Change prototype.
8675 * lra.c (lra): Add code to deal with fails by splitting hard reg
8678 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8681 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8683 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8686 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8687 (dw_val_node): Add val_symbolic_view.
8688 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8689 (symview_upper_bound): New.
8690 (new_line_info_table): Initialize symviews_since_reset.
8691 (dwarf2out_source_line): Count symviews_since_reset and set
8692 symview_upper_bound.
8693 (dw_val_equal_p): Handle symview.
8694 (add_AT_symview): New.
8695 (print_dw_val): Handle symview.
8696 (attr_checksum, attr_checksum_ordered): Likewise.
8697 (same_dw_val_p, size_of_die): Likewise.
8698 (value_format, output_die): Likewise.
8699 (add_high_low_attributes): Use add_AT_symview for entry_view.
8700 (dwarf2out_finish): Reset symview_upper_bound, clear
8703 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8706 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8707 Add strict argument and use it.
8708 (rs6000_split_multireg_move): Update for new strict argument.
8709 (mem_operand_gpr): Disallow all non-offsettable addresses.
8710 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8712 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8715 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8716 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8717 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8720 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8721 decl, use remap_type if we want to use the type.
8723 2018-03-09 Martin Sebor <msebor@redhat.com>
8725 PR tree-optimization/84526
8726 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8728 (builtin_access::generic_overlap): Be prepared to handle non-array
8731 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8733 PR rtl-optimization/84682
8734 * lra-constraints.c (process_address_1): Check is_address flag
8735 for address constraints.
8736 (process_alt_operands): Likewise.
8737 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8738 preprocess_constraints.
8739 * recog.h (preprocess_constraints): Add oploc parameter.
8741 * recog.c (preprocess_constraints): Test address_operand for
8742 CT_ADDRESS constraints.
8744 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8747 * lra-assigns.c (assign_by_spills): Return a flag of reload
8748 assignment failure. Do not process the reload assignment
8749 failures. Do not spill other reload pseudos if they has the same
8751 (lra_assign): Add a return arg. Set up from the result of
8752 assign_by_spills call.
8753 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8754 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8755 usage_insns if it is not NULL.
8756 (spill_hard_reg_in_range): New function.
8757 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8758 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8759 function prototypes.
8760 (lra_assign): Change prototype.
8761 * lra.c (lra): Add code to deal with fails by splitting hard reg
8764 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8767 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8768 Accept complain bool parameter. Only emit errors if it is true.
8769 (arm_parse_cpu_option_name): Likewise.
8770 (arm_target_thumb_only): Adjust callers of the above.
8771 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8772 prototype to take a default true bool parameter.
8773 (arm_parse_arch_option_name): Likewise.
8775 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8776 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8780 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8781 * configure: Regenerate.
8782 * configure.ac ("linker --version-script option"): New.
8783 ("linker soname option"): New.
8785 2018-03-09 Richard Biener <rguenther@suse.de>
8787 PR tree-optimization/84775
8788 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8789 immediate uses of predicate stmts and mark them modified.
8792 PR tree-optimization/84178
8793 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8795 (version_loop_for_if_conversion): Delay update_ssa call.
8796 (tree_if_conversion): Delay update_ssa until after predicate
8799 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8802 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8803 when the function accesses prior frames.
8805 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8808 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8809 gen_llsym, otherwise call maybe_gen_llsym.
8812 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8813 has ',' character inside of it.
8815 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8818 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8819 as clobbering CC_REGNUM.
8821 2018-03-08 Richard Biener <rguenther@suse.de>
8824 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8825 (follow_copies_to_constant): Do not follow SSA names registered
8828 2018-03-08 Richard Biener <rguenther@suse.de>
8830 PR tree-optimization/84178
8831 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8833 (version_loop_for_if_conversion): Delay update_ssa call.
8834 (tree_if_conversion): Delay update_ssa until after predicate
8837 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8839 PR tree-optimization/84178
8840 * tree-if-conv.c (release_bb_predicate): Remove the
8841 the assertion that the stmts have NULL use_ops.
8842 Discard the statements, asserting that they haven't
8843 yet been added to a BB.
8845 2018-03-08 Richard Biener <rguenther@suse.de>
8847 PR tree-optimization/84746
8848 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8849 (phi_translate): Pass in destination ANTIC_OUT set.
8850 (phi_translate_1): Likewise. For a simplified result lookup
8851 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8852 (phi_translate_set): Adjust.
8853 (do_pre_regular_insertion): Likewise.
8854 (do_pre_partial_partial_insertion): Likewise.
8856 2018-03-08 Martin Liska <mliska@suse.cz>
8858 PR gcov-profile/84735
8859 * doc/gcov.texi: Document usage of profile files.
8860 * gcov-io.h: Document changes in the format.
8862 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8866 * dwarf2out.c (struct dw_line_info_table): Update comments for
8868 (FORCE_RESET_NEXT_VIEW): New.
8869 (FORCE_RESETTING_VIEW_P): New.
8870 (RESETTING_VIEW_P): Check for -1 too.
8871 (ZERO_VIEW_P): Likewise.
8872 (new_line_info_table): Force-reset next view.
8873 (dwarf2out_begin_function): Likewise.
8874 (dwarf2out_source_line): Simplify zero_view_p initialization.
8875 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8876 view directly. Omit view when omitting .loc at line 0.
8878 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8880 PR tree-optimization/84740
8881 * tree-switch-conversion.c (process_switch): Call build_constructors
8882 only if info.phi_count is non-zero.
8884 PR tree-optimization/84739
8885 * tree-tailcall.c (find_tail_calls): Check call arguments against
8886 DECL_ARGUMENTS (current_function_decl) rather than
8887 DECL_ARGUMENTS (func) when checking for tail recursion.
8889 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8891 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8892 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8893 Volker Reichelt's entry and add entries for people that perform
8894 GCC fuzzy testing and report numerous bugs.
8896 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8899 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8900 readonly data in sdata, if that is disabled.
8901 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8902 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8903 -mreadonly-in-sdata option.
8905 2018-03-07 Martin Sebor <msebor@redhat.com>
8907 PR tree-optimization/84468
8908 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8909 basic block when looking for nul assignment.
8911 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8914 * except.h (output_function_exception_table): Adjust prototype.
8915 * except.c (output_function_exception_table): Remove FNNAME parameter
8916 and add SECTION parameter. Ouput one part of the table at a time.
8917 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8918 the first part of the exception table and emit unwind directives.
8919 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8920 (i386_pe_seh_cold_init): Likewise.
8921 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8922 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8923 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8924 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8925 * config/i386/winnt.c: Include except.h.
8926 (struct seh_frame_state): Add reg_offset, after_prologue and
8927 in_cold_section fields.
8928 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8929 (i386_pe_seh_cold_init): New function.
8930 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8931 to seh->in_cold_section.
8932 (seh_emit_push): Record the offset of the push.
8933 (seh_emit_save): Record the offet of the save.
8934 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8935 Test seh->after_prologue to disregard the epilogue.
8936 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8937 (i386_pe_end_cold_function): New function.
8939 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8942 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8943 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8946 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8948 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8949 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8952 * multiple_target.c: Include tree-inline.h and intl.h.
8953 (expand_target_clones): Diagnose and fail if node->definition and
8954 !tree_versionable_function_p (node->decl).
8956 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8958 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8960 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8961 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8962 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8964 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8967 * combine.c (try_combine): Use reg_or_subregno instead of handling
8968 just paradoxical SUBREGs and REGs.
8970 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8972 * config/arc/arc.c (arc_finalize_pic): Remove function.
8973 (arc_must_save_register): We use single base PIC register, remove
8974 checks to save/restore the PIC register.
8975 (arc_expand_prologue): Likewise.
8976 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8978 (arc_verify_short): Likewise.
8979 (arc_attr_type): Likewise.
8980 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8981 (walk_stores): Likewise.
8982 (arc_address_cost): Make it static.
8983 (arc_verify_short): Likewise.
8984 (branch_dest): Likewise.
8985 (arc_attr_type): Likewise.
8986 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8987 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8988 (arc_final_prescan_insn): Remove inserting the nops due to
8989 hardware hazards. It is done in reorg step.
8990 (insn_length_variant_t): Remove.
8991 (insn_length_parameters_t): Likewise.
8992 (arc_insn_length_parameters): Likewise.
8993 (arc_get_insn_variants): Likewise.
8994 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8996 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8999 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9002 PR tree-optimization/84687
9003 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9005 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9007 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9009 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9010 Rename to ppc_speculation_barrier.
9011 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9012 __builtin_ppc_speculation_barrier.
9014 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9017 * combine.c (combine_simplify_rtx): Don't try to simplify if
9018 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9021 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9023 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9024 to 32 bytes when compiling for POWER9.
9026 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9029 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9030 regparm >= 3 with no arg reg available also for calls with
9031 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9034 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9036 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9038 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9041 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9043 2018-03-05 Richard Biener <rguenther@suse.de>
9045 PR tree-optimization/84486
9046 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9047 When inserting a __builtin_assume_aligned call set the LHS
9048 SSA name alignment info accordingly.
9050 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9052 PR tree-optimization/84114
9053 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9054 Avoid reassociation of FLOAT_MODE addition.
9056 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9058 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9059 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9060 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9061 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9062 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9063 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9064 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9066 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9067 __builtin_ia32_wbinvd): New builtins.
9068 (SPECIAL_ARGS2): New.
9069 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9070 (SPECIAL_ARGS2): New.
9071 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9072 (ix86_valid_target_attribute_inner_p): Ditto.
9073 (ix86_init_mmx_sse_builtins): Add special_args2.
9074 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9075 TARGET_WBNOINVD_P): New.
9076 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9077 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9078 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9079 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9080 * config/i386/pconfigintrin.h: New file.
9081 * config/i386/wbnoinvdintrin.h: Ditto.
9082 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9084 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9086 2018-03-05 Richard Biener <rguenther@suse.de>
9088 PR tree-optimization/84670
9089 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9091 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9092 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9093 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9094 doesn't grow if all successors (recursively) were visited at least
9097 2018-03-05 Richard Biener <rguenther@suse.de>
9099 PR tree-optimization/84650
9100 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9101 if executed in the loop pipeline.
9103 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9105 * doc/configfiles.texi (Configuration Files): Move info about
9106 conditionalizing $target-protos.h to...
9107 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9108 differs from $target-protos.h.
9110 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9111 Chung-Ju Wu <jasonwucj@gmail.com>
9113 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9114 * config/nds32/nds32-multiple.md (setmemsi): Define.
9115 * config/nds32/nds32-memory-manipulation.c
9116 (nds32_gen_dup_4_byte_to_word_value): New.
9117 (emit_setmem_word_loop): New.
9118 (emit_setmem_byte_loop): New.
9119 (nds32_expand_setmem_loop): New.
9120 (nds32_expand_setmem_loop_v3m): New.
9121 (nds32_expand_setmem_unroll): New.
9122 (nds32_expand_setmem): New.
9124 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9125 Chung-Ju Wu <jasonwucj@gmail.com>
9127 * config/nds32/nds32-memory-manipulation.c
9128 (nds32_emit_load_store): New.
9129 (nds32_emit_post_inc_load_store): New.
9130 (nds32_emit_mem_move): New.
9131 (nds32_emit_mem_move_block): New.
9132 (nds32_expand_movmemsi_loop_unknown_size): New.
9133 (nds32_expand_movmemsi_loop_known_size): New.
9134 (nds32_expand_movmemsi_loop): New.
9135 (nds32_expand_movmemsi_unroll): New.
9136 (nds32_expand_movmemqi): Rename ...
9137 (nds32_expand_movmemsi): ... to this.
9138 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9139 (movmemsi): ... to this.
9140 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9141 (nds32_expand_movmemsi): ... to this.
9143 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9144 Monk Chiang <sh.chiang04@gmail.com>
9145 Chung-Ju Wu <jasonwucj@gmail.com>
9147 * config/nds32/nds32-protos.h
9148 (nds32_expand_load_multiple): New arguments.
9149 (nds32_expand_store_multiple): Ditto.
9150 (nds32_valid_multiple_load_store): Rename ...
9151 (nds32_valid_multiple_load_store_p): ... to this.
9152 * config/nds32/nds32-memory-manipulation.c
9153 (nds32_expand_load_multiple): Refine implementation.
9154 (nds32_expand_store_multiple): Ditto.
9155 * config/nds32/nds32-multiple.md
9156 (load_multiple): Update nds32_expand_load_multiple interface.
9157 (store_multiple): Update nds32_expand_store_multiple interface.
9158 * config/nds32/nds32-predicates.c
9159 (nds32_valid_multiple_load_store): Rename ...
9160 (nds32_valid_multiple_load_store_p): ... to this and refine
9162 * config/nds32/predicates.md
9163 (nds32_load_multiple_and_update_address_operation): New predicate.
9164 (nds32_store_multiple_and_update_address_operation): New predicate.
9166 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9167 Chung-Ju Wu <jasonwucj@gmail.com>
9169 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9170 (combo): New attribute.
9171 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9173 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9175 * config/nds32/nds32.opt: Change -mcmodel= default value.
9177 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9178 Monk Chiang <sh.chiang04@gmail.com>
9179 Chung-Ju Wu <jasonwucj@gmail.com>
9181 * config/nds32/constants.md (unspec_element): New enum.
9182 * config/nds32/constraints.md (Umw): New constraint.
9183 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9184 * config/nds32/nds32-intrinsic.md: Likewise.
9185 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9186 (nds32_valid_smw_lwm_base_p): New.
9187 (nds32_output_smw_single_word): New.
9188 (nds32_output_lmw_single_word): New.
9189 (nds32_expand_unaligned_load): New.
9190 (nds32_expand_unaligned_store): New.
9191 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9192 (nds32_output_smw_single_word): Declare.
9193 (nds32_output_lmw_single_word): Declare.
9194 (nds32_expand_unaligned_load): Declare.
9195 (nds32_expand_unaligned_store): Declare.
9196 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9197 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9198 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9199 NDS32_BUILTIN_UASTORE_DW.
9200 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9203 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9204 Kito Cheng <kito.cheng@gmail.com>
9205 Chung-Ju Wu <jasonwucj@gmail.com>
9207 * config/nds32/nds32-intrinsic.c
9208 (nds32_expand_builtin_null_ftype_reg): Delete.
9209 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9210 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9211 (nds32_read_argument): New.
9212 (nds32_legitimize_target): Ditto.
9213 (nds32_legitimize_argument): Ditto.
9214 (nds32_check_constant_argument): Ditto.
9215 (nds32_expand_unop_builtin): Ditto.
9216 (nds32_expand_unopimm_builtin): Ditto.
9217 (nds32_expand_binop_builtin): Ditto.
9218 (nds32_builtin_decl_impl): Ditto.
9219 (builtin_description): Ditto.
9220 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9221 (nds32_init_builtins_impl): Ditto.
9222 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9223 (nds32_builtin_decl): New.
9224 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9225 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9227 2018-03-02 Jeff Law <law@redhat.com>
9229 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9230 (redundant_insn, fill_simple_delay_slots): Likewise.
9231 (fill_slots_from_thread): Likewise.
9232 * resource.c (mark_referenced_resources): Likewise.
9233 (mark_set_resources, find_dead_or_set_registers): Likewise.
9235 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9237 * substring-locations.h (format_warning_va): Formatting fix for
9239 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9241 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9242 * substring-locations.c: Include intl.h.
9243 (format_warning_va): Turned into small wrapper around
9244 format_warning_n_va, renamed to ...
9245 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9246 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9248 (format_warning_at_substring_n): New function.
9249 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9250 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9251 format_warning_at_substring with just a shorter name instead of
9252 const function pointer.
9253 (fmtwarn_n): New function.
9254 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9255 appropriate, get rid of all the fmtstr temporaries, move conditionals
9256 with G_() wrapped string literals directly into fmtwarn arguments,
9257 cast dir.len to (int), formatting fixes.
9259 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9261 * doc/invoke.texi: Remove "Cilk Plus" references.
9263 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9264 Richard Biener <rguenther@suse.de>
9267 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9268 for error or warning attributes if CALL_FROM_THUNK_P is set.
9271 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9274 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9275 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9278 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9279 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9281 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9282 instead of -1U in last predictors element's probability member.
9284 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9287 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9288 arguments if they are comparable.
9290 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9292 PR tree-optimization/84634
9293 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9294 masks and masked_loop_p with a single loop_masks, making sure it's
9295 null for bb vectorization.
9297 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9299 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9300 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9301 loop->force_vectorize to check whether there is no alias.
9303 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9306 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9308 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9310 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9311 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9312 instead of a loop around prev_real_insn.
9313 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9317 * config/i386/i386.c (ix86_print_operand): Use conditional
9318 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9321 2018-03-02 Richard Biener <rguenther@suse.de>
9323 PR tree-optimization/84427
9324 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9325 (bitmap_set_subtract_values): Rewrite to handle multiple
9328 (prune_clobbered_mems): Likewise.
9329 (phi_translate): Take edge instead of pred/phiblock.
9330 (phi_translate_1): Likewise.
9331 (phi_translate_set): Likewise. Insert all translated
9332 exprs for a value into the set, keeping possibly multiple
9333 expressions per value.
9334 (compute_antic_aux): Adjust for phi_translate changes.
9335 When intersecting union the expressions and prune those
9336 not in the final value set, keeping possibly multiple
9337 expressions per value. Do not use value-insertion
9338 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9339 all expressions. Add verification that the value-sets
9340 only shrink during iteration.
9341 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9342 (do_pre_regular_insertion): Likewise.
9343 (do_pre_partial_partial_insertion): Likewise.
9345 2018-03-02 Richard Biener <rguenther@suse.de>
9348 * config/darwin.c (saved_debug_info_level): New static global.
9349 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9350 (darwin_asm_lto_end): Restore debug info generation settings.
9352 2018-03-01 Martin Liska <mliska@suse.cz>
9355 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9358 2018-03-01 Richard Biener <rguenther@suse.de>
9361 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9362 type annotation with LTO when debug was disabled at compile-time.
9364 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9366 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9368 (mips_final_postscan_insn): Likewise.
9370 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9372 PR rtl-optimization/84528
9373 * alias.c (init_alias_target): Add commentary.
9374 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9375 a unique base value if the frame pointer is not eliminated
9376 to the stack pointer.
9378 2018-03-01 Tom de Vries <tom@codesourcery.com>
9380 PR rtl-optimization/83327
9381 * lra-int.h (hard_regs_spilled_into): Declare.
9382 * lra.c (hard_regs_spilled_into): Define.
9383 (init_reg_info): Init hard_regs_spilled_into.
9384 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9385 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9386 (process_bb_lives): Handle hard_regs_spilled_into.
9387 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9388 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9390 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9392 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9393 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9394 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9395 * config/rs6000/aix72.h: New file.
9397 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9399 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9400 instead of warning_at with conditional singular and plural messages
9404 * stor-layout.c (update_alignment_for_field): For
9405 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9406 && !DECL_PACKED (field), do the alignment update, just use
9407 only desired_align instead of MAX (type_align, desired_align)
9409 (place_field): Don't do known_align < desired_align handling
9410 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9411 is non-NULL, instead do it after rli->prev_field handling and
9412 only if not within a bitfield word. For DECL_PACKED (field)
9413 use type_align of BITS_PER_UNIT.
9415 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9417 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9418 superfluous parentheses and trailing spaces.
9420 2018-02-28 Richard Biener <rguenther@suse.de>
9422 PR tree-optimization/84584
9423 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9424 SCoPs with fake exit edge.
9426 2018-02-28 Martin Liska <mliska@suse.cz>
9429 * timevar.c (timer::print): Fix format to properly print 100%
9432 2018-02-28 Richard Biener <rguenther@suse.de>
9435 * genmatch.c (capture_info::walk_match): Do not mark
9436 captured expressions without operands as expr_p given
9437 they act more like predicates and should be subject to
9438 "lost tail" side-effect preserving.
9440 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9442 PR rtl-optimization/81611
9443 * auto-inc-dec.c (attempt_change): Move dead note from
9444 mem_insn if it's the next use of regno
9445 (find_address): Take address use of reg holding
9446 non-incremented value. Add parm to limit search to the named
9448 (merge_in_block): Attempt to use a mem insn that is the next
9449 use of the original regno.
9451 2018-02-27 Martin Sebor <msebor@redhat.com>
9454 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9455 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9457 2018-02-27 Martin Sebor <msebor@redhat.com>
9459 PR translation/84207
9460 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9461 n argument to unsigned HOST_WIDE_INT.
9462 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9463 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9464 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9465 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9467 2018-02-27 Richard Biener <rguenther@suse.de>
9469 PR tree-optimization/84512
9470 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9471 Do not use the estimate returned from record_stmt_cost for
9472 the scalar iteration cost but sum properly using add_stmt_cost.
9474 2018-02-27 Richard Biener <rguenther@suse.de>
9476 PR tree-optimization/84466
9477 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9478 Adjust last change to less strictly validate use operands.
9480 2018-02-27 Martin Liska <mliska@suse.cz>
9482 PR gcov-profile/84548
9483 * gcov.c (process_file): Allow partial overlap and consider it
9484 also as group functions.
9485 (output_lines): Properly calculate range of lines for a group.
9487 2018-02-27 Martin Liska <mliska@suse.cz>
9489 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9490 'ggc' suffixes. Change first column width.
9491 (timer::print): Fix formatting of the column.
9493 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9495 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9496 preserve inline entry blocks for the sake of debug inline
9497 entry point markers alone.
9498 (remove_unused_locals): Suggest in comments a better place to
9499 force the preservation of inline entry blocks that are
9500 otherwise unused, but do not preserve them.
9502 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9504 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9506 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9509 * config/i386/constraints.md (Bs): Replace
9510 ix86_indirect_branch_register with
9511 TARGET_INDIRECT_BRANCH_REGISTER.
9513 * config/i386/i386.md (indirect_jump): Likewise.
9514 (tablejump): Likewise.
9515 (*sibcall_memory): Likewise.
9516 (*sibcall_value_memory): Likewise.
9517 Peepholes of indirect call and jump via memory: Likewise.
9518 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9519 (*sibcall_value_GOT_32): Likewise.
9520 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9521 (GOT_memory_operand): Likewise.
9522 (call_insn_operand): Likewise.
9523 (sibcall_insn_operand): Likewise.
9524 (GOT32_symbol_operand): Likewise.
9525 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9527 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9529 PR rtl-optimization/83496
9530 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9531 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9532 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9533 redundant insn, if any.
9534 (relax_delay_slots): Likewise.
9535 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9537 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9539 PR tree-optimization/83965
9540 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9541 that grouped statements are part of a reduction chain. Return
9542 true if the statement is not marked as a reduction itself but
9544 (vect_recog_dot_prod_pattern): Don't check whether the statement
9545 is part of a group here.
9546 (vect_recog_sad_pattern): Likewise.
9547 (vect_recog_widen_sum_pattern): Likewise.
9549 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9552 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9554 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9557 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9559 (ix86_output_indirect_function_return): New prototype.
9560 (ix86_split_simple_return_pop_internal): Likewise.
9561 * config/i386/i386.c (indirect_return_via_cx): New.
9562 (indirect_return_via_cx_bnd): Likewise.
9563 (indirect_thunk_name): Handle return va CX_REG.
9564 (output_indirect_thunk_function): Create alias for
9565 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9566 (ix86_output_indirect_jmp): Remove the bool argument.
9567 (ix86_output_indirect_function_return): New function.
9568 (ix86_split_simple_return_pop_internal): Likewise.
9569 * config/i386/i386.md (*indirect_jump): Don't pass false
9570 to ix86_output_indirect_jmp.
9571 (*tablejump_1): Likewise.
9572 (simple_return_pop_internal): Change it to define_insn_and_split.
9573 Call ix86_split_simple_return_pop_internal to split it for
9575 (simple_return_indirect_internal): Call
9576 ix86_output_indirect_function_return instead of
9577 ix86_output_indirect_jmp.
9579 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9582 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9583 memset and value initialization afterwards.
9585 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9587 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9589 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9592 * common/config/aarch64/aarch64-common.c
9593 (aarch_option_optimization_table[]): Switch
9594 off fomit-frame-pointer
9596 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9597 Chung-Ju Wu <jasonwucj@gmail.com>
9599 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9601 (store_multiple): Ditto.
9603 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9605 * config.gcc: Add --with-cpu support for nds32 target.
9606 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9607 * config/nds32/nds32.opt: Add -mcpu= option.
9609 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9611 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9612 isel=yes): Warn for these deprecated options.
9614 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9616 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9617 ISA_2_5_MASKS_EMBEDDED.
9619 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9621 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9622 p->max as pointers rather than using iterative_hash_expr.
9624 2018-02-23 Carl Love <cel@us.ibm.com>
9626 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9627 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9629 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9630 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9631 P8V_BUILTIN_VEC_VUNSIGNED2.
9633 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9636 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9637 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9639 * lra-constraints.c (curr_insn_transform): Set up
9640 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9642 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9643 * lra-eliminations.c (spill_pseudos): Ditto.
9644 (process_insn_for_elimination): Ditto.
9645 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9646 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9647 LRA_NON_CLOBBERED_ALT.
9649 2018-02-22 Martin Sebor <msebor@redhat.com>
9651 PR tree-optimization/84480
9652 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9653 to maybe_diag_stxncpy_trunc. Call it.
9654 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9655 from gimple_fold_builtin_strcpy. Print inlining stack.
9656 (handle_builtin_stxncpy): Print inlining stack.
9657 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9659 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9662 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9663 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9664 and -fcheck-pointer-bounds are used together.
9665 (indirect_thunk_prefix): New enum.
9666 (indirect_thunk_need_prefix): New function.
9667 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9668 "_nt" instead of "_bnd" for NOTRACK prefix.
9669 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9670 (output_indirect_thunk_function): Likewise.
9672 (ix86_code_end): Update output_indirect_thunk_function calls.
9673 (ix86_output_indirect_branch_via_reg): Replace
9674 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9675 (ix86_output_indirect_branch_via_push): Likewise.
9676 (ix86_output_function_return): Likewise.
9677 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9678 incompatible with -fcf-protection=branch and
9679 -fcheck-pointer-bounds.
9681 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9684 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9685 Change gcc_assert call to output_operand_lossage.
9687 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9689 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9691 2018-02-22 DJ Delorie <dj@redhat.com>
9692 Sebastian Perta <sebastian.perta@renesas.com>
9693 Oleg Endo <olegendo@gcc.gnu.org>
9695 * config/rx/rx.c (rx_rtx_costs): New function.
9696 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9698 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9700 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9702 2018-02-22 Martin Liska <mliska@suse.cz>
9705 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9706 Add "native" as a possible value.
9708 2018-02-22 Martin Liska <mliska@suse.cz>
9711 * config/i386/i386.c (ix86_option_override_internal):
9712 Add "native" as a possible value for -march and -mtune.
9714 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9717 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9718 to all type variants.
9720 PR tree-optimization/84503
9721 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9722 width as info->bitpos + info->bitsize - start.
9723 (merged_store_group::merge_overlapping): Simplify width computation.
9724 (check_no_overlap): New function.
9725 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9726 start + width and last_order of the group, fail if check_no_overlap
9728 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9729 to group if check_no_overlap fails.
9731 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9733 * config/rs6000/altivec.md: Delete contraint arguments to
9734 define_expand, define_split, and define_peephole2, and in
9735 define_insn_and_split if always unused.
9736 * config/rs6000/darwin.md: Ditto.
9737 * config/rs6000/dfp.md: Ditto.
9738 * config/rs6000/rs6000.md: Ditto.
9739 * config/rs6000/sync.md: Ditto.
9740 * config/rs6000/vector.md: Ditto.
9741 * config/rs6000/vsx.md: Ditto.
9743 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9745 * config/rs6000/altivec.md: Write output control strings as braced
9746 blocks instead of double-quoted strings.
9747 * config/rs6000/darwin.md: Ditto.
9748 * config/rs6000/rs6000.md: Ditto.
9749 * config/rs6000/vector.md: Ditto.
9750 * config/rs6000/vsx.md: Ditto.
9752 2018-02-21 Jason Merrill <jason@redhat.com>
9754 PR c++/84314 - ICE with templates and fastcall attribute.
9755 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9757 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9759 * ipa-cp.c (determine_versionability): Fix comment typos.
9761 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9764 * ipa-cp.c (determine_versionability): Do not version functions caling
9767 2018-02-21 Martin Liska <mliska@suse.cz>
9770 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9771 Add "native" as a possible value.
9772 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9773 the macro when native cpu detection is available.
9775 2018-02-21 Martin Liska <mliska@suse.cz>
9778 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9779 Add "native" as a possible value.
9780 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9781 when native cpu detection is available.
9783 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9784 Martin Sebor <msebor@redhat.com>
9786 PR tree-optimization/84478
9787 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9789 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9790 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9791 support which is conservatively correct, for 2 only stay conservative
9792 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9793 argument to the 2 argument get_range_strlen, adjust 6 arg
9794 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9796 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9797 (gimple_fold_builtin_strlen): Pass true as last argument to
9800 2018-02-20 Martin Sebor <msebor@redhat.com>
9803 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9804 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9805 (builtin_memref::builtin_memref): Factor out parts into
9806 set_base_and_offset and call it.
9808 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9811 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9812 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9813 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9814 search at the associated MODE_INT.
9816 2018-02-20 Jeff Law <law@redhat.com>
9819 PR tree-optimization/81592
9821 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9822 for range data rather than using global data.
9823 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9824 range data rather than using global data.
9825 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9826 pass it to children as needed.
9827 (struct directive::fmtresult): Similarly.
9828 (struct directive::set_width): Similarly.
9829 (struct directive::set_precision): Similarly.
9830 (format_integer, format_directive, parse_directive): Similarly.
9831 (format_none): Accept unnamed vr_values parameter.
9832 (format_percent, format_floating, format_character): Similarly.
9833 (format_string, format_plain): Similarly.
9834 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9835 the EVRP range analyzer for range data rather than using global data.
9836 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9837 gimple-ssa-evrp-analyze.h
9838 (class sprintf_dom_walker): Add after_dom_children member function.
9839 Add evrp_range_analyzer member.
9840 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9841 range analyzer as needed.
9842 (sprintf_dom_walker::after_dom_children): New member function.
9843 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9845 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9846 (evrp_range_analyzer::pop_to_marker): Likewise.
9848 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9850 PR tree-optimization/84419
9851 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9852 with the required type if its current type is compatible but
9855 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9858 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9859 after vectorization.
9861 2018-02-20 Martin Liska <mliska@suse.cz>
9864 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9865 possible values if we don't have a hint.
9867 2018-02-20 Martin Liska <mliska@suse.cz>
9871 * final.c (shorten_branches): Build align_tab array with one
9873 * opts.c (finish_options): Add alignment option limit check.
9874 (MAX_CODE_ALIGN): Likewise.
9875 (MAX_CODE_ALIGN_VALUE): Likewise.
9876 * doc/invoke.texi: Document maximum allowed option value for
9877 all -falign-* options.
9879 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9882 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9883 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9884 * var-tracking.c (emit_note_insn_var_location): Remove all references
9885 to NOTE_INSN_CALL_ARG_LOCATION.
9886 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9887 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9888 Use copy_rtx_if_shared.
9889 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9890 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9891 (dwarf2out_var_location): Remove handling of
9892 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9894 * final.c (final_scan_insn): Remove all references to
9895 NOTE_INSN_CALL_ARG_LOCATION.
9896 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9897 before dumping final insns.
9898 * except.c (emit_note_eh_region_end): Remove all references to
9899 NOTE_INSN_CALL_ARG_LOCATION.
9900 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9901 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9902 * config/arc/arc.c (hwloop_optimize): Likewise.
9903 * config/arm/arm.c (create_fix_barrier): Likewise.
9904 * config/s390/s390.c (s390_chunkify_start): Likewise.
9905 * config/sh/sh.c (find_barrier): Likewise.
9906 * config/i386/i386.c (rest_of_insert_endbranch,
9907 ix86_seh_fixup_eh_fallthru): Likewise.
9908 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9909 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9910 * config/frv/frv.c (frv_function_prologue): Likewise.
9911 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9913 (note_outside_basic_block_p): Remove all references to
9914 NOTE_INSN_CALL_ARG_LOCATION.
9915 * gengtype.c (adjust_field_rtx_def): Likewise.
9916 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9918 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9919 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9922 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9925 PR tree-optimization/84452
9926 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9927 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9930 2018-02-19 Martin Liska <mliska@suse.cz>
9933 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9935 2018-02-19 Martin Liska <mliska@suse.cz>
9936 Richard Sandiford <richard.sandiford@linaro.org>
9938 PR tree-optimization/82491
9939 * gimple-fold.c (get_base_constructor): Make earlier bail out
9942 2018-02-19 Carl Love <cel@us.ibm.com>
9944 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9945 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9947 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9948 P8V_BUILTIN_VEC_NEG.
9950 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9952 * config/rl78/rl78.md (movdf): New define expand.
9954 2018-02-19 Martin Liska <mliska@suse.cz>
9957 * doc/invoke.texi: Fix typo.
9958 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9960 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9962 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9963 handle rs6000_single_float and rs6000_double_float specially for
9966 2018-02-16 Jeff Law <law@redhat.com>
9968 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9971 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9974 * ipa-inline.c (inline_small_functions): Fix a typo.
9976 2018-02-16 Nathan Sidwell <nathan@acm.org>
9978 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9980 2018-02-16 Carl Love <cel@us.ibm.com>
9982 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9983 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9984 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9985 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9986 expansion to P8V_BUILTIN_VEC_FLOAT2.
9988 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9990 PR rtl-optimization/70023
9991 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9992 src_regno into account.
9994 2018-02-16 Carl Love <cel@us.ibm.com>
9996 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9997 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9998 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9999 * config/rs6000/rs6000.c: Remove case statements for
10000 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10001 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10002 and P9V_BUILTIN_VEC_VINSERT4B.
10003 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10004 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10005 * config/rs6000/vsx.md:
10006 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10009 2018-02-16 Carl Love <cel@us.ibm.com>
10011 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10013 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10015 * config/rs6000/rs6000-c.c: Add the definitions for
10016 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10017 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10018 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10019 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10020 definition for insert4b and define insn *insert3b_internal.
10021 * doc/extend.texi: Add documentation for vec_extract4b.
10023 2018-02-16 Nathan Sidwell <nathan@acm.org>
10025 * doc/extend.texi (Backwards Compatibility): Mention friend
10026 injection. Note for-scope is deprecated.
10027 * doc/invoke.texi (-ffriend-injection): Deprecate.
10029 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10031 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10032 that moved to I2, also allow destinations that are a paradoxical
10033 subreg (instead of a normal reg).
10035 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10038 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10041 2018-02-16 Richard Biener <rguenther@suse.de>
10043 PR tree-optimization/84037
10044 PR tree-optimization/84016
10046 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10047 Adjust vec_construct for the fact we need additional higher latency
10048 128bit inserts for AVX256 and AVX512 vector builds.
10049 (ix86_add_stmt_cost): Scale vector construction cost for
10052 2018-02-16 Richard Biener <rguenther@suse.de>
10054 PR tree-optimization/84417
10055 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10056 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10057 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10059 2018-02-16 Martin Liska <mliska@suse.cz>
10062 * internal-fn.def (ASAN_CHECK): Set proper flags.
10063 (ASAN_MARK): Likewise.
10065 2018-02-16 Julia Koval <julia.koval@intel.com>
10067 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10068 from PTA_CANNONLAKE.
10070 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10073 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10074 Use ++iter rather than iter++ for std::list iterators.
10075 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10076 defer deleting them until all nodes in the forest are processed. Do
10077 free even leaf nodes. Change to_process into auto_vec.
10080 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10081 * vec.h (vec_default_construct): Use memset instead of placement new
10082 if BROKEN_VALUE_INITIALIZATION is defined.
10083 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10084 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10087 PR rtl-optimization/83723
10088 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10089 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10090 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10092 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10094 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10096 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10098 PR rtl-optimization/81443
10099 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10100 from inner REGs to paradoxical SUBREGs.
10102 2018-02-16 Richard Biener <rguenther@suse.de>
10104 PR tree-optimization/84399
10105 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10106 For operands we can analyze at their definition make sure we can
10107 analyze them at each use as well.
10109 2018-02-16 Richard Biener <rguenther@suse.de>
10111 PR tree-optimization/84190
10112 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10113 volatile accesses if the decl isn't volatile.
10115 2018-02-15 Jason Merrill <jason@redhat.com>
10117 PR c++/84314 - ICE with templates and fastcall attribute.
10118 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10119 TYPE_CANONICAL on an existing type.
10121 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10123 PR tree-optimization/84383
10124 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10125 dstoff nor call operand_equal_p if dstbase is NULL.
10127 PR tree-optimization/84334
10128 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10129 also a CONSTANT_CLASS_P, punt.
10131 2018-02-14 Jim Wilson <jimw@sifive.com>
10133 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10134 first SMALL_OPERAND check. New local min_second_step. Move assert
10135 to where locals are set. Add TARGET_RVC support.
10136 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10138 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10140 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10142 2018-02-14 Martin Sebor <msebor@redhat.com>
10144 PR tree-optimization/83698
10145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10146 arrays constrain the offset range to their bounds.
10147 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10148 (builtin_access::overlap): Avoid setting the size of overlap if it's
10150 (maybe_diag_overlap): Also consider arrays when deciding what values
10151 of offsets to include in diagnostics.
10153 2018-02-14 Martin Sebor <msebor@redhat.com>
10156 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10157 that correspond to the kind of a declaration.
10159 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10162 * config/pa/pa.md: Load address of PIC label using the linkage table
10163 if the label is nonlocal.
10165 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10167 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10168 warning message if user requests -maltivec=be.
10169 * doc/invoke.texi: Document deprecation of -maltivec=be.
10171 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10174 * config/rs6000/rs6000-c.c: Update definitions for
10175 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10176 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10178 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10181 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10182 add _get_ssp intrinsics. Remove argument from
10183 __builtin_ia32_rdssp[d|q].
10184 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10185 * config/i386/i386-builtin.def: Remove argument from
10186 __builtin_ia32_rdssp[d|q].
10187 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10188 ix86_expand_special_args_builtin for _rdssp[d|q].
10189 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10190 Clear register before usage.
10191 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10192 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10194 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10196 PR tree-optimization/84357
10197 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10198 operand 1 of an ARRAY_REF too.
10200 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10203 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10204 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10206 (set_of_reg): New struct.
10207 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10208 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10209 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10211 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10212 Split into bitclr, bitset, bitinvert patterns if appropriate.
10213 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10214 use rx_fuse_in_memory_bitop.
10215 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10216 to named insn, correct maximum insn length.
10218 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10221 * machmode.def: Define a complex mode for PARTIAL_INT.
10222 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10224 * doc/rtl.texi: Document CSPImode.
10225 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10227 (msp430_hard_regno_nregs_with_padding): Likewise.
10229 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10232 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10234 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10236 PR rtl-optimization/84169
10237 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10238 we generated a parallel as new i3 and we split that to new i2 and i3
10239 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10240 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10241 those to i2, not i1. Partially rewrite this scan code.
10243 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10246 * stor-layout.c (place_field): For variable length fields, adjust
10247 offset_align afterwards not just based on the field's alignment,
10248 but also on the size.
10250 PR middle-end/84309
10251 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10252 of exps and logs in the use_exp2 case.
10254 2018-02-13 Jeff Law <law@redhat.com>
10256 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10257 entry for "vector".
10259 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10262 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10266 * common.opt (gas-loc-support, gas-locview-support): New.
10267 (ginline-points, ginternal-reset-location-views): New.
10268 * doc/invoke.texi: Document them. Use @itemx where intended.
10269 (gvariable-location-views): Adjust.
10270 * target.def (reset_location_view): New.
10271 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10272 (TARGET_RESET_LOCATION_VIEW): New.
10273 * doc/tm.texi: Rebuilt.
10274 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10275 (dwarf2out_default_as_locview_support): New.
10276 (output_asm_line_debug_info): Use option variables.
10277 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10278 (output_loc_list): Likewise.
10279 (add_high_low_attributes): Check option variables.
10280 Don't output entry view attribute in strict mode.
10281 (gen_inlined_subroutine_die): Check option variables.
10282 (dwarf2out_inline_entry): Likewise.
10283 (init_sections_and_labels): Likewise.
10284 (dwarf2out_early_finish): Likewise.
10285 (maybe_reset_location_view): New, from...
10286 (dwarf2out_var_location): ... here. Call it.
10287 * debug.h (dwarf2out_default_as_loc_support): Declare.
10288 (dwarf2out_default_as_locview_support): Declare.
10289 * hooks.c (hook_int_rtx_insn_0): New.
10290 * hooks.h (hook_int_rtx_insn_0): Declare.
10291 * toplev.c (process_options): Take -gas-loc-support and
10292 -gas-locview-support from dwarf2out. Enable
10293 -gvariable-location-views by default only with locview
10294 assembler support. Enable -ginternal-reset-location-views by
10295 default only if the target defines the corresponding hook.
10296 Enable -ginline-points by default if location views are
10297 enabled; force it disabled if statement frontiers are
10299 * tree-inline.c (expand_call_inline): Check option variables.
10300 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10302 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10304 PR tree-optimization/84321
10305 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10306 handling. Also check whether the anti-range contains any values
10307 that satisfy the mask; switch to a VR_RANGE if not.
10309 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10312 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10314 2018-02-13 Martin Jambor <mjambor@suse.cz>
10317 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10318 of call statements, also set location of a load to a temporary.
10320 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10322 * config/rl78/rl78.c (add_vector_labels): New function.
10323 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10324 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10325 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10326 which checks that no arguments are passed.
10327 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10328 * doc/extend.texi: Documentation for the new attribute.
10330 2018-02-13 Andreas Schwab <schwab@suse.de>
10332 * config/riscv/linux.h (CPP_SPEC): Define.
10334 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10337 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10338 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10339 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10340 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10341 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10342 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10343 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10344 temporarily for AES and PCLMUL builtins.
10346 PR tree-optimization/84339
10347 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10348 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10351 PR middle-end/84309
10352 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10353 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10354 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10356 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10358 * omp-simd-clone.h: New file.
10359 * omp-simd-clone.c: Include omp-simd-clone.h.
10360 (expand_simd_clones): No longer static.
10361 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10362 cgraph.h and omp-simd-clone.h.
10363 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10364 (vect_recog_widen_shift_pattern): Formatting fix.
10365 (vect_pattern_recog_1): Don't check optab for calls.
10368 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10369 operands[2] into a REG before using gen_lowpart on it.
10371 2018-02-12 Jeff Law <law@redhat.com>
10374 * config/sh/sh.c (find_barrier): Consider a sibling call
10377 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10378 successfully back substituting a reg.
10380 2018-02-12 Richard Biener <rguenther@suse.de>
10382 PR tree-optimization/84037
10383 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10384 parameter, move visited init to caller.
10385 (vect_slp_analyze_operations): Separate cost from validity
10386 check, initialize visited once for all instances.
10387 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10389 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10391 (vect_model_promotion_demotion_cost): Likewise.
10392 (vectorizable_bswap): Guard cost modeling with !slp_node
10393 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10395 (vectorizable_call): Likewise.
10396 (vectorizable_conversion): Likewise.
10397 (vectorizable_assignment): Likewise.
10398 (vectorizable_shift): Likewise.
10399 (vectorizable_operation): Likewise.
10400 (vectorizable_store): Likewise.
10401 (vectorizable_load): Likewise.
10402 (vectorizable_condition): Likewise.
10403 (vectorizable_comparison): Likewise.
10405 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10408 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10409 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10410 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10412 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10414 PR middle-end/83665
10415 * params.def (inline-min-speedup): Increase from 8 to 15.
10416 (max-inline-insns-auto): Decrease from 40 to 30.
10417 * ipa-split.c (consider_split): Add some buffer for function to
10418 be considered inlining candidate.
10419 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10422 2018-02-12 Richard Biener <rguenther@suse.de>
10424 PR tree-optimization/84037
10425 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10426 matched stmts if we cannot swap the non-matched ones.
10428 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10430 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10431 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10432 _mm_maskz_scalef_round_ss): New intrinsics.
10433 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10434 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10435 __builtin_ia32_scalefss_round): Remove.
10436 (__builtin_ia32_scalefsd_mask_round,
10437 __builtin_ia32_scalefss_mask_round): New intrinsics.
10438 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10439 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10440 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10441 "<round_constraint>")): Changed to ...
10442 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10443 "<round_scalar_constraint>")): ... this.
10444 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10445 %0, %1, %2<round_op3>}"): Changed to ...
10446 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10447 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10448 %2<round_scalar_mask_op3>}"): ... this.
10449 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10451 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10453 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10454 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10455 (_mm_maskz_sqrt_round_ss): New intrinsics.
10456 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10457 (__builtin_ia32_sqrtsd_mask_round)
10458 (__builtin_ia32_sqrtss_mask_round): New builtins.
10459 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10460 (__builtin_ia32_sqrtss_round): Remove.
10461 (__builtin_ia32_sqrtsd_mask_round)
10462 (__builtin_ia32_sqrtss_mask_round): New builtins.
10463 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10464 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10465 ((match_operand:VF_128 1 "vector_operand"
10466 "xBm,<round_constraint>")): Changed to ...
10467 ((match_operand:VF_128 1 "vector_operand"
10468 "xBm,<round_scalar_constraint>")): ... this.
10469 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10470 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10471 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10472 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10473 %<iptr>1<round_scalar_mask_op3>}): ... this.
10474 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10475 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10477 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10480 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10481 Cast vec_cmpeq result to correct type.
10482 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10483 Cast vec_cmpgt result to correct type.
10485 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10487 * final.c (final_scan_insn_1): Renamed from...
10488 (final_scan_insn): ... this. New wrapper, to recover
10489 seen from the outermost call in recursive ones.
10490 * config/sparc/sparc.c (output_return): Drop seen from call.
10491 (output_sibcall): Likewise.
10492 * config/visium/visium.c (output_branch): Likewise.
10494 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10496 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10499 2018-02-10 Alan Modra <amodra@gmail.com>
10502 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10503 Specify LR as an input.
10505 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10508 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10509 remove_member_access_dummy_vars): New functions.
10510 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10511 lower_omp_1, execute_lower_omp): Use them.
10513 PR rtl-optimization/84308
10514 * shrink-wrap.c (spread_components): Release todo vector.
10516 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10518 PR rtl-optimization/57193
10519 * ira-color.c (struct allocno_color_data): Add member
10520 conflict_allocno_hard_prefs.
10521 (update_conflict_allocno_hard_prefs): New.
10522 (bucket_allocno_compare_func): Add a preference based on
10523 conflict_allocno_hard_prefs.
10524 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10525 (color_allocnos): Remove a dead code. Initiate
10526 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10528 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10531 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10532 constraint from =wa to wa. Avoid a subreg on the output operand,
10533 instead use a pseudo and subreg it in a move.
10534 (p9_xxbrd_<mode>): Changed to ...
10535 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10536 (p9_xxbrd_v2df): New expander.
10537 (p9_xxbrw_<mode>): Changed to ...
10538 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10539 (p9_xxbrw_v4sf): New expander.
10541 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10543 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10545 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10548 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10549 multiply in 32-bit mode.
10550 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10551 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10554 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10556 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10557 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10558 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10559 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10561 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10564 * dwarf2out.c (is_trivial_indirect_ref): New function.
10565 (dwarf2out_late_global_decl): Do not generate a location
10566 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10567 and that are not defined in the current unit.
10569 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10571 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10572 instead of a libcall for UNORDERED.
10574 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10577 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10578 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10580 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10583 * config/s390/s390.c (s390_set_current_function): Invoke
10584 s390_indirect_branch_settings also if fndecl didn't change.
10586 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10588 * config/rs6000/rs6000.md (blockage): Set length to zero.
10590 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10592 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10594 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10597 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10598 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10602 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10603 PARALLEL incoming that failed vt_get_decl_and_offset check.
10605 PR middle-end/84237
10606 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10607 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10609 (get_variable_section): For decls in named .bss* sections pass true as
10610 second argument to bss_initializer_p.
10612 2018-02-09 Marek Polacek <polacek@redhat.com>
10613 Jakub Jelinek <jakub@redhat.com>
10616 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10617 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10618 Sync some changes from cxx_fold_indirect_ref.
10620 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10622 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10624 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10625 (BLOCK_INLINE_ENTRY_LABEL): New.
10626 (dwarf2out_var_location): Disregard inline entry markers.
10627 (inline_entry_data): New struct.
10628 (inline_entry_data_hasher): New hashtable type.
10629 (inline_entry_data_hasher::hash): New.
10630 (inline_entry_data_hasher::equal): New.
10631 (inline_entry_data_table): New variable.
10632 (add_high_low_attributes): Add DW_AT_entry_pc and
10633 DW_AT_GNU_entry_view attributes if a pending entry is found
10634 in inline_entry_data_table. Add old entry_pc attribute only
10635 if debug nonbinding markers are disabled.
10636 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10637 markers are enabled.
10638 (block_within_block_p, dwarf2out_inline_entry): New.
10639 (dwarf2out_finish): Check that no entries remained in
10640 inline_entry_data_table.
10641 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10642 (final_scan_insn, notice_source_line): Likewise.
10643 (rest_of_clean_state): Skip inline entry markers.
10644 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10646 * gimple.c (gimple_build_debug_inline_entry): New.
10647 * gimple.h (enum gimple_debug_subcode): Add
10648 GIMPLE_DEBUG_INLINE_ENTRY.
10649 (gimple_build_debug_inline_entry): Declare.
10650 (gimple_debug_inline_entry_p): New.
10651 (gimple_debug_nonbind_marker_p): Adjust.
10652 * insn-notes.def (INLINE_ENTRY): New.
10653 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10654 inline entry marker notes.
10655 (print_insn): Likewise.
10656 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10657 (INSN_DEBUG_MARKER_KIND): Likewise.
10658 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10659 * tree-inline.c (expand_call_inline): Build and insert
10660 debug_inline_entry stmt.
10661 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10662 inline entry blocks early, if nonbind markers are enabled.
10663 (dump_scope_block): Dump fragment info.
10664 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10665 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10666 (gimple_build_debug_inline_entry): New.
10667 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10668 Enable/disable inline entry points too.
10669 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10670 (DEBUG_INSN): Describe inline entry markers.
10672 * common.opt (gvariable-location-views): New.
10673 (gvariable-location-views=incompat5): New.
10674 * config.in: Rebuilt.
10675 * configure: Rebuilt.
10676 * configure.ac: Test assembler for view support.
10677 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10678 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10679 * dwarf2out.c (var_loc_view): New typedef.
10680 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10681 (dwarf2out_locviews_in_attribute): New.
10682 (dwarf2out_locviews_in_loclist): New.
10683 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10684 (enum dw_line_info_opcode): Add LI_adv_address.
10685 (struct dw_line_info_table): Add view.
10686 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10687 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10688 (zero_view_p): New variable.
10689 (ZERO_VIEW_P): New macro.
10690 (output_asm_line_debug_info): New.
10691 (struct var_loc_node): Add view.
10692 (add_AT_view_list, AT_loc_list): New.
10693 (add_var_loc_to_decl): Add view param. Test it against last.
10694 (new_loc_list): Add view params. Record them.
10695 (AT_loc_list_ptr): Handle loc and view lists.
10696 (view_list_to_loc_list_val_node): New.
10697 (print_dw_val): Handle dw_val_class_view_list.
10698 (size_of_die): Likewise.
10699 (value_format): Likewise.
10700 (loc_list_has_views): New.
10701 (gen_llsym): Set vl_symbol too.
10702 (maybe_gen_llsym, skip_loc_list_entry): New.
10703 (dwarf2out_maybe_output_loclist_view_pair): New.
10704 (output_loc_list): Output view list or entries too.
10705 (output_view_list_offset): New.
10706 (output_die): Handle dw_val_class_view_list.
10707 (output_dwarf_version): New.
10708 (output_compilation_unit_header): Use it.
10709 (output_skeleton_debug_sections): Likewise.
10710 (output_rnglists, output_line_info): Likewise.
10711 (output_pubnames, output_aranges): Update version comments.
10712 (output_one_line_info_table): Output view numbers in asm comments.
10713 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10714 Call maybe_gen_llsym.
10715 (loc_list_from_tree_1): Adjust.
10716 (add_AT_location_description): Create view list attribute if
10717 needed, check it's absent otherwise.
10718 (convert_cfa_to_fb_loc_list): Adjust.
10719 (maybe_emit_file): Call output_asm_line_debug_info for test.
10720 (dwarf2out_var_location): Reset views as needed. Precompute
10721 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10722 attribute. Set view.
10723 (new_line_info_table): Reset next view.
10724 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10725 (dwarf2out_source_line): Likewise. Output view resets and labels to
10726 the assembler, or select appropriate line info opcodes.
10727 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10728 (optimize_string_length): Catch it. Adjust.
10729 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10730 dw_val_class_view_list, and remove it if no longer needed.
10731 (hash_loc_list): Hash view numbers.
10732 (loc_list_hasher::equal): Compare them.
10733 (optimize_location_lists): Check whether a view list symbol is
10734 needed, and whether the locview attribute is present, and
10735 whether they match. Remove the locview attribute if no longer
10737 (index_location_lists): Call skip_loc_list_entry for test.
10738 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10739 Use output_dwarf_version.
10740 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10741 (struct dw_val_node): Add val_view_list.
10742 * final.c (SEEN_NEXT_VIEW): New.
10743 (set_next_view_needed): New.
10744 (clear_next_view_needed): New.
10745 (maybe_output_next_view): New.
10746 (final_start_function): Rename to...
10747 (final_start_function_1): ... this. Take pointer to FIRST,
10748 add SEEN parameter. Emit param bindings in the initial view.
10749 (final_start_function): Reintroduce SEEN-less interface.
10750 (final): Rename to...
10751 (final_1): ... this. Take SEEN parameter. Output final pending
10752 next view at the end.
10753 (final): Reintroduce seen-less interface.
10754 (final_scan_insn): Output pending next view before switching
10755 sections or ending a block. Mark the next view as needed when
10756 outputting variable locations. Notify debug backend of section
10757 changes, and of location view changes.
10758 (rest_of_handle_final): Adjust.
10759 * toplev.c (process_options): Autodetect value for debug variable
10760 location views option. Warn on incompat5 without -gdwarf-5.
10761 * doc/invoke.texi (gvariable-location-views): New.
10762 (gvariable-location-views=incompat5): New.
10763 (gno-variable-location-views): New.
10765 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10767 PR tree-optimization/84136
10768 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10769 that the result of find_edge is non-NULL.
10771 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10774 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10775 storing integer register in SImode. Fix cost of 256 and 512
10776 byte aligned SSE register store.
10778 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10780 * config/i386/i386.c (ix86_multiplication_cost): Fix
10781 multiplication cost for TARGET_AVX512DQ.
10783 2018-02-08 Marek Polacek <polacek@redhat.com>
10785 PR tree-optimization/84238
10786 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10789 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10791 PR tree-optimization/84265
10792 * tree-vect-stmts.c (vectorizable_store): Don't treat
10793 VMAT_CONTIGUOUS accesses as grouped.
10794 (vectorizable_load): Likewise.
10796 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10798 PR tree-optimization/81635
10799 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10800 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10801 (test_round_for_mask): New functions.
10802 (wide_int_cc_tests): Call test_round_for_mask.
10803 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10804 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10805 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10806 range returned by get_range_info.
10808 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10811 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10812 * symtab.c: Include builtins.h
10813 (symtab_node::output_to_lto_symbol_table_p): Move here
10814 from lto-streamer-out.c:output_symbol_p.
10815 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10816 (output_symbol_p): Move all logic to symtab.c
10817 (produce_symtab): Update.
10819 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10821 * config/s390/s390-opts.h (enum indirect_branch): Define.
10822 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10823 (s390_indirect_branch_via_thunk)
10824 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10825 (enum s390_indirect_branch_type): Define.
10826 * config/s390/s390.c (struct s390_frame_layout, struct
10827 machine_function): Remove.
10828 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10829 (indirect_branch_table_label_no, indirect_branch_table_name):
10831 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10832 (enum s390_indirect_branch_option): Define.
10833 (s390_return_addr_from_memory): New function.
10834 (s390_handle_string_attribute): New function.
10835 (s390_attribute_table): Add new attribute handler.
10836 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10837 (s390_indirect_branch_via_thunk): New function.
10838 (s390_indirect_branch_via_inline_thunk): New function.
10839 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10840 sibling call optimization for non z10 compiles.
10841 (s390_emit_call): Force indirect branch target to be a single
10842 register. Add r1 clobber for non-z10 compiles.
10843 (s390_emit_epilogue): Emit return jump via return_use expander.
10844 (s390_reorg): Handle JUMP_INSNs as execute targets.
10845 (s390_option_override_internal): Perform validity checks for the
10846 new command line options.
10847 (s390_indirect_branch_attrvalue): New function.
10848 (s390_indirect_branch_settings): New function.
10849 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10850 (s390_output_indirect_thunk_function): New function.
10851 (s390_code_end): Implement target hook.
10852 (s390_case_values_threshold): Implement target hook.
10853 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10855 * config/s390/s390.h (struct s390_frame_layout)
10856 (struct machine_function): Move here from s390.c.
10857 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10858 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10859 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10860 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10861 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10862 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10863 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10864 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10865 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10866 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10867 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10868 (mnemonic attribute): Add values which aren't recognized
10870 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10871 pattern for branch conversion. Fix mnemonic attribute.
10872 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10873 indirect branch via thunk if requested.
10874 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10875 ("*indirect_jump"): Disable for branch conversion using out of
10877 ("indirect_jump_via_thunk<mode>_z10")
10878 ("indirect_jump_via_thunk<mode>")
10879 ("indirect_jump_via_inlinethunk<mode>_z10")
10880 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10881 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10882 ("casesi_jump_via_inlinethunk<mode>_z10")
10883 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10884 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10885 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10886 ("*indirect2_jump"): Disable for branch conversion.
10887 ("casesi_jump"): Turn into expander and expand patterns for branch
10889 ("return_use"): New expander.
10890 ("*return"): Emit return via thunk and rename it to ...
10891 ("*return<mode>"): ... this one.
10892 * config/s390/s390.opt: Add new options and and enum for the
10895 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10897 * lra-constraints.c (match_reload): Unconditionally use
10898 gen_lowpart_SUBREG, rather than selecting between that
10899 and equivalent gen_rtx_SUBREG code.
10901 2018-02-08 Richard Biener <rguenther@suse.de>
10903 PR tree-optimization/84233
10904 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10905 changed flag instead of boguously re-using phi_inserted.
10907 2018-02-08 Martin Jambor <mjambor@suse.cz>
10909 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10910 static local variables.
10912 2018-02-08 Richard Biener <rguenther@suse.de>
10914 PR tree-optimization/84278
10915 * tree-vect-stmts.c (vectorizable_store): When looking for
10916 smaller vector types to perform grouped strided loads/stores
10917 make sure the mode is supported by the target.
10918 (vectorizable_load): Likewise.
10920 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10922 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10923 Increase LDP/STP opportunities by adding adjacent callee-saves.
10925 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10927 PR rtl-optimization/84068
10928 PR rtl-optimization/83459
10929 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10931 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10933 PR tree-optimization/84224
10934 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10935 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10936 non-zero arguments.
10938 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10941 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10942 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10943 count, remove one USE.
10945 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10947 * doc/install.texi (Configuration): Document the
10948 --with-long-double-format={ibm,ieee} PowerPC configuration
10952 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10953 Convert from define_expand to be define_insn_and_split. Rework
10954 float/double/_Float128 conversions to QI/HI/SImode to work with
10955 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10956 conversions to QI/HImode types did a store and then a load to
10957 truncate the value. For conversions to VSX registers, don't split
10958 the insn, instead emit the code directly. Use the code iterator
10959 any_fix to combine signed and unsigned conversions.
10960 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10961 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10962 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10963 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10964 (fix_<mode>di2_hw): Likewise.
10965 (fixuns_<mode>di2_hw): Likewise.
10966 (fix_<mode>si2_hw): Likewise.
10967 (fixuns_<mode>si2_hw): Likewise.
10968 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10969 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10970 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10971 fix<uns>_trunc<SFDF:mode>si2_p8.
10972 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10973 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10974 (fix<uns>_<mode>_mem): Likewise.
10975 (fctiw<u>z_<mode>_mem): Likewise.
10976 (fix<uns>_<mode>_mem): Likewise.
10977 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10978 the register allocator from doing a direct move to the GPRs to do
10979 a store, and instead use the ISA 3.0 store byte/half-word from
10980 vector register instruction. For IEEE 128-bit floating point,
10981 also optimize stores of 32-bit ints.
10982 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10984 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10986 * genextract.c (push_pathstr_operand): New function to support
10988 (walk_rtx): Call push_pathstr_operand.
10989 (print_path): Support [a-zA-Z].
10991 2018-02-07 Richard Biener <rguenther@suse.de>
10993 PR tree-optimization/84037
10994 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10995 (cse_and_gimplify_to_preheader): Declare.
10996 (vect_get_place_in_interleaving_chain): Likewise.
10997 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10999 (_loop_vec_info::~_loop_vec_info): Delete it.
11000 (cse_and_gimplify_to_preheader): New function.
11001 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11002 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11003 (vectorizable_load): Likewise. For grouped stores always base
11004 the IV on the first element.
11005 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11006 condition before gimplifying.
11008 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11010 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11011 *DIV_EXPR and *MOD_EXPR.
11013 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11016 * config/i386/i386.c (ix86_option_override_internal): Mask out
11017 the CF_SET bit when checking -fcf-protection.
11019 2018-02-07 Tom de Vries <tom@codesourcery.com>
11022 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11025 2018-02-07 Richard Biener <rguenther@suse.de>
11027 PR tree-optimization/84204
11028 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11031 PR tree-optimization/84205
11032 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11033 special-case isl_ast_op_zdiv_r.
11035 PR tree-optimization/84223
11036 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11037 Only add conditions from within the region.
11038 (gather_bbs::after_dom_children): Adjust.
11040 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11043 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11044 * config/avr/avr.md: Only post-reload split REG-REG moves if
11045 either register is GENERAL_REG_P.
11047 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11049 PR tree-optimization/84235
11050 * tree-ssa-scopedtables.c
11051 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11052 if the subtraction is performed in floating point type where NaNs are
11053 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11054 build 1. Formatting fix.
11056 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11059 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11060 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11061 and skip it regardless of bb boundaries. Use CALL_P macro,
11062 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11063 unnecessarily, formatting fix.
11065 2018-02-06 Michael Collison <michael.collison@arm.com>
11067 * config/arm/thumb2.md:
11068 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11069 (*thumb_mov_notscc): Ditto.
11071 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11074 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11075 unsigned_fix, not "s".
11077 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11079 * configure.ac (gcc_fn_eh_frame_ro): New function.
11080 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11081 correct .eh_frame permissions.
11082 * configure: Regenerate.
11084 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11086 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11087 irrelevant options.
11089 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11091 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11092 Display warning message for -mno-speculate-indirect-jumps.
11094 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11096 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11098 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11100 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11102 PR tree-optimization/84225
11103 * tree-eh.c (find_trapping_overflow): Only call
11104 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11106 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11109 * config/i386/i386.c: Reimplement the check of possible options
11110 -mibt/-mshstk conbination. Change error messages.
11111 * doc/invoke.texi: Fix a typo: remove extra '='.
11113 2018-02-06 Marek Polacek <polacek@redhat.com>
11115 PR tree-optimization/84228
11116 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11118 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11121 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11122 emitted arch directives.
11123 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11124 __ARM_FEATURE_COPROC before changing architectures.
11126 2018-02-06 Richard Biener <rguenther@suse.de>
11128 * config/i386/i386.c (print_reg): Fix typo.
11129 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11131 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11133 * configure: Regenerate.
11135 2018-02-05 Martin Sebor <msebor@redhat.com>
11137 PR tree-optimization/83369
11138 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11141 2018-02-05 Martin Liska <mliska@suse.cz>
11143 * doc/invoke.texi: Cherry-pick upstream r323995.
11145 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11147 * ira.c (ira_init_register_move_cost): Adjust comment.
11149 2018-02-05 Martin Liska <mliska@suse.cz>
11151 PR gcov-profile/84137
11152 * doc/gcov.texi: Fix typo in documentation.
11154 2018-02-05 Martin Liska <mliska@suse.cz>
11156 PR gcov-profile/83879
11157 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11158 using dlopen functionality.
11160 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11162 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11163 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11164 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11165 _mm_maskz_range_round_ss): New intrinsics.
11166 (__builtin_ia32_rangesd128_round)
11167 (__builtin_ia32_rangess128_round): Remove.
11168 (__builtin_ia32_rangesd128_mask_round,
11169 __builtin_ia32_rangess128_mask_round): New builtins.
11170 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11171 __builtin_ia32_rangess128_round): Remove.
11172 (__builtin_ia32_rangesd128_mask_round,
11173 __builtin_ia32_rangess128_mask_round): New builtins.
11174 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11175 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11176 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11177 "<round_saeonly_constraint>")): Changed to ...
11178 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11179 "<round_saeonly_scalar_constraint>")): ... this.
11180 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11181 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11182 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11183 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11184 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11186 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11188 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11190 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11191 Remove all values except native, 8540 and 8548.
11193 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11195 * config/i386/i386.c (ix86_output_function_return): Pass
11196 INVALID_REGNUM, instead of -1, as invalid register number to
11197 indirect_thunk_name and output_indirect_thunk.
11199 2018-02-02 Julia Koval <julia.koval@intel.com>
11201 * config.gcc: Add -march=icelake.
11202 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11203 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11204 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11205 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11206 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11207 (processor_target_table): Add icelake.
11208 (ix86_option_override_internal): Handle new PTAs.
11209 (get_builtin_code_for_version): Handle icelake.
11210 (M_INTEL_COREI7_ICELAKE): New.
11211 (fold_builtin_cpu): Handle icelake.
11212 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11213 * doc/invoke.texi: Add -march=icelake.
11215 2018-02-02 Julia Koval <julia.koval@intel.com>
11217 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11218 to wide_int_bitmask.
11219 * wide-int-bitmask.h: New.
11221 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11224 * config/i386/i386.md: Replace Pmode with word_mode in
11225 builtin_setjmp_setup and builtin_longjmp to support x32.
11227 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11231 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11233 (rs6000_supported_cpu_names): New static variable.
11234 (linux_cpu_translation_table): Likewise.
11235 (elf_platform) <cpu>: Define new static variable and use it.
11236 Translate kernel AT_PLATFORM name to canonical name if needed.
11237 Error if platform name is unknown.
11239 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11242 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11244 2018-02-01 Jeff Law <law@redhat.com>
11247 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11248 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11249 the scratch if RELEASE_VIA_POP is false.
11250 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11251 If we have to save a temporary register, decrement SIZE appropriately.
11252 Pass new arguments to release_scratch_register_on_entry.
11253 (ix86_adjust_stack_and_probe): Likewise.
11254 (ix86_emit_probe_stack_range): Pass new arguments to
11255 release_scratch_register_on_entry.
11257 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11259 PR rtl-optimization/84157
11260 * combine.c (change_zero_ext): Use REG_P predicate in
11261 front of HARD_REGISTER_P predicate.
11263 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11265 * config/avr/avr.c (avr_option_override): Move disabling of
11266 -fdelete-null-pointer-checks to...
11267 * common/config/avr/avr-common.c (avr_option_optimization_table):
11270 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11272 PR tree-optimization/81635
11273 * tree-data-ref.c (split_constant_offset_1): For types that
11274 wrap on overflow, try to use range info to prove that wrapping
11277 2018-02-01 Renlin Li <renlin.li@arm.com>
11280 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11281 TAILCALL_ADDR_REGS.
11282 (aarch64_register_move_cost): Likewise.
11283 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11284 TAILCALL_ADDR_REGS.
11285 (REG_CLASS_NAMES): Likewise.
11286 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11287 TAILCALL_ADDR_REGS. Remove IP registers.
11288 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11290 2018-02-01 Richard Biener <rguenther@suse.de>
11292 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11293 for specifying RPO order and allow NULL for that.
11294 * domwalk.c (dom_walker::dom_walker): Likewise.
11295 (dom_walker::walk): Handle NULL RPO order.
11296 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11298 (rewrite_update_dom_walker): Likewise.
11299 (mark_def_dom_walker): Likewise.
11301 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11303 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11304 (aarch64_maybe_expand_sve_subreg_move): Declare.
11305 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11306 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11308 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11309 that are semantically a reverse operation.
11310 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11311 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11312 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11314 (aarch64_can_change_mode_class): For big-endian, forbid changes
11315 between two SVE modes if they have different element sizes.
11317 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11319 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11320 the TImode handling for big-endian targets.
11322 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11324 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11325 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11327 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11328 Remove BSWAP handing for big-endian targets and use the form of
11329 LD1RQ appropariate for the mode.
11331 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11333 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11334 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11335 duplicated element.
11337 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11340 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11341 check for operands that need to go through aarch64_sve_reload_be.
11343 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11345 PR tree-optimization/81661
11346 PR tree-optimization/84117
11347 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11348 * tree-eh.c: Include gimplify.h.
11349 (find_trapping_overflow, replace_trapping_overflow,
11350 rewrite_to_non_trapping_overflow): New functions.
11351 * tree-vect-loop.c: Include tree-eh.h.
11352 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11353 * tree-data-ref.c: Include tree-eh.h.
11354 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11356 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11358 PR rtl-optimization/84123
11359 * combine.c (change_zero_ext): Check if hard register satisfies
11360 can_change_dest_mode before calling gen_lowpart_SUBREG.
11362 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11365 * ira.c (ira_init_register_move_cost): Remove assert.
11367 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11369 PR rtl-optimization/84071
11370 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11371 * doc/tm.texi: Regenerate.
11373 2018-01-31 Richard Biener <rguenther@suse.de>
11375 PR tree-optimization/84132
11376 * tree-data-ref.c (analyze_miv_subscript): Properly
11377 check whether evolution_function_is_affine_multivariate_p
11378 before calling gcd_of_steps_may_divide_p.
11380 2018-01-31 Julia Koval <julia.koval@intel.com>
11383 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11384 * config/i386/i386.md (rdpid_rex64) New.
11385 (rdpid): Make 32bit only.
11387 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11390 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11391 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11393 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11396 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11398 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11400 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11402 PR rtl-optimization/84071
11403 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11404 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11406 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11408 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11409 (arc_attribute_table): Add 'aux' attribute.
11410 (arc_in_small_data_p): Consider aux like variables.
11411 (arc_is_aux_reg_p): New function.
11412 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11413 (arc_get_aux_arg): New function.
11414 (prepare_move_operands): Handle aux-register access.
11415 (arc_handle_aux_attribute): New function.
11416 * doc/extend.texi (ARC Variable attributes): Add subsection.
11418 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11420 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11421 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11422 (arc_attribute_table): Add 'uncached' attribute.
11423 (arc_print_operand): Print '.di' flag for uncached memory
11425 (arc_in_small_data_p): Do not consider for small data the uncached
11427 (arc_is_uncached_mem_p): New function.
11428 * config/arc/predicates.md (compact_store_memory_operand): Check
11429 for uncached memory accesses.
11430 (nonvol_nonimm_operand): Likewise.
11431 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11433 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11436 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11437 falign-loops=): Add Optimization flag.
11439 2018-01-30 Jeff Law <law@redhat.com>
11442 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11443 INT_REGISTERS_SAVED. Check it prior to calling
11444 get_scratch_register_on_entry.
11445 (ix86_adjust_stack_and_probe): Similarly.
11446 (ix86_emit_probe_stack_range): Similarly.
11447 (ix86_expand_prologue): Corresponding changes.
11449 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11452 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11453 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11455 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11458 * lra-constraints.c (curr_insn_transform): Process AND in the
11461 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11463 PR rtl-optimization/83986
11464 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11465 dependence against last_pending_memory_flush in addition to
11466 pending_jump_insns.
11468 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11470 PR tree-optimization/81611
11471 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11474 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11477 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11480 2018-01-30 Richard Biener <rguenther@suse.de>
11481 Jakub Jelinek <jakub@redhat.com>
11483 PR tree-optimization/84111
11484 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11485 inner loops added during recursion, as they don't have up-to-date
11488 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11491 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11492 (can_inline_edge_by_limits_p): ... here.
11493 (can_early_inline_edge_p, check_callers,
11494 update_caller_keys, update_callee_keys, recursive_inlining,
11495 add_new_edges_to_heap, speculation_useful_p,
11496 inline_small_functions,
11497 inline_small_functions, flatten_function,
11498 inline_to_all_callers_1): Update.
11500 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11502 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11505 2018-01-30 Richard Biener <rguenther@suse.de>
11507 PR tree-optimization/83008
11508 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11509 invariant and constant vector uses in stmts when they need
11510 more than one stmt.
11512 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11515 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11516 * configure: Regenerate.
11518 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11520 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11522 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11523 Use gen_rtx_REG rather than gen_lowpart.
11525 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11527 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11528 rather than 0 when creating partial subregs.
11530 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11532 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11535 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11538 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11539 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11540 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11541 flags. This restores the settings used before the 2017-07-24.
11542 Turning off pre increment/decrement/modify allows IVOPTS to
11543 optimize DF/SF loops where the index is an int.
11545 2018-01-29 Richard Biener <rguenther@suse.de>
11546 Kelvin Nilsen <kelvin@gcc.gnu.org>
11549 * tree-vect-stmts.c (vectorizable_call): Don't call
11550 targetm.vectorize_builtin_md_vectorized_function if callee is
11553 2018-01-22 Carl Love <cel@us.ibm.com>
11555 * doc/extend.tex: Fix typo in second arg in
11556 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11558 2018-01-29 Richard Biener <rguenther@suse.de>
11560 PR tree-optimization/84086
11561 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11562 (flush_ssaname_freelist): When SSA names were released reset
11563 the SCEV hash table.
11565 2018-01-29 Richard Biener <rguenther@suse.de>
11567 PR tree-optimization/84057
11568 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11569 removed paths when removing edges.
11571 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11573 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11574 -mfunction-return=@var{choice}.
11576 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11578 PR diagnostic/84034
11579 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11580 Handle CR like TAB.
11581 (layout::print_source_line): Likewise.
11582 (test_get_line_width_without_trailing_whitespace): Add test cases.
11584 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11586 PR middle-end/84040
11587 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11590 2018-01-26 Jim Wilson <jimw@sifive.com>
11592 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11594 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11597 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11599 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11600 and CMP + SUB-immediate -> SUBS.
11602 2018-01-26 Martin Sebor <msebor@redhat.com>
11604 PR tree-optimization/83896
11605 * tree-ssa-strlen.c (get_string_len): Rename...
11606 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11607 Avoid assuming length is constant.
11608 (handle_char_store): Use HOST_WIDE_INT for string length.
11610 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11613 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11614 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11616 2018-01-26 Richard Biener <rguenther@suse.de>
11618 PR rtl-optimization/84003
11619 * dse.c (record_store): Only record redundant stores when
11620 the earlier store aliases at least all accesses the later one does.
11622 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11624 PR rtl-optimization/83985
11625 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11626 REG_CFA_RESTORE insns.
11627 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11628 REG_CFA_RESTORE insns here.
11631 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11632 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11634 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11636 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11637 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11638 (arc_init): Likewise.
11639 (arc_override_options): Likewise.
11640 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11642 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11644 * config/arc/arc.h (TARGET_DBNZ): Define.
11645 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11646 properly set the tune attribute.
11647 (dbnz): Use TARGET_DBNZ guard.
11648 * config/arc/arc.opt (mtune): Add core3 option.
11650 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11652 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11653 recognize new pic like addresses.
11654 (arc_delegitimize_address): Clean up.
11656 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11658 * config/arc/arc-arches.def: Option mrf16 valid for all
11660 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11661 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11662 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11663 * config/arc/arc-tables.opt: Regenerate.
11664 * config/arc/arc.c (arc_conditional_register_usage): Handle
11665 reduced register file case.
11666 (arc_file_start): Set must have build attributes.
11667 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11668 mrf16 option value.
11669 * config/arc/arc.opt (mrf16): Add new option.
11670 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11671 * config/arc/genmultilib.awk: Handle new mrf16 option.
11672 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11673 * config/arc/t-multilib: Regenerate.
11674 * doc/invoke.texi (ARC Options): Document mrf16 option.
11676 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11678 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11679 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11680 (arc_attribute_table): Add 'secure_call' attribute.
11681 (arc_print_operand): Print secure call operand.
11682 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11684 (arc_is_secure_call_p): New function. * config/arc/arc.md
11685 (call_i): Add support for sjli instruction.
11686 (call_value_i): Likewise.
11687 * config/arc/constraints.md (Csc): New constraint.
11689 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11690 John Eric Martin <John.Martin@emmicro-us.com>
11692 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11693 * config/arc/arc.c (_arc_jli_section): New struct.
11694 (arc_jli_section): New type.
11695 (rc_jli_sections): New static variable.
11696 (arc_handle_jli_attribute): New function.
11697 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11698 (arc_file_end): New function.
11699 (TARGET_ASM_FILE_END): Define.
11700 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11701 (arc_add_jli_section): New function.
11702 (jli_call_scan): Likewise.
11703 (arc_reorg): Call jli_call_scan.
11704 (arc_output_addsi): Remove 'S' from printing asm operand.
11705 (arc_is_jli_call_p): New function.
11706 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11708 (movhi_insn): Likewise.
11709 (movsi_insn): Likewise.
11710 (movsi_set_cc_insn): Likewise.
11711 (loadqi_update): Likewise.
11712 (load_zeroextendqisi_update): Likewise.
11713 (load_signextendqisi_update): Likewise.
11714 (loadhi_update): Likewise.
11715 (load_zeroextendhisi_update): Likewise.
11716 (load_signextendhisi_update): Likewise.
11717 (loadsi_update): Likewise.
11718 (loadsf_update): Likewise.
11719 (movsicc_insn): Likewise.
11720 (bset_insn): Likewise.
11721 (bxor_insn): Likewise.
11722 (bclr_insn): Likewise.
11723 (bmsk_insn): Likewise.
11724 (bicsi3_insn): Likewise.
11725 (cmpsi_cc_c_insn): Likewise.
11726 (movsi_ne): Likewise.
11727 (movsi_cond_exec): Likewise.
11728 (clrsbsi2): Likewise.
11729 (norm_f): Likewise.
11739 (call_i): Remove 'S' asm letter, add jli instruction.
11740 (call_value_i): Likewise.
11741 * config/arc/arc.op (mjli-always): New option.
11742 * config/arc/constraints.md (Cji): New constraint.
11743 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11745 (subsf3_fpx): Likewise.
11746 (mulsf3_fpx): Likewise.
11747 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11749 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11750 function attrbutes.
11751 * doc/invoke.texi (ARC): Document mjli-always option.
11753 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11755 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11756 avoid addition with 0 and use incw and decw where possible.
11758 2018-01-26 Richard Biener <rguenther@suse.de>
11760 PR tree-optimization/81082
11761 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11762 association if it requires casting to unsigned.
11763 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11764 from fold_plusminus_mult_expr to catch important cases late when
11765 range info is available.
11767 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11769 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11770 * configure.ac (hidden_linkonce): New test.
11771 * configure: Regenerate.
11772 * config.in: Regenerate.
11774 2018-01-26 Julia Koval <julia.koval@intel.com>
11776 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11777 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11778 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11779 _mm_mask_bitshuffle_epi64_mask): Fix type.
11780 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11781 USI_FTYPE_V4DI_V4DI_USI): Remove.
11782 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11783 __builtin_ia32_vpshufbitqmb256_mask,
11784 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11785 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11786 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11788 2018-01-26 Alan Modra <amodra@gmail.com>
11791 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11792 UNSPEC_VBPERMQ. Sort other unspecs.
11794 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11796 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11798 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11800 PR middle-end/83055
11801 * predict.c (drop_profile): Do not push/pop cfun; update also
11803 (handle_missing_profiles): Fix logic looking for zero profiles.
11805 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11807 PR middle-end/83977
11808 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11809 on functions with #pragma omp declare simd or functions with simd
11811 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11812 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11813 Remove trailing \n from warning_at calls.
11815 2018-01-25 Tom de Vries <tom@codesourcery.com>
11818 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11819 for neutered workers.
11821 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11824 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11825 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11827 2018-01-24 Jeff Law <law@redhat.com>
11830 * i386.c (get_probe_interval): Move to earlier point.
11831 (ix86_compute_frame_layout): If -fstack-clash-protection and
11832 the frame is larger than the probe interval, then use pushes
11833 to save registers rather than reg->mem moves.
11834 (ix86_expand_prologue): Remove conditional for int_registers_saved
11837 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11840 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11841 min/max for never referenced object.
11843 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11845 PR middle-end/83977
11846 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11848 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11849 attributes from DECL_ATTRIBUTES (decl) without affecting
11850 DECL_ATTRIBUTES (current_function_decl).
11851 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11852 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11854 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11856 PR tree-optimization/83979
11857 * fold-const.c (fold_comparison): Use constant_boolean_node
11858 instead of boolean_{true,false}_node.
11860 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11862 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11865 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11867 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11868 Simplify the clause that sets the length attribute.
11869 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11870 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11871 clause that sets the length attribute.
11872 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11874 2018-01-24 Tom de Vries <tom@codesourcery.com>
11877 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11878 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11879 Add strict parameter.
11880 (prevent_branch_around_nothing): Insert dummy insn between branch to
11881 label and label with no ptx insn inbetween.
11882 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11884 2018-01-24 Tom de Vries <tom@codesourcery.com>
11887 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11888 for neutered threads in warp.
11889 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11891 2018-01-24 Richard Biener <rguenther@suse.de>
11893 PR tree-optimization/83176
11894 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11897 2018-01-24 Richard Biener <rguenther@suse.de>
11899 PR tree-optimization/82819
11900 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11901 code generating pluses that are no-ops in the target precision.
11903 2018-01-24 Richard Biener <rguenther@suse.de>
11905 PR middle-end/84000
11906 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11908 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11910 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11911 to merge probabilities.
11912 * predict.c (probably_never_executed): Also mark as cold functions
11913 with global 0 profile and guessed local profile.
11914 * profile-count.c (profile_probability::combine_with_count): New
11916 * profile-count.h (profile_probability::operator*,
11917 profile_probability::operator*=, profile_probability::operator/,
11918 profile_probability::operator/=): Reduce precision to adjusted
11919 and set value to guessed on contradictory divisions.
11920 (profile_probability::combine_with_freq): Remove.
11921 (profile_probability::combine_wiht_count): Declare.
11922 (profile_count::force_nonzero):: Set to adjusted.
11923 (profile_count::probability_in):: Set quality to adjusted.
11924 * tree-ssa-tail-merge.c (replace_block_by): Use
11925 combine_with_count.
11927 2018-01-23 Andrew Waterman <andrew@sifive.com>
11928 Jim Wilson <jimw@sifive.com>
11930 * config/riscv/riscv.c (riscv_stack_boundary): New.
11931 (riscv_option_override): Set riscv_stack_boundary. Handle
11932 riscv_preferred_stack_boundary_arg.
11933 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11934 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11935 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11936 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11937 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11938 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11940 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11943 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11944 of struct ix86_frame.
11945 (ix86_expand_epilogue): Likewise. Add a local variable for
11946 the reg_save_offset field in struct ix86_frame.
11948 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11950 PR tree-optimization/82604
11951 * tree-loop-distribution.c (enum partition_kind): New enum item
11952 PKIND_PARTIAL_MEMSET.
11953 (partition_builtin_p): Support above new enum item.
11954 (generate_code_for_partition): Ditto.
11955 (compute_access_range): Differentiate cases that equality can be
11956 proven at all loops, the innermost loops or no loops.
11957 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11958 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11959 (finalize_partitions, distribute_loop): Don't fuse partition of
11960 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11961 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11962 parloop is enabled.
11964 2018-01-23 Martin Liska <mliska@suse.cz>
11966 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11967 order to ignore the predictor.
11968 (PRED_POLYMORPHIC_CALL): Likewise.
11969 (PRED_RECURSIVE_CALL): Likewise.
11971 2018-01-23 Martin Liska <mliska@suse.cz>
11973 * tree-profile.c (tree_profiling): Print function header to
11974 aware reader which function we are working on.
11975 * value-prof.c (gimple_find_values_to_profile): Do not print
11976 not interesting value histograms.
11978 2018-01-23 Martin Liska <mliska@suse.cz>
11980 * profile-count.h (enum profile_quality): Add
11981 profile_uninitialized as the first value. Do not number values
11982 as they are zero based.
11983 (profile_count::verify): Update sanity check.
11984 (profile_probability::verify): Likewise.
11986 2018-01-23 Nathan Sidwell <nathan@acm.org>
11988 * doc/invoke.texi (ffor-scope): Deprecate.
11990 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11992 PR tree-optimization/83510
11993 * domwalk.c (set_all_edges_as_executable): New function.
11994 (dom_walker::dom_walker): Convert bool param
11995 "skip_unreachable_blocks" to enum reachability. Move setup of
11996 edge flags to set_all_edges_as_executable and only do it when
11997 reachability is REACHABLE_BLOCKS.
11998 * domwalk.h (enum dom_walker::reachability): New enum.
11999 (dom_walker::dom_walker): Convert bool param
12000 "skip_unreachable_blocks" to enum reachability.
12001 (set_all_edges_as_executable): New decl.
12002 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12003 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12005 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12006 but converting true to REACHABLE_BLOCKS.
12007 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12009 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12010 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12011 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12013 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12014 if check_all_array_refs will be called.
12016 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12018 * tree.c (selftest::test_location_wrappers): Add more test
12021 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12023 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12024 (selftest::test_bit_in_range): Likewise.
12026 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12029 * doc/sourcebuild.texi (vect_float): Say that the selector
12030 only describes the situation when -funsafe-math-optimizations is on.
12031 (vect_float_strict): Document.
12033 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12035 PR tree-optimization/83965
12036 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12037 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12038 instead of checking only for a reduction.
12039 (vect_recog_widen_sum_pattern): Likewise.
12041 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12043 * predict.c (probably_never_executed): Only use precise profile info.
12044 (compute_function_frequency): Skip after inlining hack since we now
12045 have quality checking.
12047 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12049 * profile-count.h (profile_probability::very_unlikely,
12050 profile_probability::unlikely, profile_probability::even): Set
12051 precision to guessed.
12053 2018-01-23 Richard Biener <rguenther@suse.de>
12055 PR tree-optimization/83963
12056 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12057 Properly terminate dominator walk when crossing the exit edge not
12058 when visiting its source block.
12060 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12063 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12064 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12066 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12068 PR tree-optimization/83957
12069 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12070 semicolon after for body surrounded by braces.
12072 PR tree-optimization/83081
12073 * profile-count.h (profile_probability::split): New method.
12074 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12075 Use profile_probability::split.
12076 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12077 when splitting a single conditional jump into 2.
12079 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12081 PR tree-optimization/69452
12082 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12085 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12087 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12088 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12089 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12091 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12093 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12095 * config/rl78/rl78.md (movdi): New define_expand.
12096 * config/rl78/rl78.c (rl78_split_movdi): New function.
12098 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12101 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12103 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12104 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12105 128-bit to produce an UNSPEC move to get the double word with the
12106 signbit and then a shift directly to do signbit.
12107 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12108 implementation with a new version that just does either a direct
12109 move or a regular move. Move memory interface to separate insns.
12110 Move insns so they are next to the expander.
12111 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12112 with signbit move. Split big and little endian case.
12113 (signbit<mode>2_dm_mem_le): Likewise.
12114 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12115 (signbit<mode>2_dm2): Likewise.
12117 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12119 * config/rl78/rl78.md (anddi3): New define_expand.
12121 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12123 * config/rl78/rl78.md (umindi3): New define_expand.
12125 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12127 * config/rl78/rl78.md (smindi3): New define_expand.
12129 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12131 * config/rl78/rl78.md (smaxdi3): New define_expand.
12133 2018-01-22 Carl Love <cel@us.ibm.com>
12135 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12136 LVX_V1TI): Add macro expansion.
12137 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12138 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12139 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12140 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12141 Change check to determine if the instruction is a byte reversing
12142 entry. Fix typo in comment.
12143 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12144 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12145 Add def_builtin calls for new builtins.
12146 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12147 Add define_insn expansion.
12149 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12151 * config/rl78/rl78.md (umaxdi3): New define_expand.
12153 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12155 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12156 for non-QImode registers.
12158 2018-01-22 Richard Biener <rguenther@suse.de>
12160 PR tree-optimization/83963
12161 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12162 including the loop exit block.
12163 (scop_detection::merge_sese): Likewise.
12164 (scop_detection::add_scop): Do it here instead.
12166 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12168 * doc/sourcebuild.texi (arm_softfloat): Document.
12170 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12173 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12174 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12175 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12177 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12178 David Edelsohn <dje.gcc@gmail.com>
12181 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12182 Change "crset eq" to "crset 2".
12183 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12184 (*call_indirect_aix<mode>_nospec): Likewise.
12185 (*call_value_indirect_aix<mode>_nospec): Likewise.
12186 (*call_indirect_elfv2<mode>_nospec): Likewise.
12187 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12188 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12189 change assembly output from . to $.
12190 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12191 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12192 (*tablejump<mode>_internal1_nospec): Likewise.
12194 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12197 * config/sh/sh_optimize_sett_clrt.cc:
12198 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12200 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12202 PR tree-optimization/83940
12203 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12204 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12205 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12206 use it to pass back the definition type.
12207 (vect_check_store_rhs): Likewise rhs_dt_out.
12208 (vect_build_gather_load_calls): Add a mask_dt argument and use
12209 it instead of a call to vect_is_simple_use.
12210 (vectorizable_store): Update calls to vect_check_load_store_mask
12211 and vect_check_store_rhs. Use the dt returned by the latter instead
12212 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12213 instead of calls to vect_is_simple_use. Pass the scalar rather
12214 than the vector operand to vect_is_simple_use when handling
12215 second and subsequent copies of an rhs value.
12216 (vectorizable_load): Update calls to vect_check_load_store_mask
12217 and vect_build_gather_load_calls. Use the cached mask_dt and
12218 gs_info.offset_dt instead of calls to vect_is_simple_use.
12220 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12222 PR middle-end/83945
12223 * tree-emutls.c: Include gimplify.h.
12224 (lower_emutls_2): New function.
12225 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12226 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12227 it before further processing.
12230 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12231 UINTVAL (trueop1) instead of INTVAL (op1).
12233 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12237 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12238 INCOMING_FRAME_SP_OFFSET if not defined.
12239 (scan_trace): Add ENTRY argument. If true and
12240 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12241 emit a note to adjust the CFA offset.
12242 (create_cfi_notes): Adjust scan_trace callers.
12243 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12244 INCOMING_FRAME_SP_OFFSET in the CIE.
12245 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12246 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12248 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12249 * doc/tm.texi: Regenerated.
12251 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12253 PR rtl-optimization/83147
12254 * lra-constraints.c (remove_inheritance_pseudos): Use
12255 lra_substitute_pseudo_within_insn.
12257 2018-01-19 Tom de Vries <tom@codesourcery.com>
12258 Cesar Philippidis <cesar@codesourcery.com>
12261 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12263 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12266 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12267 spaces for function labels.
12269 2018-01-19 Martin Liska <mliska@suse.cz>
12271 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12272 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12273 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12274 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12275 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12276 (PRED_CONST_RETURN): Change from 69 to 65.
12277 (PRED_NULL_RETURN): Change from 91 to 71.
12278 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12279 (PRED_LOOP_GUARD): Change from 66 to 73.
12281 2018-01-19 Martin Liska <mliska@suse.cz>
12283 * predict.c (predict_insn_def): Add new assert.
12284 (struct branch_predictor): Change type to signed integer.
12285 (test_prediction_value_range): Amend test to cover
12286 PROB_UNINITIALIZED.
12287 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12288 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12289 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12290 (PRED_LOOP_IV_COMPARE): Likewise.
12291 * predict.h (PROB_UNINITIALIZED): Define new constant.
12293 2018-01-19 Martin Liska <mliska@suse.cz>
12295 * predict.c (dump_prediction): Add new format for
12296 analyze_brprob.py script which is enabled with -details
12298 * profile-count.h (precise_p): New function.
12300 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12302 PR tree-optimization/83922
12303 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12304 there are no statements that need masking.
12305 (vect_active_double_reduction_p): New function.
12306 (vect_analyze_loop_operations): Use it when handling phis that
12307 are not in the loop header.
12309 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12311 PR tree-optimization/83914
12312 * tree-vect-loop.c (vectorizable_induction): Don't convert
12313 init_expr or apply the peeling adjustment for inductions
12314 that are nested within the vectorized loop.
12316 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12318 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12321 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12325 * function.h (gimplify_parameters): Add gimple_seq * argument.
12326 * function.c: Include gimple.h and options.h.
12327 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12328 for the added local temporaries if needed.
12329 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12330 if there are any parameter cleanups, wrap whole body into a
12331 try/finally with the cleanups.
12333 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12336 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12337 Use GET_MODE_CLASS for scalar floating point.
12339 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12343 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12344 Fix call of call_cgraph_insertion_hooks.
12346 2018-01-18 Martin Sebor <msebor@redhat.com>
12348 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12350 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12353 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12356 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12359 * common.opt: (-ffile-prefix-map): New option.
12360 * opts.c (common_handle_option): Defer it.
12361 * opts-global.c (handle_common_deferred_options): Handle it.
12362 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12363 * file-prefix-map.h: New file.
12364 (remap_debug_filename, add_debug_prefix_map): ...here.
12365 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12366 * final.c (debug_prefix_map, add_debug_prefix_map
12367 remap_debug_filename): Move to...
12368 * file-prefix-map.c: New file.
12369 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12370 generalize, get rid of alloca(), use strrchr() instead of strchr().
12371 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12372 Implement in terms of add_prefix_map().
12373 (remap_macro_filename, remap_debug_filename): Implement in term of
12375 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12376 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12377 * dbxout.c: Include file-prefix-map.h.
12378 * varasm.c: Likewise.
12379 * vmsdbgout.c: Likewise.
12380 * xcoffout.c: Likewise.
12381 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12382 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12383 * doc/invoke.texi (-ffile-prefix-map): Document.
12384 (-fdebug-prefix-map): Update description.
12386 2018-01-18 Martin Liska <mliska@suse.cz>
12388 * config/i386/i386.c (indirect_thunk_name): Document that also
12390 (output_indirect_thunk): Document why both instructions
12391 (pause and lfence) are generated.
12393 2018-01-18 Richard Biener <rguenther@suse.de>
12395 PR tree-optimization/83887
12396 * graphite-scop-detection.c
12397 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12398 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12399 (scop_detection::merge_sese): Re-implement with a flood-fill
12400 algorithm that properly finds a SESE region if it exists.
12402 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12405 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12406 pointer_diff optimizations use view_convert instead of convert.
12408 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12410 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12411 Generate different code for -mno-speculate-indirect-jumps.
12412 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12413 (*call_indirect_aix<mode>): Disable for
12414 -mno-speculate-indirect-jumps.
12415 (*call_indirect_aix<mode>_nospec): New define_insn.
12416 (*call_value_indirect_aix<mode>): Disable for
12417 -mno-speculate-indirect-jumps.
12418 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12419 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12420 -mno-speculate-indirect-jumps.
12421 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12423 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12425 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12426 long double type, set the flags for noting the default long double
12427 type, even if we don't pass or return a long double type.
12429 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12432 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12435 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12437 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12438 support for merge[hl].
12439 (fold_mergehl_helper): New helper function.
12440 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12441 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12442 (altivec_vmrglw_direct): Add xxmrglw insn.
12444 2018-01-17 Andrew Waterman <andrew@sifive.com>
12446 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12447 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12449 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12452 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12453 call the lto_location_cache before reading the
12454 DECL_SOURCE_LOCATION of the types.
12456 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12457 Richard Sandiford <richard.sandiford@linaro.org>
12459 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12460 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12461 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12462 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12463 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12465 * config/aarch64/constraints.md (aarch64_movti_operand):
12467 * config/aarch64/predicates.md (Uti): Add new constraint.
12469 2018-01-17 Carl Love <cel@us.ibm.com>
12471 * config/rs6000/vsx.md (define_expand xl_len_r,
12472 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12473 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12475 (define_expand, define_insn): Move the shift left from the
12476 define_insn to the define_expand for lxvl and stxvl instructions.
12477 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12478 and XL_LEN_R definitions to PURE.
12480 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12482 * config/i386/i386.c (indirect_thunk_name): Declare regno
12483 as unsigned int. Compare regno with INVALID_REGNUM.
12484 (output_indirect_thunk): Ditto.
12485 (output_indirect_thunk_function): Ditto.
12486 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12487 in the call to output_indirect_thunk_function.
12489 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12491 PR middle-end/83884
12492 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12493 rather than the size of inner_type to determine the stack slot size
12494 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12496 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12499 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12502 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12504 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12505 endian Linux systems to optionally enable multilibs for selecting
12506 the long double type if the user configured an explicit type.
12507 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12508 have no long double multilibs if not defined.
12509 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12510 warn if the user used -mabi={ieee,ibm}longdouble and we built
12511 multilibs for long double.
12512 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12513 appropriate multilib option.
12514 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12516 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12517 for building long double multilibs.
12518 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12520 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12522 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12525 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12527 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12530 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12533 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12536 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12538 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12539 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12541 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12543 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12544 different rtl trees depending on TARGET_64BIT.
12545 (rs6000_gen_lvx): Likewise.
12547 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12549 * config/visium/visium.md (nop): Tweak comment.
12550 (hazard_nop): Likewise.
12552 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12554 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12555 -mspeculate-indirect-jumps.
12556 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12557 for -mno-speculate-indirect-jumps.
12558 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12559 (*call_value_indirect_elfv2<mode>): Disable for
12560 -mno-speculate-indirect-jumps.
12561 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12562 (indirect_jump): Emit different RTL for
12563 -mno-speculate-indirect-jumps.
12564 (*indirect_jump<mode>): Disable for
12565 -mno-speculate-indirect-jumps.
12566 (*indirect_jump<mode>_nospec): New define_insn.
12567 (tablejump): Emit different RTL for
12568 -mno-speculate-indirect-jumps.
12569 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12570 (tablejumpsi_nospec): New define_expand.
12571 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12572 (tablejumpdi_nospec): New define_expand.
12573 (*tablejump<mode>_internal1): Disable for
12574 -mno-speculate-indirect-jumps.
12575 (*tablejump<mode>_internal1_nospec): New define_insn.
12576 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12579 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12581 * caller-save.c (insert_save): Drop unnecessary parameter. All
12584 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12585 Richard Biener <rguenth@suse.de>
12588 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12589 return early, inline manually is_gimple_sizepos. Make sure if we
12590 call gimplify_expr we don't end up with a gimple constant.
12591 * tree.c (variably_modified_type_p): Don't return true for
12592 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12593 * gimplify.h (is_gimple_sizepos): Remove.
12595 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12597 PR tree-optimization/83857
12598 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12599 vectorizable_live_operation for pure SLP statements.
12600 (vectorizable_live_operation): Handle PHIs.
12602 2018-01-16 Richard Biener <rguenther@suse.de>
12604 PR tree-optimization/83867
12605 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12606 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12608 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12611 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12612 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12613 If off is not INTEGER_CST, issue a may not be aligned warning
12614 rather than isn't aligned. Use isn%'t rather than isn't.
12615 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12617 <case MULT_EXPR>: Improve the case when bottom and one of the
12618 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12619 operand, in that case check if the other operand is multiple of
12620 bottom divided by the INTEGER_CST operand.
12622 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12625 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12626 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12627 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12628 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12629 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12630 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12631 (pa_function_arg_size): New function.
12633 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12635 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12636 in a separate statement.
12638 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12640 PR tree-optimization/83847
12641 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12642 group gathers and scatters.
12644 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12646 PR rtl-optimization/86620
12647 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12649 PR rtl-optimization/83213
12650 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12651 to last if both are JUMP_INSNs.
12653 PR tree-optimization/83843
12654 * gimple-ssa-store-merging.c
12655 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12656 store_immediate_info for bswap/nop orig_stores.
12658 2018-01-15 Andrew Waterman <andrew@sifive.com>
12660 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12662 <UDIV>: Increase cost if !TARGET_DIV.
12664 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12666 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12667 (define_attr "cr_logical_3op"): New.
12668 (cceq_ior_compare): Adjust.
12669 (cceq_ior_compare_complement): Adjust.
12670 (*cceq_rev_compare): Adjust.
12671 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12672 (is_cracked_insn): Adjust.
12673 (insn_must_be_first_in_group): Adjust.
12674 * config/rs6000/40x.md: Adjust.
12675 * config/rs6000/440.md: Adjust.
12676 * config/rs6000/476.md: Adjust.
12677 * config/rs6000/601.md: Adjust.
12678 * config/rs6000/603.md: Adjust.
12679 * config/rs6000/6xx.md: Adjust.
12680 * config/rs6000/7450.md: Adjust.
12681 * config/rs6000/7xx.md: Adjust.
12682 * config/rs6000/8540.md: Adjust.
12683 * config/rs6000/cell.md: Adjust.
12684 * config/rs6000/e300c2c3.md: Adjust.
12685 * config/rs6000/e500mc.md: Adjust.
12686 * config/rs6000/e500mc64.md: Adjust.
12687 * config/rs6000/e5500.md: Adjust.
12688 * config/rs6000/e6500.md: Adjust.
12689 * config/rs6000/mpc.md: Adjust.
12690 * config/rs6000/power4.md: Adjust.
12691 * config/rs6000/power5.md: Adjust.
12692 * config/rs6000/power6.md: Adjust.
12693 * config/rs6000/power7.md: Adjust.
12694 * config/rs6000/power8.md: Adjust.
12695 * config/rs6000/power9.md: Adjust.
12696 * config/rs6000/rs64.md: Adjust.
12697 * config/rs6000/titan.md: Adjust.
12699 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12701 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12702 ix86_indirect_branch_register logic.
12704 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12706 * config/i386/constraints.md (Bs): Update
12707 ix86_indirect_branch_register check. Don't check
12708 ix86_indirect_branch_register with GOT_memory_operand.
12710 * config/i386/predicates.md (GOT_memory_operand): Don't check
12711 ix86_indirect_branch_register here.
12712 (GOT32_symbol_operand): Likewise.
12714 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12716 * config/i386/predicates.md (constant_call_address_operand):
12717 Rewrite ix86_indirect_branch_register logic.
12718 (sibcall_insn_operand): Likewise.
12720 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12722 * config/i386/constraints.md (Bs): Replace
12723 ix86_indirect_branch_thunk_register with
12724 ix86_indirect_branch_register.
12726 * config/i386/i386.md (indirect_jump): Likewise.
12727 (tablejump): Likewise.
12728 (*sibcall_memory): Likewise.
12729 (*sibcall_value_memory): Likewise.
12730 Peepholes of indirect call and jump via memory: Likewise.
12731 * config/i386/i386.opt: Likewise.
12732 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12733 (GOT_memory_operand): Likewise.
12734 (call_insn_operand): Likewise.
12735 (sibcall_insn_operand): Likewise.
12736 (GOT32_symbol_operand): Likewise.
12738 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12740 PR middle-end/83837
12741 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12742 type rather than type addr's type points to.
12743 (expand_omp_atomic_mutex): Likewise.
12744 (expand_omp_atomic): Likewise.
12746 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12749 * config/i386/i386.c (output_indirect_thunk_function): Use
12750 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12751 for __x86_return_thunk.
12753 2018-01-15 Richard Biener <rguenther@suse.de>
12755 PR middle-end/83850
12756 * expmed.c (extract_bit_field_1): Fix typo.
12758 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12761 * config/arm/iterators.md (VF): New mode iterator.
12762 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12763 Remove integer-related logic from pattern.
12764 (neon_vabd<mode>_3): Likewise.
12766 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12768 PR middle-end/82694
12769 * common.opt (fstrict-overflow): No longer an alias.
12770 (fwrapv-pointer): New option.
12771 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12772 also for pointer types based on flag_wrapv_pointer.
12773 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12774 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12775 opts->x_flag_wrapv got set.
12776 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12777 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12778 POINTER_TYPE_OVERFLOW_UNDEFINED.
12779 * match.pd: Likewise in address comparison pattern.
12780 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12782 2018-01-15 Richard Biener <rguenther@suse.de>
12785 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12786 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12787 Reset type names to their identifier if their TYPE_DECL doesn't
12788 have linkage (and thus is used for ODR and devirt).
12789 (save_debug_info_for_decl): Remove.
12790 (save_debug_info_for_type): Likewise.
12791 (add_tree_to_fld_list): Adjust.
12792 * tree-pretty-print.c (dump_generic_node): Make dumping of
12793 type names more robust.
12795 2018-01-15 Richard Biener <rguenther@suse.de>
12797 * BASE-VER: Bump to 8.0.1.
12799 2018-01-14 Martin Sebor <msebor@redhat.com>
12802 * builtins.c (check_access): Avoid warning when the no-warning bit
12805 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12807 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12808 * ira-color (allocno_hard_regs_compare): Likewise.
12810 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12813 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12814 Use .pushsection/.popsection.
12816 2018-01-14 Martin Sebor <msebor@redhat.com>
12819 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12821 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12823 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12824 entry from extra_headers.
12825 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12826 extra_headers, make the list bitwise identical to the i?86-*-* one.
12828 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12830 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12831 -mcmodel=large with -mindirect-branch=thunk,
12832 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12833 -mfunction-return=thunk-extern.
12834 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12835 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12836 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12838 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12840 * config/i386/i386.c (print_reg): Print the name of the full
12841 integer register without '%'.
12842 (ix86_print_operand): Handle 'V'.
12843 * doc/extend.texi: Document 'V' modifier.
12845 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12847 * config/i386/constraints.md (Bs): Disallow memory operand for
12848 -mindirect-branch-register.
12850 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12851 (GOT_memory_operand): Likewise.
12852 (call_insn_operand): Likewise.
12853 (sibcall_insn_operand): Likewise.
12854 (GOT32_symbol_operand): Likewise.
12855 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12856 for -mindirect-branch-register.
12857 (tablejump): Likewise.
12858 (*sibcall_memory): Likewise.
12859 (*sibcall_value_memory): Likewise.
12860 Disallow peepholes of indirect call and jump via memory for
12861 -mindirect-branch-register.
12862 (*call_pop): Replace m with Bw.
12863 (*call_value_pop): Likewise.
12864 (*sibcall_pop_memory): Replace m with Bs.
12865 * config/i386/i386.opt (mindirect-branch-register): New option.
12866 * doc/invoke.texi: Document -mindirect-branch-register option.
12868 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12870 * config/i386/i386-protos.h (ix86_output_function_return): New.
12871 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12872 set function_return_type.
12873 (indirect_thunk_name): Add ret_p to indicate thunk for function
12875 (output_indirect_thunk_function): Pass false to
12876 indirect_thunk_name.
12877 (ix86_output_indirect_branch_via_reg): Likewise.
12878 (ix86_output_indirect_branch_via_push): Likewise.
12879 (output_indirect_thunk_function): Create alias for function
12880 return thunk if regno < 0.
12881 (ix86_output_function_return): New function.
12882 (ix86_handle_fndecl_attribute): Handle function_return.
12883 (ix86_attribute_table): Add function_return.
12884 * config/i386/i386.h (machine_function): Add
12885 function_return_type.
12886 * config/i386/i386.md (simple_return_internal): Use
12887 ix86_output_function_return.
12888 (simple_return_internal_long): Likewise.
12889 * config/i386/i386.opt (mfunction-return=): New option.
12890 (indirect_branch): Mention -mfunction-return=.
12891 * doc/extend.texi: Document function_return function attribute.
12892 * doc/invoke.texi: Document -mfunction-return= option.
12894 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12896 * config/i386/i386-opts.h (indirect_branch): New.
12897 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12898 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12899 with local indirect jump when converting indirect call and jump.
12900 (ix86_set_indirect_branch_type): New.
12901 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12902 (indirectlabelno): New.
12903 (indirect_thunk_needed): Likewise.
12904 (indirect_thunk_bnd_needed): Likewise.
12905 (indirect_thunks_used): Likewise.
12906 (indirect_thunks_bnd_used): Likewise.
12907 (INDIRECT_LABEL): Likewise.
12908 (indirect_thunk_name): Likewise.
12909 (output_indirect_thunk): Likewise.
12910 (output_indirect_thunk_function): Likewise.
12911 (ix86_output_indirect_branch_via_reg): Likewise.
12912 (ix86_output_indirect_branch_via_push): Likewise.
12913 (ix86_output_indirect_branch): Likewise.
12914 (ix86_output_indirect_jmp): Likewise.
12915 (ix86_code_end): Call output_indirect_thunk_function if needed.
12916 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12918 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12919 (ix86_attribute_table): Add indirect_branch.
12920 * config/i386/i386.h (machine_function): Add indirect_branch_type
12921 and has_local_indirect_jump.
12922 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12924 (tablejump): Likewise.
12925 (*indirect_jump): Use ix86_output_indirect_jmp.
12926 (*tablejump_1): Likewise.
12927 (simple_return_indirect_internal): Likewise.
12928 * config/i386/i386.opt (mindirect-branch=): New option.
12929 (indirect_branch): New.
12932 (thunk-inline): Likewise.
12933 (thunk-extern): Likewise.
12934 * doc/extend.texi: Document indirect_branch function attribute.
12935 * doc/invoke.texi: Document -mindirect-branch= option.
12937 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12940 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12942 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12944 * ipa-inline.c (want_inline_small_function_p): Return false if
12945 inlining has already failed with CIF_FINAL_ERROR.
12946 (update_caller_keys): Call want_inline_small_function_p before
12948 (update_callee_keys): Likewise.
12950 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12952 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12954 (rs6000_quadword_masked_address_p): Likewise.
12955 (quad_aligned_load_p): Likewise.
12956 (quad_aligned_store_p): Likewise.
12957 (const_load_sequence_p): Add comment to describe the outer-most loop.
12958 (mimic_memory_attributes_and_flags): New function.
12959 (rs6000_gen_stvx): Likewise.
12960 (replace_swapped_aligned_store): Likewise.
12961 (rs6000_gen_lvx): Likewise.
12962 (replace_swapped_aligned_load): Likewise.
12963 (replace_swapped_load_constant): Capitalize argument name in
12964 comment describing this function.
12965 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12966 and stores that access quad-word aligned addresses and replace
12967 with stvx or lvx instructions when appropriate.
12968 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12969 New function prototype.
12970 (rs6000_quadword_masked_address_p): Likewise.
12971 (rs6000_gen_lvx): Likewise.
12972 (rs6000_gen_stvx): Likewise.
12973 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12974 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12975 when memory address is aligned.
12976 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12977 this split to select lvx instruction when memory address is aligned.
12978 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12979 instruction when memory address is aligned.
12980 (*vsx_le_perm_load_v16qi): Likewise.
12981 (four unnamed splitters): Modify to select the stvx instruction
12982 when memory is aligned.
12984 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12986 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12987 which appears in the queue multiple times.
12989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12990 Alan Hayward <alan.hayward@arm.com>
12991 David Sherwood <david.sherwood@arm.com>
12993 * tree-vectorizer.h (vec_lower_bound): New structure.
12994 (_loop_vec_info): Add check_nonzero and lower_bounds.
12995 (LOOP_VINFO_CHECK_NONZERO): New macro.
12996 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12997 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12998 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12999 fields. Make seg_len the distance travelled, not including the
13001 (dr_direction_indicator): Declare.
13002 (dr_zero_step_indicator): Likewise.
13003 (dr_known_forward_stride_p): Likewise.
13004 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13006 (runtime_alias_check_p): Allow runtime alias checks with
13008 (operator ==): Compare access_size and align.
13009 (prune_runtime_alias_test_list): Rework for new distinction between
13010 the access_size and seg_len.
13011 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13013 (get_segment_min_max): New function.
13014 (create_intersect_range_checks): Use it.
13015 (dr_step_indicator): New function.
13016 (dr_direction_indicator): Likewise.
13017 (dr_zero_step_indicator): Likewise.
13018 (dr_known_forward_stride_p): Likewise.
13019 * tree-loop-distribution.c (data_ref_segment_size): Return
13020 DR_STEP * (niters - 1).
13021 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13023 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13024 (vect_preserves_scalar_order_p): New function, split out from...
13025 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13026 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13027 (vect_vfa_access_size): New function.
13028 (vect_vfa_align): Likewise.
13029 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13030 (dump_lower_bound): New function.
13031 (vect_check_lower_bound): Likewise.
13032 (vect_small_gap_p): Likewise.
13033 (vectorizable_with_step_bound_p): Likewise.
13034 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13035 depencies if the vectorization factor is 1. Convert the checks
13036 for nonzero steps into checks on the bounds of DR_STEP. Try using
13037 a bunds check for variable steps if the minimum required step is
13038 relatively small. Update calls to the dr_with_seg_len
13039 constructor and to vect_compile_time_alias.
13040 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13042 (vect_loop_versioning): Call it.
13043 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13045 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13048 Alan Hayward <alan.hayward@arm.com>
13049 David Sherwood <david.sherwood@arm.com>
13051 * doc/sourcebuild.texi (vect_scatter_store): Document.
13052 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13054 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13056 * genopinit.c (main): Add supports_vec_scatter_store and
13057 supports_vec_scatter_store_cached to target_optabs.
13058 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13059 IFN_MASK_SCATTER_STORE.
13060 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13062 * internal-fn.h (internal_store_fn_p): Declare.
13063 (internal_fn_stored_value_index): Likewise.
13064 * internal-fn.c (scatter_store_direct): New macro.
13065 (expand_scatter_store_optab_fn): New function.
13066 (direct_scatter_store_optab_supported_p): New macro.
13067 (internal_store_fn_p): New function.
13068 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13069 IFN_MASK_SCATTER_STORE.
13070 (internal_fn_mask_index): Likewise.
13071 (internal_fn_stored_value_index): New function.
13072 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13073 for scatter stores.
13074 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13075 * optabs-query.c (supports_vec_scatter_store_p): New function.
13076 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13077 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13078 true for scatter stores.
13079 (vect_gather_scatter_fn_p): Handle scatter stores too.
13080 (vect_check_gather_scatter): Consider using scatter stores if
13081 supports_vec_scatter_store_p.
13082 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13083 scatter stores too.
13084 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13085 internal_fn_stored_value_index.
13086 (check_load_store_masking): Handle scatter stores too.
13087 (vect_get_store_rhs): Make public.
13088 (vectorizable_call): Use internal_store_fn_p.
13089 (vectorizable_store): Handle scatter store internal functions.
13090 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13091 when deciding whether the end of the group has been reached.
13092 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13093 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13094 (mask_scatter_store<mode>): New insns.
13096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13097 Alan Hayward <alan.hayward@arm.com>
13098 David Sherwood <david.sherwood@arm.com>
13100 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13101 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13102 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13104 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13105 Use vect_truncate_gather_scatter_offset if we can't treat the
13106 operation as a normal gather load or scatter store.
13107 (get_group_load_store_type): Take the gather_scatter_info
13108 as argument. Try using a gather load or scatter store for
13109 single-element groups.
13110 (get_load_store_type): Update calls to get_group_load_store_type
13111 and vect_use_strided_gather_scatters_p.
13113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13114 Alan Hayward <alan.hayward@arm.com>
13115 David Sherwood <david.sherwood@arm.com>
13117 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13118 optional tree argument.
13119 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13121 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13122 but continue to use the current value as a fallback.
13123 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13124 to compare the updates.
13125 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13126 (get_load_store_type): Use it when handling a strided access.
13127 (vect_get_strided_load_store_ops): New function.
13128 (vect_get_data_ptr_increment): Likewise.
13129 (vectorizable_load): Handle strided gather loads. Always pass
13130 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13133 Alan Hayward <alan.hayward@arm.com>
13134 David Sherwood <david.sherwood@arm.com>
13136 * doc/md.texi (gather_load@var{m}): Document.
13137 (mask_gather_load@var{m}): Likewise.
13138 * genopinit.c (main): Add supports_vec_gather_load and
13139 supports_vec_gather_load_cached to target_optabs.
13140 * optabs-tree.c (init_tree_optimization_optabs): Use
13141 ggc_cleared_alloc to allocate target_optabs.
13142 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13143 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13145 * internal-fn.h (internal_load_fn_p): Declare.
13146 (internal_gather_scatter_fn_p): Likewise.
13147 (internal_fn_mask_index): Likewise.
13148 (internal_gather_scatter_fn_supported_p): Likewise.
13149 * internal-fn.c (gather_load_direct): New macro.
13150 (expand_gather_load_optab_fn): New function.
13151 (direct_gather_load_optab_supported_p): New macro.
13152 (direct_internal_fn_optab): New function.
13153 (internal_load_fn_p): Likewise.
13154 (internal_gather_scatter_fn_p): Likewise.
13155 (internal_fn_mask_index): Likewise.
13156 (internal_gather_scatter_fn_supported_p): Likewise.
13157 * optabs-query.c (supports_at_least_one_mode_p): New function.
13158 (supports_vec_gather_load_p): Likewise.
13159 * optabs-query.h (supports_vec_gather_load_p): Declare.
13160 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13161 and memory_type field.
13162 (NUM_PATTERNS): Bump to 15.
13163 * tree-vect-data-refs.c: Include internal-fn.h.
13164 (vect_gather_scatter_fn_p): New function.
13165 (vect_describe_gather_scatter_call): Likewise.
13166 (vect_check_gather_scatter): Try using internal functions for
13167 gather loads. Recognize existing calls to a gather load function.
13168 (vect_analyze_data_refs): Consider using gather loads if
13169 supports_vec_gather_load_p.
13170 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13171 (vect_get_gather_scatter_offset_type): Likewise.
13172 (vect_convert_mask_for_vectype): Likewise.
13173 (vect_add_conversion_to_patterm): Likewise.
13174 (vect_try_gather_scatter_pattern): Likewise.
13175 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13176 (vect_vect_recog_func_ptrs): Add it.
13177 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13178 internal_fn_mask_index and internal_gather_scatter_fn_p.
13179 (check_load_store_masking): Take the gather_scatter_info as an
13180 argument and handle gather loads.
13181 (vect_get_gather_scatter_ops): New function.
13182 (vectorizable_call): Check internal_load_fn_p.
13183 (vectorizable_load): Likewise. Handle gather load internal
13185 (vectorizable_store): Update call to check_load_store_masking.
13186 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13187 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13188 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13189 (aarch64_gather_scale_operand_d): New predicates.
13190 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13191 (mask_gather_load<mode>): New insns.
13193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13194 Alan Hayward <alan.hayward@arm.com>
13195 David Sherwood <david.sherwood@arm.com>
13197 * optabs.def (fold_left_plus_optab): New optab.
13198 * doc/md.texi (fold_left_plus_@var{m}): Document.
13199 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13200 * internal-fn.c (fold_left_direct): Define.
13201 (expand_fold_left_optab_fn): Likewise.
13202 (direct_fold_left_optab_supported_p): Likewise.
13203 * fold-const-call.c (fold_const_fold_left): New function.
13204 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13205 * tree-parloops.c (valid_reduction_p): New function.
13206 (gather_scalar_reductions): Use it.
13207 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13208 (vect_finish_replace_stmt): Declare.
13209 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13210 (needs_fold_left_reduction_p): New function, split out from...
13211 (vect_is_simple_reduction): ...here. Accept reductions that
13212 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13213 (vect_force_simple_reduction): Also store the reduction type in
13214 the assignment's STMT_VINFO_REDUC_TYPE.
13215 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13216 (merge_with_identity): New function.
13217 (vect_expand_fold_left): Likewise.
13218 (vectorize_fold_left_reduction): Likewise.
13219 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13220 scalar phi in place for it. Check for target support and reject
13221 cases that would reassociate the operation. Defer the transform
13222 phase to vectorize_fold_left_reduction.
13223 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13224 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13225 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13229 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13230 call to ifc_temp_var.
13232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13233 Alan Hayward <alan.hayward@arm.com>
13234 David Sherwood <david.sherwood@arm.com>
13236 * target.def (legitimize_address_displacement): Take the original
13237 offset as a poly_int.
13238 * targhooks.h (default_legitimize_address_displacement): Update
13240 * targhooks.c (default_legitimize_address_displacement): Likewise.
13241 * doc/tm.texi: Regenerate.
13242 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13243 as an argument, moving assert of ad->disp == ad->disp_term to...
13244 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13245 Try calling targetm.legitimize_address_displacement before expanding
13246 the address rather than afterwards, and adjust for the new interface.
13247 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13248 Match the new hook interface. Handle SVE addresses.
13249 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13250 new hook interface.
13252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13254 * Makefile.in (OBJS): Add early-remat.o.
13255 * target.def (select_early_remat_modes): New hook.
13256 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13257 * doc/tm.texi: Regenerate.
13258 * targhooks.h (default_select_early_remat_modes): Declare.
13259 * targhooks.c (default_select_early_remat_modes): New function.
13260 * timevar.def (TV_EARLY_REMAT): New timevar.
13261 * passes.def (pass_early_remat): New pass.
13262 * tree-pass.h (make_pass_early_remat): Declare.
13263 * early-remat.c: New file.
13264 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13266 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13269 Alan Hayward <alan.hayward@arm.com>
13270 David Sherwood <david.sherwood@arm.com>
13272 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13273 vfm1 with a bound_epilog parameter.
13274 (vect_do_peeling): Update calls accordingly, and move the prologue
13275 call earlier in the function. Treat the base bound_epilog as 0 for
13276 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13277 this base when peeling for gaps.
13278 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13279 with fully-masked loops.
13280 (vect_estimate_min_profitable_iters): Handle the single peeled
13281 iteration in that case.
13283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13284 Alan Hayward <alan.hayward@arm.com>
13285 David Sherwood <david.sherwood@arm.com>
13287 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13288 single-element interleaving even if the size is not a power of 2.
13289 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13290 accesses for single-element interleaving if the group size is
13293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13294 Alan Hayward <alan.hayward@arm.com>
13295 David Sherwood <david.sherwood@arm.com>
13297 * doc/md.texi (fold_extract_last_@var{m}): Document.
13298 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13299 * optabs.def (fold_extract_last_optab): New optab.
13300 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13301 * internal-fn.c (fold_extract_direct): New macro.
13302 (expand_fold_extract_optab_fn): Likewise.
13303 (direct_fold_extract_optab_supported_p): Likewise.
13304 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13305 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13306 EXTRACT_LAST_REDUCTION.
13307 (get_initial_def_for_reduction): Do not create an initial vector
13308 for EXTRACT_LAST_REDUCTION reductions.
13309 (vectorizable_reduction): Leave the scalar phi in place for
13310 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13311 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13312 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13313 transform phase to vectorizable_condition.
13314 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13316 (vect_finish_stmt_generation): ...here.
13317 (vect_finish_replace_stmt): New function.
13318 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13319 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13321 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13324 Alan Hayward <alan.hayward@arm.com>
13325 David Sherwood <david.sherwood@arm.com>
13327 * doc/md.texi (extract_last_@var{m}): Document.
13328 * optabs.def (extract_last_optab): New optab.
13329 * internal-fn.def (EXTRACT_LAST): New internal function.
13330 * internal-fn.c (cond_unary_direct): New macro.
13331 (expand_cond_unary_optab_fn): Likewise.
13332 (direct_cond_unary_optab_supported_p): Likewise.
13333 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13334 loops using EXTRACT_LAST.
13335 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13336 (extract_last_<mode>): ...this optab.
13337 (vec_extract<mode><Vel>): Update accordingly.
13339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13340 Alan Hayward <alan.hayward@arm.com>
13341 David Sherwood <david.sherwood@arm.com>
13343 * target.def (empty_mask_is_expensive): New hook.
13344 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13345 * doc/tm.texi: Regenerate.
13346 * targhooks.h (default_empty_mask_is_expensive): Declare.
13347 * targhooks.c (default_empty_mask_is_expensive): New function.
13348 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13349 if the target says that empty masks are expensive.
13350 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13352 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13355 Alan Hayward <alan.hayward@arm.com>
13356 David Sherwood <david.sherwood@arm.com>
13358 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13359 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13360 (vect_use_loop_mask_for_alignment_p): New function.
13361 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13362 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13363 niters_skip argument. Make sure that the first niters_skip elements
13364 of the first iteration are inactive.
13365 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13366 Update call to vect_set_loop_masks_directly.
13367 (get_misalign_in_elems): New function, split out from...
13368 (vect_gen_prolog_loop_niters): ...here.
13369 (vect_update_init_of_dr): Take a code argument that specifies whether
13370 the adjustment should be added or subtracted.
13371 (vect_update_init_of_drs): Likewise.
13372 (vect_prepare_for_masked_peels): New function.
13373 (vect_do_peeling): Skip prologue peeling if we're using a mask
13374 instead. Update call to vect_update_inits_of_drs.
13375 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13377 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13378 alignment. Do not include the number of peeled iterations in
13379 the minimum threshold in that case.
13380 (vectorizable_induction): Adjust the start value down by
13381 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13382 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13383 Take the number of skipped iterations into account when calculating
13385 * tree-vect-stmts.c (vect_gen_while_not): New function.
13387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13388 Alan Hayward <alan.hayward@arm.com>
13389 David Sherwood <david.sherwood@arm.com>
13391 * doc/sourcebuild.texi (vect_fully_masked): Document.
13392 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13393 default value to 0.
13394 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13396 (vect_analyze_loop_2): ...here. Don't check the vectorization
13397 factor against the number of loop iterations if the loop is
13400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13401 Alan Hayward <alan.hayward@arm.com>
13402 David Sherwood <david.sherwood@arm.com>
13404 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13405 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13406 (dump_groups): Update accordingly.
13407 (iv_use::mem_type): New member variable.
13408 (address_p): New function.
13409 (record_use): Add a mem_type argument and initialize the new
13411 (record_group_use): Add a mem_type argument. Use address_p.
13412 Remove obsolete null checks of base_object. Update call to record_use.
13413 (find_interesting_uses_op): Update call to record_group_use.
13414 (find_interesting_uses_cond): Likewise.
13415 (find_interesting_uses_address): Likewise.
13416 (get_mem_type_for_internal_fn): New function.
13417 (find_address_like_use): Likewise.
13418 (find_interesting_uses_stmt): Try find_address_like_use before
13419 calling find_interesting_uses_op.
13420 (addr_offset_valid_p): Use the iv mem_type field as the type
13421 of the addressed memory.
13422 (add_autoinc_candidates): Likewise.
13423 (get_address_cost): Likewise.
13424 (split_small_address_groups_p): Use address_p.
13425 (split_address_groups): Likewise.
13426 (add_iv_candidate_for_use): Likewise.
13427 (autoinc_possible_for_pair): Likewise.
13428 (rewrite_groups): Likewise.
13429 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13430 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13431 (get_alias_ptr_type_for_ptr_address): New function.
13432 (rewrite_use_address): Rewrite address uses in calls that were
13433 identified by find_address_like_use.
13435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13436 Alan Hayward <alan.hayward@arm.com>
13437 David Sherwood <david.sherwood@arm.com>
13439 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13441 * gimple-expr.h (is_gimple_addressable: Likewise.
13442 * gimple-expr.c (is_gimple_address): Likewise.
13443 * internal-fn.c (expand_call_mem_ref): New function.
13444 (expand_mask_load_optab_fn): Use it.
13445 (expand_mask_store_optab_fn): Likewise.
13447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13448 Alan Hayward <alan.hayward@arm.com>
13449 David Sherwood <david.sherwood@arm.com>
13451 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13452 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13453 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13454 (cond_umax@var{mode}): Document.
13455 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13456 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13457 (cond_umin_optab, cond_umax_optab): New optabs.
13458 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13459 (COND_IOR, COND_XOR): New internal functions.
13460 * internal-fn.h (get_conditional_internal_fn): Declare.
13461 * internal-fn.c (cond_binary_direct): New macro.
13462 (expand_cond_binary_optab_fn): Likewise.
13463 (direct_cond_binary_optab_supported_p): Likewise.
13464 (get_conditional_internal_fn): New function.
13465 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13466 Cope with reduction statements that are vectorized as calls rather
13468 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13469 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13470 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13471 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13472 (UNSPEC_COND_EOR): New unspecs.
13473 (optab): Add mappings for them.
13474 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13475 (sve_int_op, sve_fp_op): New int attributes.
13477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13478 Alan Hayward <alan.hayward@arm.com>
13479 David Sherwood <david.sherwood@arm.com>
13481 * optabs.def (while_ult_optab): New optab.
13482 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13483 * internal-fn.def (WHILE_ULT): New internal function.
13484 * internal-fn.h (direct_internal_fn_supported_p): New override
13485 that takes two types as argument.
13486 * internal-fn.c (while_direct): New macro.
13487 (expand_while_optab_fn): New function.
13488 (convert_optab_supported_p): Likewise.
13489 (direct_while_optab_supported_p): New macro.
13490 * wide-int.h (wi::udiv_ceil): New function.
13491 * tree-vectorizer.h (rgroup_masks): New structure.
13492 (vec_loop_masks): New typedef.
13493 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13494 and fully_masked_p.
13495 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13496 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13497 (vect_max_vf): New function.
13498 (slpeel_make_loop_iterate_ntimes): Delete.
13499 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13500 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13501 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13502 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13503 internal-fn.h, stor-layout.h and optabs-query.h.
13504 (vect_set_loop_mask): New function.
13505 (add_preheader_seq): Likewise.
13506 (add_header_seq): Likewise.
13507 (interleave_supported_p): Likewise.
13508 (vect_maybe_permute_loop_masks): Likewise.
13509 (vect_set_loop_masks_directly): Likewise.
13510 (vect_set_loop_condition_masked): Likewise.
13511 (vect_set_loop_condition_unmasked): New function, split out from
13512 slpeel_make_loop_iterate_ntimes.
13513 (slpeel_make_loop_iterate_ntimes): Rename to..
13514 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13515 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13516 (vect_do_peeling): Update call accordingly.
13517 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13519 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13520 mask_compare_type, can_fully_mask_p and fully_masked_p.
13521 (release_vec_loop_masks): New function.
13522 (_loop_vec_info): Use it to free the loop masks.
13523 (can_produce_all_loop_masks_p): New function.
13524 (vect_get_max_nscalars_per_iter): Likewise.
13525 (vect_verify_full_masking): Likewise.
13526 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13527 retries, and free the mask rgroups before retrying. Check loop-wide
13528 reasons for disallowing fully-masked loops. Make the final decision
13529 about whether use a fully-masked loop or not.
13530 (vect_estimate_min_profitable_iters): Do not assume that peeling
13531 for the number of iterations will be needed for fully-masked loops.
13532 (vectorizable_reduction): Disable fully-masked loops.
13533 (vectorizable_live_operation): Likewise.
13534 (vect_halve_mask_nunits): New function.
13535 (vect_double_mask_nunits): Likewise.
13536 (vect_record_loop_mask): Likewise.
13537 (vect_get_loop_mask): Likewise.
13538 (vect_transform_loop): Handle the case in which the final loop
13539 iteration might handle a partial vector. Call vect_set_loop_condition
13540 instead of slpeel_make_loop_iterate_ntimes.
13541 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13542 (check_load_store_masking): New function.
13543 (prepare_load_store_mask): Likewise.
13544 (vectorizable_store): Handle fully-masked loops.
13545 (vectorizable_load): Likewise.
13546 (supportable_widening_operation): Use vect_halve_mask_nunits for
13548 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13549 (vect_gen_while): New function.
13550 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13551 (aarch64_uqdec<mode>): New insn.
13553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13554 Alan Hayward <alan.hayward@arm.com>
13555 David Sherwood <david.sherwood@arm.com>
13557 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13558 (reduc_xor_scal_optab): New optabs.
13559 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13560 (reduc_xor_scal_@var{m}): Document.
13561 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13562 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13563 internal functions.
13564 * fold-const-call.c (fold_const_call): Handle them.
13565 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13566 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13567 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13568 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13569 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13570 (UNSPEC_XORV): New unspecs.
13571 (optab): Add entries for them.
13572 (BITWISEV): New int iterator.
13573 (bit_reduc_op): New int attributes.
13575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13576 Alan Hayward <alan.hayward@arm.com>
13577 David Sherwood <david.sherwood@arm.com>
13579 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13580 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13581 * optabs.def (vec_shl_insert_optab): New optab.
13582 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13583 (duplicate_and_interleave): Likewise.
13584 * tree-vect-loop.c: Include internal-fn.h.
13585 (neutral_op_for_slp_reduction): New function, split out from
13586 get_initial_defs_for_reduction.
13587 (get_initial_def_for_reduction): Handle option 2 for variable-length
13588 vectors by loading the neutral value into a vector and then shifting
13589 the initial value into element 0.
13590 (get_initial_defs_for_reduction): Replace the code argument with
13591 the neutral value calculated by neutral_op_for_slp_reduction.
13592 Use gimple_build_vector for constant-length vectors.
13593 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13594 but the first group_size elements have a neutral value.
13595 Use duplicate_and_interleave otherwise.
13596 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13597 Update call to get_initial_defs_for_reduction. Handle SLP
13598 reductions for variable-length vectors by creating one vector
13599 result for each scalar result, with the elements associated
13600 with other scalar results stubbed out with the neutral value.
13601 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13602 Require IFN_VEC_SHL_INSERT for double reductions on
13603 variable-length vectors, or SLP reductions that have
13604 a neutral value. Require can_duplicate_and_interleave_p
13605 support for variable-length unchained SLP reductions if there
13606 is no neutral value, such as for MIN/MAX reductions. Also require
13607 the number of vector elements to be a multiple of the number of
13608 SLP statements when doing variable-length unchained SLP reductions.
13609 Update call to vect_create_epilog_for_reduction.
13610 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13611 and remove initial values.
13612 (duplicate_and_interleave): Make public.
13613 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13614 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13617 Alan Hayward <alan.hayward@arm.com>
13618 David Sherwood <david.sherwood@arm.com>
13620 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13621 (can_duplicate_and_interleave_p): New function.
13622 (vect_get_and_check_slp_defs): Take the vector of statements
13623 rather than just the current one. Remove excess parentheses.
13624 Restriction rejectinon of vect_constant_def and vect_external_def
13625 for variable-length vectors to boolean types, or types for which
13626 can_duplicate_and_interleave_p is false.
13627 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13628 (duplicate_and_interleave): New function.
13629 (vect_get_constant_vectors): Use gimple_build_vector for
13630 constant-length vectors and suitable variable-length constant
13631 vectors. Use duplicate_and_interleave for other variable-length
13632 vectors. Don't defer the update when inserting new statements.
13634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13635 Alan Hayward <alan.hayward@arm.com>
13636 David Sherwood <david.sherwood@arm.com>
13638 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13639 min_profitable_iters doesn't go negative.
13641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13642 Alan Hayward <alan.hayward@arm.com>
13643 David Sherwood <david.sherwood@arm.com>
13645 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13646 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13647 * optabs.def (vec_mask_load_lanes_optab): New optab.
13648 (vec_mask_store_lanes_optab): Likewise.
13649 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13650 (MASK_STORE_LANES): Likewise.
13651 * internal-fn.c (mask_load_lanes_direct): New macro.
13652 (mask_store_lanes_direct): Likewise.
13653 (expand_mask_load_optab_fn): Handle masked operations.
13654 (expand_mask_load_lanes_optab_fn): New macro.
13655 (expand_mask_store_optab_fn): Handle masked operations.
13656 (expand_mask_store_lanes_optab_fn): New macro.
13657 (direct_mask_load_lanes_optab_supported_p): Likewise.
13658 (direct_mask_store_lanes_optab_supported_p): Likewise.
13659 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13661 (vect_load_lanes_supported): Likewise.
13662 * tree-vect-data-refs.c (strip_conversion): New function.
13663 (can_group_stmts_p): Likewise.
13664 (vect_analyze_data_ref_accesses): Use it instead of checking
13665 for a pair of assignments.
13666 (vect_store_lanes_supported): Take a masked_p parameter.
13667 (vect_load_lanes_supported): Likewise.
13668 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13669 vect_store_lanes_supported and vect_load_lanes_supported.
13670 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13671 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13672 parameter. Don't allow gaps for masked accesses.
13673 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13674 and vect_load_lanes_supported.
13675 (get_load_store_type): Take a masked_p parameter and update
13676 call to get_group_load_store_type.
13677 (vectorizable_store): Update call to get_load_store_type.
13678 Handle IFN_MASK_STORE_LANES.
13679 (vectorizable_load): Update call to get_load_store_type.
13680 Handle IFN_MASK_LOAD_LANES.
13682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13683 Alan Hayward <alan.hayward@arm.com>
13684 David Sherwood <david.sherwood@arm.com>
13686 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13688 * config/aarch64/aarch64-protos.h
13689 (aarch64_sve_struct_memory_operand_p): Declare.
13690 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13691 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13692 (VPRED, vpred): Handle SVE structure modes.
13693 * config/aarch64/constraints.md (Utx): New constraint.
13694 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13695 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13696 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13697 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13698 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13699 structure modes. Split into pieces after RA.
13700 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13701 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13703 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13704 SVE structure modes.
13705 (aarch64_classify_address): Likewise.
13706 (sizetochar): Move earlier in file.
13707 (aarch64_print_operand): Handle SVE register lists.
13708 (aarch64_array_mode): New function.
13709 (aarch64_sve_struct_memory_operand_p): Likewise.
13710 (TARGET_ARRAY_MODE): Redefine.
13712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13713 Alan Hayward <alan.hayward@arm.com>
13714 David Sherwood <david.sherwood@arm.com>
13716 * target.def (array_mode): New target hook.
13717 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13718 * doc/tm.texi: Regenerate.
13719 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13720 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13721 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13722 targetm.array_mode.
13723 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13727 Alan Hayward <alan.hayward@arm.com>
13728 David Sherwood <david.sherwood@arm.com>
13730 * fold-const.c (fold_binary_loc): Check the argument types
13731 rather than the result type when testing for a vector operation.
13733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13735 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13736 * doc/tm.texi: Regenerate.
13738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13739 Alan Hayward <alan.hayward@arm.com>
13740 David Sherwood <david.sherwood@arm.com>
13742 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13743 (sve): Document new AArch64 extension.
13744 * doc/md.texi (w): Extend the description of the AArch64
13745 constraint to include SVE vectors.
13746 (Upl, Upa): Document new AArch64 predicate constraints.
13747 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13749 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13750 (msve-vector-bits=): New option.
13751 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13752 SVE when these are disabled.
13753 (sve): New extension.
13754 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13755 modes. Adjust their number of units based on aarch64_sve_vg.
13756 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13757 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13758 aarch64_addr_query_type.
13759 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13760 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13761 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13762 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13763 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13764 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13765 (aarch64_simd_imm_zero_p): Delete.
13766 (aarch64_check_zero_based_sve_index_immediate): Declare.
13767 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13768 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13769 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13770 (aarch64_sve_float_mul_immediate_p): Likewise.
13771 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13772 rather than an rtx.
13773 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13774 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13775 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13776 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13777 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13778 (aarch64_regmode_natural_size): Likewise.
13779 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13780 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13782 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13783 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13784 for VG and the SVE predicate registers.
13785 (V_ALIASES): Add a "z"-prefixed alias.
13786 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13787 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13788 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13789 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13790 (REG_CLASS_NAMES): Add entries for them.
13791 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13792 and the predicate registers.
13793 (aarch64_sve_vg): Declare.
13794 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13795 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13796 (REGMODE_NATURAL_SIZE): Define.
13797 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13799 * config/aarch64/aarch64.c: Include cfgrtl.h.
13800 (simd_immediate_info): Add a constructor for series vectors,
13801 and an associated step field.
13802 (aarch64_sve_vg): New variable.
13803 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13804 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13805 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13806 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13807 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13808 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13809 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13810 (aarch64_get_mask_mode): New functions.
13811 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13812 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13813 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13814 predicate modes and predicate registers. Explicitly restrict
13815 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13816 to store a vector mode if it is recognized by
13817 aarch64_classify_vector_mode.
13818 (aarch64_regmode_natural_size): New function.
13819 (aarch64_hard_regno_caller_save_mode): Return the original mode
13821 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13822 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13823 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13824 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13826 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13827 does not overlap dest if the function is frame-related. Handle
13829 (aarch64_split_add_offset): New function.
13830 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13831 them aarch64_add_offset.
13832 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13833 and update call to aarch64_sub_sp.
13834 (aarch64_add_cfa_expression): New function.
13835 (aarch64_expand_prologue): Pass extra temporary registers to the
13836 functions above. Handle the case in which we need to emit new
13837 DW_CFA_expressions for registers that were originally saved
13838 relative to the stack pointer, but now have to be expressed
13839 relative to the frame pointer.
13840 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13842 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13843 IP0 and IP1 values for SVE frames.
13844 (aarch64_expand_vec_series): New function.
13845 (aarch64_expand_sve_widened_duplicate): Likewise.
13846 (aarch64_expand_sve_const_vector): Likewise.
13847 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13848 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13849 into the register, rather than emitting a SET directly.
13850 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13851 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13852 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13853 (offset_9bit_signed_scaled_p): New functions.
13854 (aarch64_replicate_bitmask_imm): New function.
13855 (aarch64_bitmask_imm): Use it.
13856 (aarch64_cannot_force_const_mem): Reject expressions involving
13857 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13858 (aarch64_classify_index): Handle SVE indices, by requiring
13859 a plain register index with a scale that matches the element size.
13860 (aarch64_classify_address): Handle SVE addresses. Assert that
13861 the mode of the address is VOIDmode or an integer mode.
13862 Update call to aarch64_classify_symbol.
13863 (aarch64_classify_symbolic_expression): Update call to
13864 aarch64_classify_symbol.
13865 (aarch64_const_vec_all_in_range_p): New function.
13866 (aarch64_print_vector_float_operand): Likewise.
13867 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13868 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13869 and the FP immediates 1.0 and 0.5.
13870 (aarch64_print_address_internal): Handle SVE addresses.
13871 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13872 (aarch64_regno_regclass): Handle predicate registers.
13873 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13875 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13876 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13877 (aarch64_convert_sve_vector_bits): New function.
13878 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13879 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13880 rather than an rtx.
13881 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13882 Handle SVE vector and predicate modes. Accept VL-based constants
13883 that need only one temporary register, and VL offsets that require
13884 no temporary registers.
13885 (aarch64_conditional_register_usage): Mark the predicate registers
13886 as fixed if SVE isn't available.
13887 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13888 Return true for SVE vector and predicate modes.
13889 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13890 rather than an unsigned int. Handle SVE modes.
13891 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13893 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13895 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13896 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13897 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13898 (aarch64_sve_float_mul_immediate_p): New functions.
13899 (aarch64_sve_valid_immediate): New function.
13900 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13901 Explicitly reject structure modes. Check for INDEX constants.
13902 Handle PTRUE and PFALSE constants.
13903 (aarch64_check_zero_based_sve_index_immediate): New function.
13904 (aarch64_simd_imm_zero_p): Delete.
13905 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13906 vector modes. Accept constants in the range of CNT[BHWD].
13907 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13908 ask for an Advanced SIMD mode.
13909 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13910 (aarch64_simd_vector_alignment): Handle SVE predicates.
13911 (aarch64_vectorize_preferred_vector_alignment): New function.
13912 (aarch64_simd_vector_alignment_reachable): Use it instead of
13914 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13915 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13917 (MAX_VECT_LEN): Delete.
13918 (expand_vec_perm_d): Add a vec_flags field.
13919 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13920 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13921 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13923 (aarch64_evpc_rev): Rename to...
13924 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13925 (aarch64_evpc_rev_global): New function.
13926 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13927 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13929 (aarch64_evpc_sve_tbl): New function.
13930 (aarch64_expand_vec_perm_const_1): Update after rename of
13931 aarch64_evpc_rev. Handle SVE permutes too, trying
13932 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13933 than aarch64_evpc_tbl.
13934 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13935 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13936 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13937 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13938 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13939 (aarch64_expand_sve_vcond): New functions.
13940 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13941 of aarch64_vector_mode_p.
13942 (aarch64_dwarf_poly_indeterminate_value): New function.
13943 (aarch64_compute_pressure_classes): Likewise.
13944 (aarch64_can_change_mode_class): Likewise.
13945 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13946 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13947 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13948 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13949 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13950 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13951 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13952 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13954 (Dn, Dl, Dr): Accept const as well as const_vector.
13955 (Dz): Likewise. Compare against CONST0_RTX.
13956 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13957 of "vector" where appropriate.
13958 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13959 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13960 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13961 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13962 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13963 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13964 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13965 (v_int_equiv): Extend to SVE modes.
13966 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13968 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13969 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13970 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13971 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13972 (SVE_COND_FP_CMP): New int iterators.
13973 (perm_hilo): Handle the new unpack unspecs.
13974 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13976 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13977 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13978 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13979 (aarch64_equality_operator, aarch64_constant_vector_operand)
13980 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13981 (aarch64_sve_nonimmediate_operand): Likewise.
13982 (aarch64_sve_general_operand): Likewise.
13983 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13984 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13985 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13986 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13987 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13988 (aarch64_sve_float_arith_immediate): Likewise.
13989 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13990 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13991 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13992 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13993 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13994 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13995 (aarch64_sve_float_arith_operand): Likewise.
13996 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13997 (aarch64_sve_float_mul_operand): Likewise.
13998 (aarch64_sve_vec_perm_operand): Likewise.
13999 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14000 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14001 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14002 as well as const_vector.
14003 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14004 in file. Use CONST0_RTX and CONSTM1_RTX.
14005 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14006 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14007 Use aarch64_simd_imm_zero.
14008 * config/aarch64/aarch64-sve.md: New file.
14009 * config/aarch64/aarch64.md: Include it.
14010 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14011 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14012 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14013 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14014 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14015 (sve): New attribute.
14016 (enabled): Disable instructions with the sve attribute unless
14018 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14019 aarch64_expand_mov_immediate.
14020 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14021 CNT[BHSD] immediates.
14022 (movti): Split CONST_POLY_INT moves into two halves.
14023 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14024 Split additions that need a temporary here if the destination
14025 is the stack pointer.
14026 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14027 (*add<mode>3_poly_1): New instruction.
14028 (set_clobber_cc): New expander.
14030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14032 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14033 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14034 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14035 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14036 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14037 Change innermode from fixed_mode_size to machine_mode.
14038 (simplify_subreg): Update call accordingly. Handle a constant-sized
14039 subreg of a variable-length CONST_VECTOR.
14041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14042 Alan Hayward <alan.hayward@arm.com>
14043 David Sherwood <david.sherwood@arm.com>
14045 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14046 (add_offset_to_base): New function, split out from...
14047 (create_mem_ref): ...here. When handling a scale other than 1,
14048 check first whether the address is valid without the offset.
14049 Add it into the base if so, leaving the index and scale as-is.
14051 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14054 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14055 fold_for_warn before checking if arg2 is INTEGER_CST.
14057 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14059 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14060 (store_multiple_operation): Delete.
14061 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14062 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14063 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14064 guarded by TARGET_STRING.
14065 (rs6000_output_load_multiple): Delete.
14066 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14067 OPTION_MASK_STRING / TARGET_STRING handling.
14068 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14069 (const rs6000_opt_masks) <"string">: Change mask to 0.
14070 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14071 (MASK_STRING): Delete.
14072 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14074 (load_multiple): Delete.
14081 (store_multiple): Delete.
14088 (movmemsi_8reg): Delete.
14089 (corresponding unnamed define_insn): Delete.
14090 (movmemsi_6reg): Delete.
14091 (corresponding unnamed define_insn): Delete.
14092 (movmemsi_4reg): Delete.
14093 (corresponding unnamed define_insn): Delete.
14094 (movmemsi_2reg): Delete.
14095 (corresponding unnamed define_insn): Delete.
14096 (movmemsi_1reg): Delete.
14097 (corresponding unnamed define_insn): Delete.
14098 * config/rs6000/rs6000.opt (mno-string): New.
14099 (mstring): Replace by deprecation warning stub.
14100 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14102 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14104 * regrename.c (regrename_do_replace): If replacing the same
14105 reg multiple times, try to reuse last created gen_raw_REG.
14108 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14109 main to workaround a bug in GDB.
14111 2018-01-12 Tom de Vries <tom@codesourcery.com>
14114 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14116 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14118 PR rtl-optimization/80481
14119 * ira-color.c (get_cap_member): New function.
14120 (allocnos_conflict_by_live_ranges_p): Use it.
14121 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14122 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14124 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14127 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14128 (*saddl_se_1): Ditto.
14129 (*ssubsi_1): Ditto.
14130 (*ssubl_se_1): Ditto.
14132 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14134 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14135 rather than wi::to_widest for DR_INITs.
14136 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14137 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14138 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14140 (vect_analyze_group_access_1): Note that here.
14142 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14144 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14145 polynomial type sizes.
14147 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14149 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14150 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14151 (gimple_add_tmp_var): Likewise.
14153 2018-01-12 Martin Liska <mliska@suse.cz>
14155 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14156 (gimple_alloc_sizes): Likewise.
14157 (dump_gimple_statistics): Use PRIu64 in printf format.
14158 * gimple.h: Change uint64_t to int.
14160 2018-01-12 Martin Liska <mliska@suse.cz>
14162 * tree-core.h: Use uint64_t instead of int.
14163 * tree.c (tree_node_counts): Likewise.
14164 (tree_node_sizes): Likewise.
14165 (dump_tree_statistics): Use PRIu64 in printf format.
14167 2018-01-12 Martin Liska <mliska@suse.cz>
14169 * Makefile.in: As qsort_chk is implemented in vec.c, add
14170 vec.o to linkage of gencfn-macros.
14171 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14172 passing the info to record_node_allocation_statistics.
14173 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14175 * ggc-common.c (struct ggc_usage): Add operator== and use
14176 it in operator< and compare function.
14177 * mem-stats.h (struct mem_usage): Likewise.
14178 * vec.c (struct vec_usage): Remove operator< and compare
14179 function. Can be simply inherited.
14181 2018-01-12 Martin Jambor <mjambor@suse.cz>
14184 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14185 * tree-ssa-math-opts.c: Include domwalk.h.
14186 (convert_mult_to_fma_1): New function.
14187 (fma_transformation_info): New type.
14188 (fma_deferring_state): Likewise.
14189 (cancel_fma_deferring): New function.
14190 (result_of_phi): Likewise.
14191 (last_fma_candidate_feeds_initial_phi): Likewise.
14192 (convert_mult_to_fma): Added deferring logic, split actual
14193 transformation to convert_mult_to_fma_1.
14194 (math_opts_dom_walker): New type.
14195 (math_opts_dom_walker::after_dom_children): New method, body moved
14196 here from pass_optimize_widening_mul::execute, added deferring logic
14198 (pass_optimize_widening_mul::execute): Moved most of code to
14199 math_opts_dom_walker::after_dom_children.
14200 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14201 * config/i386/i386.c (ix86_option_override_internal): Added
14202 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14204 2018-01-12 Richard Biener <rguenther@suse.de>
14207 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14208 inline instance vars.
14210 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14213 * config/rx/rx.c (rx_is_restricted_memory_address):
14214 Handle SUBREG case.
14216 2018-01-12 Richard Biener <rguenther@suse.de>
14218 PR tree-optimization/80846
14219 * target.def (split_reduction): New target hook.
14220 * targhooks.c (default_split_reduction): New function.
14221 * targhooks.h (default_split_reduction): Declare.
14222 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14223 target requests first reduce vectors by combining low and high
14225 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14226 (get_vectype_for_scalar_type_and_size): Export.
14227 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14228 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14229 * doc/tm.texi: Regenerate.
14230 * config/i386/i386.c (ix86_split_reduction): Implement
14231 TARGET_VECTORIZE_SPLIT_REDUCTION.
14233 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14236 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14237 in PIC mode except for TARGET_VXWORKS_RTP.
14238 * config/sparc/sparc.c: Include cfgrtl.h.
14239 (TARGET_INIT_PIC_REG): Define.
14240 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14241 (sparc_pic_register_p): New predicate.
14242 (sparc_legitimate_address_p): Use it.
14243 (sparc_legitimize_pic_address): Likewise.
14244 (sparc_delegitimize_address): Likewise.
14245 (sparc_mode_dependent_address_p): Likewise.
14246 (gen_load_pcrel_sym): Remove 4th parameter.
14247 (load_got_register): Adjust call to above. Remove obsolete stuff.
14248 (sparc_expand_prologue): Do not call load_got_register here.
14249 (sparc_flat_expand_prologue): Likewise.
14250 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14251 (sparc_use_pseudo_pic_reg): New function.
14252 (sparc_init_pic_reg): Likewise.
14253 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14254 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14256 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14258 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14259 Add item for branch_cost.
14261 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14263 PR rtl-optimization/83565
14264 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14265 not extend the result to a larger mode for rotate operations.
14266 (num_sign_bit_copies1): Likewise.
14268 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14271 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14273 Use values-Xc.o for -pedantic.
14274 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14276 2018-01-12 Martin Liska <mliska@suse.cz>
14279 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14281 (possible_polymorphic_call_targets): Use it.
14282 (ipa_devirt): Likewise.
14284 2018-01-12 Martin Liska <mliska@suse.cz>
14286 * profile-count.h (enum profile_quality): Use 0 as invalid
14287 enum value of profile_quality.
14289 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14291 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14292 -mext-string options.
14294 2018-01-12 Richard Biener <rguenther@suse.de>
14296 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14297 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14298 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14300 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14302 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14304 * configure.ac (--with-long-double-format): Add support for the
14305 configuration option to change the default long double format on
14307 * config.gcc (powerpc*-linux*-*): Likewise.
14308 * configure: Regenerate.
14309 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14310 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14311 used without modification.
14313 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14315 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14316 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14317 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14318 MISC_BUILTIN_SPEC_BARRIER.
14319 (rs6000_init_builtins): Likewise.
14320 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14322 (speculation_barrier): New define_insn.
14323 * doc/extend.texi: Document __builtin_speculation_barrier.
14325 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14328 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14329 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14330 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14332 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14333 integral modes instead of "ss" and "sd".
14334 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14335 vectors with 32-bit and 64-bit elements.
14336 (vecdupssescalarmodesuffix): New mode attribute.
14337 (vec_dup<mode>): Use it.
14339 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14342 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14343 frame if argument is passed on stack.
14345 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14348 * ree.c (combine_reaching_defs): Optimize also
14349 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14350 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14352 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14354 PR middle-end/83189
14355 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14357 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14359 PR middle-end/83718
14360 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14361 after they are computed.
14363 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14365 PR tree-optimization/83695
14366 * gimple-loop-linterchange.cc
14367 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14368 reset cached scev information after interchange.
14369 (pass_linterchange::execute): Remove call to scev_reset_htab.
14371 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14373 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14374 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14375 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14376 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14377 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14378 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14379 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14380 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14381 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14382 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14383 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14384 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14385 (V_lane_reg): Likewise.
14386 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14388 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14389 (vfmal_lane_low<mode>_intrinsic,
14390 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14391 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14392 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14393 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14394 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14395 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14397 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14399 * config/arm/arm-cpus.in (fp16fml): New feature.
14400 (ALL_SIMD): Add fp16fml.
14401 (armv8.2-a): Add fp16fml as an option.
14402 (armv8.3-a): Likewise.
14403 (armv8.4-a): Add fp16fml as part of fp16.
14404 * config/arm/arm.h (TARGET_FP16FML): Define.
14405 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14407 * config/arm/arm-modes.def (V2HF): Define.
14408 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14409 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14410 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14411 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14412 vfmsl_low, vfmsl_high): New set of builtins.
14413 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14414 (vfml_op): New code attribute.
14415 (VFMLHALVES): New int iterator.
14416 (VFML, VFMLSEL): New mode attributes.
14417 (V_reg): Define mapping for V2HF.
14418 (V_hi, V_lo): New mode attributes.
14419 (VF_constraint): Likewise.
14420 (vfml_half, vfml_half_selector): New int attributes.
14421 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14423 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14424 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14426 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14427 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14428 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14429 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14431 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14432 Document new effective target and option set.
14434 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14436 * config/arm/arm-cpus.in (armv8_4): New feature.
14437 (ARMv8_4a): New fgroup.
14438 (armv8.4-a): New arch.
14439 * config/arm/arm-tables.opt: Regenerate.
14440 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14441 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14442 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14443 Add matching rules for -march=armv8.4-a and extensions.
14444 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14446 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14449 * config/rx/rx.md (BW): New mode attribute.
14450 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14452 2018-01-11 Richard Biener <rguenther@suse.de>
14454 PR tree-optimization/83435
14455 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14456 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14457 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14459 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14460 Alan Hayward <alan.hayward@arm.com>
14461 David Sherwood <david.sherwood@arm.com>
14463 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14465 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14466 (aarch64_print_address_internal): Use it to check for a zero offset.
14468 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14469 Alan Hayward <alan.hayward@arm.com>
14470 David Sherwood <david.sherwood@arm.com>
14472 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14473 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14474 Return a poly_int64 rather than a HOST_WIDE_INT.
14475 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14476 rather than a HOST_WIDE_INT.
14477 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14478 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14479 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14480 final_offset from HOST_WIDE_INT to poly_int64.
14481 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14482 to_constant when getting the number of units in an Advanced SIMD
14484 (aarch64_builtin_vectorized_function): Check for a constant number
14486 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14488 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14489 attribute instead of GET_MODE_NUNITS.
14490 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14491 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14492 GET_MODE_SIZE for fixed-size registers.
14493 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14494 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14495 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14496 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14497 (aarch64_print_operand, aarch64_print_address_internal)
14498 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14499 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14500 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14501 Handle polynomial GET_MODE_SIZE.
14502 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14503 wider than SImode without modification.
14504 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14505 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14506 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14507 passing and returning SVE modes.
14508 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14509 rather than GEN_INT.
14510 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14511 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14512 (aarch64_allocate_and_probe_stack_space): Likewise.
14513 (aarch64_layout_frame): Cope with polynomial offsets.
14514 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14515 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14516 polynomial offsets.
14517 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14518 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14519 poly_int64 rather than a HOST_WIDE_INT.
14520 (aarch64_get_separate_components, aarch64_process_components)
14521 (aarch64_expand_prologue, aarch64_expand_epilogue)
14522 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14523 (aarch64_anchor_offset): New function, split out from...
14524 (aarch64_legitimize_address): ...here.
14525 (aarch64_builtin_vectorization_cost): Handle polynomial
14526 TYPE_VECTOR_SUBPARTS.
14527 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14529 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14530 number of elements from the PARALLEL rather than the mode.
14531 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14532 rather than GET_MODE_BITSIZE.
14533 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14534 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14535 (aarch64_expand_vec_perm_const_1): Handle polynomial
14536 d->perm.length () and d->perm elements.
14537 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14538 Apply to_constant to d->perm elements.
14539 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14540 polynomial CONST_VECTOR_NUNITS.
14541 (aarch64_move_pointer): Take amount as a poly_int64 rather
14543 (aarch64_progress_pointer): Avoid temporary variable.
14544 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14545 the mode attribute instead of GET_MODE.
14547 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14548 Alan Hayward <alan.hayward@arm.com>
14549 David Sherwood <david.sherwood@arm.com>
14551 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14552 x exists before using it.
14553 (aarch64_add_constant_internal): Rename to...
14554 (aarch64_add_offset_1): ...this. Replace regnum with separate
14555 src and dest rtxes. Handle the case in which they're different,
14556 including when the offset is zero. Replace scratchreg with an rtx.
14557 Use 2 additions if there is no spare register into which we can
14558 move a 16-bit constant.
14559 (aarch64_add_constant): Delete.
14560 (aarch64_add_offset): Replace reg with separate src and dest
14561 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14562 Use aarch64_add_offset_1.
14563 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14564 an rtx rather than an int. Take the delta as a poly_int64
14565 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14566 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14567 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14568 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14569 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14570 and aarch64_add_sp.
14571 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14572 aarch64_add_constant.
14574 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14576 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14577 Use scalar_float_mode.
14579 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14581 * config/aarch64/aarch64-simd.md
14582 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14583 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14584 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14585 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14586 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14587 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14588 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14589 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14590 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14591 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14593 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14596 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14597 targ_options->x_arm_arch_string is non NULL.
14599 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14601 * config/aarch64/aarch64.h
14602 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14604 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14607 * expmed.c (emit_store_flag_force): Swap if const op0
14608 and change VOIDmode to mode of op0.
14610 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14612 PR rtl-optimization/83761
14613 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14614 than bytes to mode_for_size.
14616 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14618 PR middle-end/83189
14619 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14620 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14623 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14625 PR middle-end/83575
14626 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14627 when in layout mode.
14628 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14629 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14632 2018-01-10 Michael Collison <michael.collison@arm.com>
14634 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14635 * config/aarch64/aarch64-option-extension.def: Add
14636 AARCH64_OPT_EXTENSION of 'fp16fml'.
14637 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14638 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14639 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14640 * config/aarch64/constraints.md (Ui7): New constraint.
14641 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14642 (VFMLA_SEL_W): Ditto.
14645 (VFMLA16_LOW): New int iterator.
14646 (VFMLA16_HIGH): Ditto.
14647 (UNSPEC_FMLAL): New unspec.
14648 (UNSPEC_FMLSL): Ditto.
14649 (UNSPEC_FMLAL2): Ditto.
14650 (UNSPEC_FMLSL2): Ditto.
14651 (f16mac): New code attribute.
14652 * config/aarch64/aarch64-simd-builtins.def
14653 (aarch64_fmlal_lowv2sf): Ditto.
14654 (aarch64_fmlsl_lowv2sf): Ditto.
14655 (aarch64_fmlalq_lowv4sf): Ditto.
14656 (aarch64_fmlslq_lowv4sf): Ditto.
14657 (aarch64_fmlal_highv2sf): Ditto.
14658 (aarch64_fmlsl_highv2sf): Ditto.
14659 (aarch64_fmlalq_highv4sf): Ditto.
14660 (aarch64_fmlslq_highv4sf): Ditto.
14661 (aarch64_fmlal_lane_lowv2sf): Ditto.
14662 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14663 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14664 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14665 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14666 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14667 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14668 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14669 (aarch64_fmlal_lane_highv2sf): Ditto.
14670 (aarch64_fmlsl_lane_highv2sf): Ditto.
14671 (aarch64_fmlal_laneq_highv2sf): Ditto.
14672 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14673 (aarch64_fmlalq_lane_highv4sf): Ditto.
14674 (aarch64_fmlsl_lane_highv4sf): Ditto.
14675 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14676 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14677 * config/aarch64/aarch64-simd.md:
14678 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14679 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14680 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14681 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14682 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14683 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14684 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14685 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14686 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14687 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14688 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14689 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14690 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14691 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14692 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14693 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14694 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14695 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14696 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14697 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14698 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14699 (vfmlsl_low_u32): Ditto.
14700 (vfmlalq_low_u32): Ditto.
14701 (vfmlslq_low_u32): Ditto.
14702 (vfmlal_high_u32): Ditto.
14703 (vfmlsl_high_u32): Ditto.
14704 (vfmlalq_high_u32): Ditto.
14705 (vfmlslq_high_u32): Ditto.
14706 (vfmlal_lane_low_u32): Ditto.
14707 (vfmlsl_lane_low_u32): Ditto.
14708 (vfmlal_laneq_low_u32): Ditto.
14709 (vfmlsl_laneq_low_u32): Ditto.
14710 (vfmlalq_lane_low_u32): Ditto.
14711 (vfmlslq_lane_low_u32): Ditto.
14712 (vfmlalq_laneq_low_u32): Ditto.
14713 (vfmlslq_laneq_low_u32): Ditto.
14714 (vfmlal_lane_high_u32): Ditto.
14715 (vfmlsl_lane_high_u32): Ditto.
14716 (vfmlal_laneq_high_u32): Ditto.
14717 (vfmlsl_laneq_high_u32): Ditto.
14718 (vfmlalq_lane_high_u32): Ditto.
14719 (vfmlslq_lane_high_u32): Ditto.
14720 (vfmlalq_laneq_high_u32): Ditto.
14721 (vfmlslq_laneq_high_u32): Ditto.
14722 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14723 (AARCH64_FL_FOR_ARCH8_4): New.
14724 (AARCH64_ISA_F16FML): New ISA flag.
14725 (TARGET_F16FML): New feature flag for fp16fml.
14726 (doc/invoke.texi): Document new fp16fml option.
14728 2018-01-10 Michael Collison <michael.collison@arm.com>
14730 * config/aarch64/aarch64-builtins.c:
14731 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14732 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14733 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14734 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14735 (AARCH64_ISA_SHA3): New ISA flag.
14736 (TARGET_SHA3): New feature flag for sha3.
14737 * config/aarch64/iterators.md (sha512_op): New int attribute.
14738 (CRYPTO_SHA512): New int iterator.
14739 (UNSPEC_SHA512H): New unspec.
14740 (UNSPEC_SHA512H2): Ditto.
14741 (UNSPEC_SHA512SU0): Ditto.
14742 (UNSPEC_SHA512SU1): Ditto.
14743 * config/aarch64/aarch64-simd-builtins.def
14744 (aarch64_crypto_sha512hqv2di): New builtin.
14745 (aarch64_crypto_sha512h2qv2di): Ditto.
14746 (aarch64_crypto_sha512su0qv2di): Ditto.
14747 (aarch64_crypto_sha512su1qv2di): Ditto.
14748 (aarch64_eor3qv8hi): Ditto.
14749 (aarch64_rax1qv2di): Ditto.
14750 (aarch64_xarqv2di): Ditto.
14751 (aarch64_bcaxqv8hi): Ditto.
14752 * config/aarch64/aarch64-simd.md:
14753 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14754 (aarch64_crypto_sha512su0qv2di): Ditto.
14755 (aarch64_crypto_sha512su1qv2di): Ditto.
14756 (aarch64_eor3qv8hi): Ditto.
14757 (aarch64_rax1qv2di): Ditto.
14758 (aarch64_xarqv2di): Ditto.
14759 (aarch64_bcaxqv8hi): Ditto.
14760 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14761 (vsha512h2q_u64): Ditto.
14762 (vsha512su0q_u64): Ditto.
14763 (vsha512su1q_u64): Ditto.
14764 (veor3q_u16): Ditto.
14765 (vrax1q_u64): Ditto.
14766 (vxarq_u64): Ditto.
14767 (vbcaxq_u16): Ditto.
14768 * config/arm/types.md (crypto_sha512): New type attribute.
14769 (crypto_sha3): Ditto.
14770 (doc/invoke.texi): Document new sha3 option.
14772 2018-01-10 Michael Collison <michael.collison@arm.com>
14774 * config/aarch64/aarch64-builtins.c:
14775 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14776 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14777 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14778 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14779 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14780 (AARCH64_ISA_SM4): New ISA flag.
14781 (TARGET_SM4): New feature flag for sm4.
14782 * config/aarch64/aarch64-simd-builtins.def
14783 (aarch64_sm3ss1qv4si): Ditto.
14784 (aarch64_sm3tt1aq4si): Ditto.
14785 (aarch64_sm3tt1bq4si): Ditto.
14786 (aarch64_sm3tt2aq4si): Ditto.
14787 (aarch64_sm3tt2bq4si): Ditto.
14788 (aarch64_sm3partw1qv4si): Ditto.
14789 (aarch64_sm3partw2qv4si): Ditto.
14790 (aarch64_sm4eqv4si): Ditto.
14791 (aarch64_sm4ekeyqv4si): Ditto.
14792 * config/aarch64/aarch64-simd.md:
14793 (aarch64_sm3ss1qv4si): Ditto.
14794 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14795 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14796 (aarch64_sm4eqv4si): Ditto.
14797 (aarch64_sm4ekeyqv4si): Ditto.
14798 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14799 (sm3part_op): Ditto.
14800 (CRYPTO_SM3TT): Ditto.
14801 (CRYPTO_SM3PART): Ditto.
14802 (UNSPEC_SM3SS1): New unspec.
14803 (UNSPEC_SM3TT1A): Ditto.
14804 (UNSPEC_SM3TT1B): Ditto.
14805 (UNSPEC_SM3TT2A): Ditto.
14806 (UNSPEC_SM3TT2B): Ditto.
14807 (UNSPEC_SM3PARTW1): Ditto.
14808 (UNSPEC_SM3PARTW2): Ditto.
14809 (UNSPEC_SM4E): Ditto.
14810 (UNSPEC_SM4EKEY): Ditto.
14811 * config/aarch64/constraints.md (Ui2): New constraint.
14812 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14813 * config/arm/types.md (crypto_sm3): New type attribute.
14814 (crypto_sm4): Ditto.
14815 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14816 (vsm3tt1aq_u32): Ditto.
14817 (vsm3tt1bq_u32): Ditto.
14818 (vsm3tt2aq_u32): Ditto.
14819 (vsm3tt2bq_u32): Ditto.
14820 (vsm3partw1q_u32): Ditto.
14821 (vsm3partw2q_u32): Ditto.
14822 (vsm4eq_u32): Ditto.
14823 (vsm4ekeyq_u32): Ditto.
14824 (doc/invoke.texi): Document new sm4 option.
14826 2018-01-10 Michael Collison <michael.collison@arm.com>
14828 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14829 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14830 (AARCH64_FL_FOR_ARCH8_4): New.
14831 (AARCH64_FL_V8_4): New flag.
14832 (doc/invoke.texi): Document new armv8.4-a option.
14834 2018-01-10 Michael Collison <michael.collison@arm.com>
14836 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14837 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14838 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14839 * config/aarch64/aarch64-option-extension.def: Add
14840 AARCH64_OPT_EXTENSION of 'sha2'.
14841 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14842 (crypto): Disable sha2 and aes if crypto disabled.
14843 (crypto): Enable aes and sha2 if enabled.
14844 (simd): Disable sha2 and aes if simd disabled.
14845 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14847 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14848 (TARGET_SHA2): New feature flag for sha2.
14849 (TARGET_AES): New feature flag for aes.
14850 * config/aarch64/aarch64-simd.md:
14851 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14852 conditional on TARGET_AES.
14853 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14854 (aarch64_crypto_sha1hsi): Make pattern conditional
14856 (aarch64_crypto_sha1hv4si): Ditto.
14857 (aarch64_be_crypto_sha1hv4si): Ditto.
14858 (aarch64_crypto_sha1su1v4si): Ditto.
14859 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14860 (aarch64_crypto_sha1su0v4si): Ditto.
14861 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14862 (aarch64_crypto_sha256su0v4si): Ditto.
14863 (aarch64_crypto_sha256su1v4si): Ditto.
14864 (doc/invoke.texi): Document new aes and sha2 options.
14866 2018-01-10 Martin Sebor <msebor@redhat.com>
14868 PR tree-optimization/83781
14869 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14872 2018-01-11 Martin Sebor <msebor@gmail.com>
14873 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14875 PR tree-optimization/83501
14876 PR tree-optimization/81703
14878 * tree-ssa-strlen.c (get_string_cst): Rename...
14879 (get_string_len): ...to this. Handle global constants.
14880 (handle_char_store): Adjust.
14882 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14883 Jim Wilson <jimw@sifive.com>
14885 * config/riscv/riscv-protos.h (riscv_output_return): New.
14886 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14887 (riscv_attribute_table, riscv_output_return),
14888 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14889 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14890 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14892 (riscv_expand_prologue): Add early return for naked function.
14893 (riscv_expand_epilogue): Likewise.
14894 (riscv_function_ok_for_sibcall): Return false for naked function.
14895 (riscv_set_current_function): New.
14896 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14897 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14898 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14899 * doc/extend.texi (RISC-V Function Attributes): New.
14901 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14903 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14904 check for 128-bit long double before checking TCmode.
14905 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14906 128-bit long doubles before checking TFmode or TCmode.
14907 (FLOAT128_IBM_P): Likewise.
14909 2018-01-10 Martin Sebor <msebor@redhat.com>
14911 PR tree-optimization/83671
14912 * builtins.c (c_strlen): Unconditionally return zero for the empty
14914 Use -Warray-bounds for warnings.
14915 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14916 for non-constant array indices with COMPONENT_REF, arrays of
14917 arrays, and pointers to arrays.
14918 (gimple_fold_builtin_strlen): Determine and set length range for
14919 non-constant character arrays.
14921 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14923 PR middle-end/81897
14924 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14927 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14929 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14931 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14934 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14935 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14936 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14937 indexed_or_indirect_operand predicate.
14938 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14939 (*vsx_le_perm_load_v8hi): Likewise.
14940 (*vsx_le_perm_load_v16qi): Likewise.
14941 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14942 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14943 (*vsx_le_perm_store_v8hi): Likewise.
14944 (*vsx_le_perm_store_v16qi): Likewise.
14945 (eight unnamed splitters): Likewise.
14947 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14949 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14950 * config/rs6000/emmintrin.h: Likewise.
14951 * config/rs6000/mmintrin.h: Likewise.
14952 * config/rs6000/xmmintrin.h: Likewise.
14954 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14957 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14959 * tree.c (tree_nop_conversion): Return true for location wrapper
14961 (maybe_wrap_with_location): New function.
14962 (selftest::check_strip_nops): New function.
14963 (selftest::test_location_wrappers): New function.
14964 (selftest::tree_c_tests): Call it.
14965 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14966 (maybe_wrap_with_location): New decl.
14967 (EXPR_LOCATION_WRAPPER_P): New macro.
14968 (location_wrapper_p): New inline function.
14969 (tree_strip_any_location_wrapper): New inline function.
14971 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14974 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14975 stack_realign_offset for the largest alignment of stack slot
14977 (ix86_find_max_used_stack_alignment): New function.
14978 (ix86_finalize_stack_frame_flags): Use it. Set
14979 max_used_stack_alignment if we don't realign stack.
14980 * config/i386/i386.h (machine_function): Add
14981 max_used_stack_alignment.
14983 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14985 * config/arm/arm.opt (-mbranch-cost): New option.
14986 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14989 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14992 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14993 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14995 2018-01-10 Richard Biener <rguenther@suse.de>
14998 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14999 early out so it also covers the case where we have a non-NULL
15002 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15004 PR tree-optimization/83753
15005 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15006 for non-strided grouped accesses if the number of elements is 1.
15008 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15011 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15012 * i386.h (TARGET_USE_GATHER): Define.
15013 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15015 2018-01-10 Martin Liska <mliska@suse.cz>
15018 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15019 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15021 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15022 CLEANUP_NO_PARTITIONING is not set.
15024 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15026 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15027 for vectors, as a partial revert of r254296.
15028 * rtl.h (const_vec_p): Delete.
15029 (const_vec_duplicate_p): Don't test for vector CONSTs.
15030 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15031 * expmed.c (make_tree): Likewise.
15034 * common.md (E, F): Use CONSTANT_P instead of checking for
15036 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15037 checking for CONST_VECTOR.
15039 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15041 PR middle-end/83575
15042 * predict.c (force_edge_cold): Handle in more sane way edges
15043 with no prediction.
15045 2018-01-09 Carl Love <cel@us.ibm.com>
15047 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15049 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15050 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15051 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15052 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15053 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15054 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15055 * config/rs6000/rs6000-protos.h: Add extern defition for
15056 rs6000_generate_float2_double_code.
15057 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15059 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15060 (float2_v2df): Add define_expand.
15062 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15065 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15066 op_mode in the force_to_mode call.
15068 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15070 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15071 instead of checking each element individually.
15072 (aarch64_evpc_uzp): Likewise.
15073 (aarch64_evpc_zip): Likewise.
15074 (aarch64_evpc_ext): Likewise.
15075 (aarch64_evpc_rev): Likewise.
15076 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15077 instead of checking each element individually. Return true without
15079 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15080 whether all selected elements come from the same input, instead of
15081 checking each element individually. Remove calls to gen_rtx_REG,
15082 start_sequence and end_sequence and instead assert that no rtl is
15085 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15087 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15088 order of HIGH and CONST checks.
15090 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15092 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15093 if the destination isn't an SSA_NAME.
15095 2018-01-09 Richard Biener <rguenther@suse.de>
15097 PR tree-optimization/83668
15098 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15100 (canonicalize_loop_form): ... here, renamed from ...
15101 (canonicalize_loop_closed_ssa_form): ... this and amended to
15102 swap successor edges for loop exit blocks to make us use
15103 the RPO order we need for initial schedule generation.
15105 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15107 PR tree-optimization/64811
15108 * match.pd: When optimizing comparisons with Inf, avoid
15109 introducing or losing exceptions from comparisons with NaN.
15111 2018-01-09 Martin Liska <mliska@suse.cz>
15114 * asan.c (shadow_mem_size): Add gcc_assert.
15116 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15118 Don't save registers in main().
15121 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15122 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15123 * config/avr/avr.c (avr_set_current_function): Don't error if
15124 naked, OS_task or OS_main are specified at the same time.
15125 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15127 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15129 * common/config/avr/avr-common.c (avr_option_optimization_table):
15130 Switch on -mmain-is-OS_task for optimizing compilations.
15132 2018-01-09 Richard Biener <rguenther@suse.de>
15134 PR tree-optimization/83572
15135 * graphite.c: Include cfganal.h.
15136 (graphite_transform_loops): Connect infinite loops to exit
15137 and remove fake edges at the end.
15139 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15141 * ipa-inline.c (edge_badness): Revert accidental checkin.
15143 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15146 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15147 symbols; not inline clones.
15149 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15152 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15153 hard registers. Formatting fixes.
15155 PR preprocessor/83722
15156 * gcc.c (try_generate_repro): Pass
15157 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15158 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15161 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15162 Kito Cheng <kito.cheng@gmail.com>
15164 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15165 (riscv_leaf_function_p): Delete.
15166 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15168 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15170 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15172 (do_ifelse): New function.
15173 (do_isel): New function.
15174 (do_sub3): New function.
15175 (do_add3): New function.
15176 (do_load_mask_compare): New function.
15177 (do_overlap_load_compare): New function.
15178 (expand_compare_loop): New function.
15179 (expand_block_compare): Call expand_compare_loop() when appropriate.
15180 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15181 option description.
15182 (-mblock-compare-inline-loop-limit): New option.
15184 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15187 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15188 Reverse order of second and third operands in first alternative.
15189 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15190 of first and second elements in UNSPEC_VPERMR vector.
15191 (altivec_expand_vec_perm_le): Likewise.
15193 2018-01-08 Jeff Law <law@redhat.com>
15195 PR rtl-optimizatin/81308
15196 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15197 (process_switch): If group_case_labels makes a change, then set
15199 (pass_convert_switch::execute): If a switch is converted, then
15200 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15202 PR rtl-optimization/81308
15203 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15206 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15208 PR target/83663 - Revert r255946
15209 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15210 generation for cases where splatting a value is not useful.
15211 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15212 across a vec_duplicate and a paradoxical subreg forming a vector
15213 mode to a vec_concat.
15215 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15217 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15218 -march=armv8.3-a variants.
15219 * config/arm/t-multilib: Likewise.
15220 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15222 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15224 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15226 (cceq_ior_compare_complement): Give it a name so I can use it, and
15227 change boolean_or_operator predicate to boolean_operator so it can
15228 be used to generate a crand.
15229 (eqne): New code iterator.
15230 (bd/bd_neg): New code_attrs.
15231 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15232 a single define_insn.
15233 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15234 decrement (bdnzt/bdnzf/bdzt/bdzf).
15235 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15236 with the new names of the branch decrement patterns, and added the
15237 names of the branch decrement conditional patterns.
15239 2018-01-08 Richard Biener <rguenther@suse.de>
15241 PR tree-optimization/83563
15242 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15245 2018-01-08 Richard Biener <rguenther@suse.de>
15247 PR middle-end/83713
15248 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15250 2018-01-08 Richard Biener <rguenther@suse.de>
15252 PR tree-optimization/83685
15253 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15254 references to abnormals.
15256 2018-01-08 Richard Biener <rguenther@suse.de>
15259 * dwarf2out.c (output_indirect_strings): Handle empty
15260 skeleton_debug_str_hash.
15261 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15263 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15265 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15266 (emit_store_direct): Likewise.
15267 (arc_trampoline_adjust_address): Likewise.
15268 (arc_asm_trampoline_template): New function.
15269 (arc_initialize_trampoline): Use asm_trampoline_template.
15270 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15271 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15272 * config/arc/arc.md (flush_icache): Delete pattern.
15274 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15276 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15277 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15280 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15283 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15284 by not USED_FOR_TARGET.
15285 (make_pass_resolve_sw_modes): Likewise.
15287 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15289 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15292 2018-01-08 Richard Biener <rguenther@suse.de>
15294 PR middle-end/83580
15295 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15297 2018-01-08 Richard Biener <rguenther@suse.de>
15299 PR middle-end/83517
15300 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15302 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15304 PR middle-end/81897
15305 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15306 basic blocks with a small number of successors.
15307 (convert_control_dep_chain_into_preds): Improve handling of
15309 (dump_predicates): Split apart into...
15310 (dump_pred_chain): ...here...
15311 (dump_pred_info): ...and here.
15312 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15313 (can_chain_union_be_invalidated_p): Improve check for invalidation
15315 (uninit_uses_cannot_happen): Avoid unnecessary if
15316 convert_control_dep_chain_into_preds yielded nothing.
15318 2018-01-06 Martin Sebor <msebor@redhat.com>
15320 PR tree-optimization/83640
15321 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15322 subtracting negative offset from size.
15323 (builtin_access::overlap): Adjust offset bounds of the access to fall
15324 within the size of the object if possible.
15326 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15328 PR rtl-optimization/83699
15329 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15330 extract_bit_field_as_subreg to cases in which the extracted
15331 value is also a vector.
15333 * lra-constraints.c (process_alt_operands): Test for the equivalence
15334 substitutions when detecting a possible reload cycle.
15336 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15339 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15340 by default if flag_selective_schedling{,2}. Formatting fixes.
15342 PR rtl-optimization/83682
15343 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15344 if it has non-VECTOR_MODE element mode.
15345 (vec_duplicate_p): Likewise.
15347 PR middle-end/83694
15348 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15349 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15351 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15354 * config/i386/i386-builtin.def
15355 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15356 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15357 Require also OPTION_MASK_ISA_AVX512F in addition to
15358 OPTION_MASK_ISA_GFNI.
15359 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15360 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15361 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15362 to OPTION_MASK_ISA_GFNI.
15363 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15364 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15365 OPTION_MASK_ISA_AVX512BW.
15366 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15367 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15368 addition to OPTION_MASK_ISA_GFNI.
15369 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15370 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15371 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15372 to OPTION_MASK_ISA_GFNI.
15373 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15374 a requirement for all ISAs rather than any of them with a few
15376 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15378 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15379 bitmasks to be enabled with 3 exceptions, instead of requiring any
15380 enabled ISA with lots of exceptions.
15381 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15382 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15383 Change avx512bw in isa attribute to avx512f.
15384 * config/i386/sgxintrin.h: Add license boilerplate.
15385 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15386 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15387 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15388 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15390 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15391 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15392 temporarily sse2 rather than sse if not enabled already.
15395 * config/i386/sse.md (VI248_VLBW): Rename to ...
15396 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15397 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15398 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15399 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15400 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15401 mode iterator instead of VI248_VLBW.
15403 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15405 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15406 (record_modified): Skip clobbers; add debug output.
15407 (param_change_prob): Use sreal frequencies.
15409 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15411 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15412 punt for user-aligned variables.
15414 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15416 * tree-chrec.c (chrec_contains_symbols): Return true for
15419 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15422 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15423 of (x|y) == x for BICS pattern.
15425 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15427 PR tree-optimization/83605
15428 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15429 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15432 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15434 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15435 * config/epiphany/rtems.h: New file.
15437 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15438 Uros Bizjak <ubizjak@gmail.com>
15441 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15442 QIreg_operand instead of register_operand predicate.
15443 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15444 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15445 comments instead of -fmitigate[-_]rop.
15447 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15450 * cgraphunit.c (symbol_table::compile): Switch to text_section
15451 before calling assembly_start debug hook.
15452 * run-rtl-passes.c (run_rtl_passes): Likewise.
15455 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15457 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15458 range_int_cst_p rather than !symbolic_range_p before calling
15459 extract_range_from_multiplicative_op_1.
15461 2018-01-04 Jeff Law <law@redhat.com>
15463 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15464 redundant test in assertion.
15466 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15468 * doc/rtl.texi: Document machine_mode wrapper classes.
15470 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15472 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15473 using tree_to_uhwi.
15475 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15477 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15478 the VEC_PERM_EXPR fold to fail.
15480 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15483 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15484 to switched_sections.
15486 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15489 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15490 test for d.testing.
15492 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15495 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15496 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15498 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15501 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15502 is BLKmode and bitpos not zero or mode change is needed.
15504 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15507 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15510 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15513 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15514 instead of MULT rtx. Update all corresponding splitters.
15515 (*saddl_se): Ditto.
15516 (*ssub<modesuffix>): Ditto.
15517 (*ssubl_se): Ditto.
15518 (*cmp_sadd_di): Update split patterns.
15519 (*cmp_sadd_si): Ditto.
15520 (*cmp_sadd_sidi): Ditto.
15521 (*cmp_ssub_di): Ditto.
15522 (*cmp_ssub_si): Ditto.
15523 (*cmp_ssub_sidi): Ditto.
15524 * config/alpha/predicates.md (const23_operand): New predicate.
15525 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15526 Look for ASHIFT, not MULT inner operand.
15527 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15529 2018-01-04 Martin Liska <mliska@suse.cz>
15531 PR gcov-profile/83669
15532 * gcov.c (output_intermediate_file): Add version to intermediate
15534 * doc/gcov.texi: Document new field 'version' in intermediate
15535 file format. Fix location of '-k' option of gcov command.
15537 2018-01-04 Martin Liska <mliska@suse.cz>
15540 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15542 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15544 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15546 2018-01-03 Martin Sebor <msebor@redhat.com>
15548 PR tree-optimization/83655
15549 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15550 checking calls with invalid arguments.
15552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15554 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15555 (vectorizable_mask_load_store): Delete.
15556 (vectorizable_call): Return false for masked loads and stores.
15557 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15558 instead of gimple_assign_rhs1.
15559 (vectorizable_load): Handle IFN_MASK_LOAD.
15560 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15564 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15566 (vectorizable_mask_load_store): ...here.
15567 (vectorizable_load): ...and here.
15569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15571 * tree-vect-stmts.c (vect_build_all_ones_mask)
15572 (vect_build_zero_merge_argument): New functions, split out from...
15573 (vectorizable_load): ...here.
15575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15577 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15579 (vectorizable_mask_load_store): ...here.
15580 (vectorizable_store): ...and here.
15582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15584 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15586 (vectorizable_mask_load_store): ...here.
15588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15590 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15591 (vect_model_store_cost): Take a vec_load_store_type instead of a
15593 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15594 (vect_model_store_cost): Take a vec_load_store_type instead of a
15596 (vectorizable_mask_load_store): Update accordingly.
15597 (vectorizable_store): Likewise.
15598 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15602 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15603 IFN_MASK_LOAD calls here rather than...
15604 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15607 Alan Hayward <alan.hayward@arm.com>
15608 David Sherwood <david.sherwood@arm.com>
15610 * expmed.c (extract_bit_field_1): For vector extracts,
15611 fall back to extract_bit_field_as_subreg if vec_extract
15614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15615 Alan Hayward <alan.hayward@arm.com>
15616 David Sherwood <david.sherwood@arm.com>
15618 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15619 they are variable or constant sized.
15620 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15621 slots for constant-sized data.
15623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15624 Alan Hayward <alan.hayward@arm.com>
15625 David Sherwood <david.sherwood@arm.com>
15627 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15628 handling COND_EXPRs with boolean comparisons, try to find a better
15629 basis for the mask type than the boolean itself.
15631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15633 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15634 is calculated and how it can be overridden.
15635 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15636 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15638 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15642 Alan Hayward <alan.hayward@arm.com>
15643 David Sherwood <david.sherwood@arm.com>
15645 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15646 Remove the mode argument.
15647 (aarch64_simd_valid_immediate): Remove the mode and inverse
15649 * config/aarch64/iterators.md (bitsize): New iterator.
15650 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15651 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15652 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15653 aarch64_simd_valid_immediate.
15654 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15655 (aarch64_reg_or_bic_imm): Likewise.
15656 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15657 with an insn_type enum and msl with a modifier_type enum.
15658 Replace element_width with a scalar_mode. Change the shift
15659 to unsigned int. Add constructors for scalar_float_mode and
15660 scalar_int_mode elements.
15661 (aarch64_vect_float_const_representable_p): Delete.
15662 (aarch64_can_const_movi_rtx_p)
15663 (aarch64_simd_scalar_immediate_valid_for_move)
15664 (aarch64_simd_make_constant): Update call to
15665 aarch64_simd_valid_immediate.
15666 (aarch64_advsimd_valid_immediate_hs): New function.
15667 (aarch64_advsimd_valid_immediate): Likewise.
15668 (aarch64_simd_valid_immediate): Remove mode and inverse
15669 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15670 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15671 and aarch64_float_const_representable_p on the result.
15672 (aarch64_output_simd_mov_immediate): Remove mode argument.
15673 Update call to aarch64_simd_valid_immediate and use of
15674 simd_immediate_info.
15675 (aarch64_output_scalar_simd_mov_immediate): Update call
15678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15679 Alan Hayward <alan.hayward@arm.com>
15680 David Sherwood <david.sherwood@arm.com>
15682 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15683 (mode_nunits): Likewise CONST_MODE_NUNITS.
15684 * machmode.def (ADJUST_NUNITS): Document.
15685 * genmodes.c (mode_data::need_nunits_adj): New field.
15686 (blank_mode): Update accordingly.
15687 (adj_nunits): New variable.
15688 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15690 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15691 listed in adj_nunits.
15692 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15693 listed in adj_nunits. Don't emit case statements for such modes.
15694 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15695 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15696 nothing if adj_nunits is nonnull.
15697 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15698 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15699 (emit_mode_fbit): Update use of print_maybe_const_decl.
15700 (emit_move_size): Likewise. Treat the array as non-const
15702 (emit_mode_adjustments): Handle adj_nunits.
15704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15706 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15707 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15708 (VECTOR_MODES): Use it.
15709 (make_vector_modes): Take the prefix as an argument.
15711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15712 Alan Hayward <alan.hayward@arm.com>
15713 David Sherwood <david.sherwood@arm.com>
15715 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15716 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15717 for MODE_VECTOR_BOOL.
15718 * machmode.def (VECTOR_BOOL_MODE): Document.
15719 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15720 (make_vector_bool_mode): New function.
15721 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15723 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15724 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15726 * stor-layout.c (int_mode_for_mode): Likewise.
15727 * tree.c (build_vector_type_for_mode): Likewise.
15728 * varasm.c (output_constant_pool_2): Likewise.
15729 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15730 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15731 for MODE_VECTOR_BOOL.
15732 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15733 of mode class checks.
15734 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15735 instead of a list of mode class checks.
15736 (expand_vector_scalar_condition): Likewise.
15737 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15740 Alan Hayward <alan.hayward@arm.com>
15741 David Sherwood <david.sherwood@arm.com>
15743 * machmode.h (mode_size): Change from unsigned short to
15745 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15746 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15747 or if measurement_type is not polynomial.
15748 (fixed_size_mode::includes_p): Check for constant-sized modes.
15749 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15750 return a poly_uint16 rather than an unsigned short.
15751 (emit_mode_size): Change the type of mode_size from unsigned short
15752 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15753 (emit_mode_adjustments): Cope with polynomial vector sizes.
15754 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15756 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15758 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15759 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15760 * caller-save.c (setup_save_areas): Likewise.
15761 (replace_reg_with_saved_mem): Likewise.
15762 * calls.c (emit_library_call_value_1): Likewise.
15763 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15764 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15765 (gen_lowpart_for_combine): Likewise.
15766 * convert.c (convert_to_integer_1): Likewise.
15767 * cse.c (equiv_constant, cse_insn): Likewise.
15768 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15769 (cselib_subst_to_values): Likewise.
15770 * dce.c (word_dce_process_block): Likewise.
15771 * df-problems.c (df_word_lr_mark_ref): Likewise.
15772 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15773 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15774 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15775 (rtl_for_decl_location): Likewise.
15776 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15777 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15778 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15779 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15780 (expand_expr_real_1): Likewise.
15781 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15782 (pad_below): Likewise.
15783 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15784 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15785 * ira.c (get_subreg_tracking_sizes): Likewise.
15786 * ira-build.c (ira_create_allocno_objects): Likewise.
15787 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15788 (ira_sort_regnos_for_alter_reg): Likewise.
15789 * ira-costs.c (record_operand_costs): Likewise.
15790 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15791 (resolve_simple_move): Likewise.
15792 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15793 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15794 (lra_constraints): Likewise.
15795 (CONST_POOL_OK_P): Reject variable-sized modes.
15796 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15797 (add_pseudo_to_slot, lra_spill): Likewise.
15798 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15799 * optabs-query.c (get_best_extraction_insn): Likewise.
15800 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15801 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15802 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15803 * recog.c (offsettable_address_addr_space_p): Likewise.
15804 * regcprop.c (maybe_mode_change): Likewise.
15805 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15806 * regrename.c (build_def_use): Likewise.
15807 * regstat.c (dump_reg_info): Likewise.
15808 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15809 (find_reloads, find_reloads_subreg_address): Likewise.
15810 * reload1.c (eliminate_regs_1): Likewise.
15811 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15812 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15813 (simplify_binary_operation_1, simplify_subreg): Likewise.
15814 * targhooks.c (default_function_arg_padding): Likewise.
15815 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15816 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15817 (verify_gimple_assign_ternary): Likewise.
15818 * tree-inline.c (estimate_move_cost): Likewise.
15819 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15820 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15821 (get_address_cost_ainc): Likewise.
15822 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15823 (vect_supportable_dr_alignment): Likewise.
15824 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15825 (vectorizable_reduction): Likewise.
15826 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15827 (vectorizable_operation, vectorizable_load): Likewise.
15828 * tree.c (build_same_sized_truth_vector_type): Likewise.
15829 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15830 * var-tracking.c (emit_note_insn_var_location): Likewise.
15831 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15832 (ADDR_VEC_ALIGN): Likewise.
15834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15835 Alan Hayward <alan.hayward@arm.com>
15836 David Sherwood <david.sherwood@arm.com>
15838 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15840 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15841 or if measurement_type is polynomial.
15842 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15843 * combine.c (make_extraction): Likewise.
15844 * dse.c (find_shift_sequence): Likewise.
15845 * dwarf2out.c (mem_loc_descriptor): Likewise.
15846 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15847 (extract_bit_field, extract_low_bits): Likewise.
15848 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15849 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15850 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15851 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15852 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15853 * reload.c (find_reloads): Likewise.
15854 * reload1.c (alter_reg): Likewise.
15855 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15856 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15857 * tree-if-conv.c (predicate_mem_writes): Likewise.
15858 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15859 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15860 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15861 * valtrack.c (dead_debug_insert_temp): Likewise.
15862 * varasm.c (mergeable_constant_section): Likewise.
15863 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15866 Alan Hayward <alan.hayward@arm.com>
15867 David Sherwood <david.sherwood@arm.com>
15869 * expr.c (expand_assignment): Cope with polynomial mode sizes
15870 when assigning to a CONCAT.
15872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15873 Alan Hayward <alan.hayward@arm.com>
15874 David Sherwood <david.sherwood@arm.com>
15876 * machmode.h (mode_precision): Change from unsigned short to
15878 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15880 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15881 or if measurement_type is not polynomial.
15882 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15883 in which the mode is already known to be a scalar_int_mode.
15884 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15885 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15887 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15888 for GET_MODE_PRECISION.
15889 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15890 for GET_MODE_PRECISION.
15891 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15893 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15894 (expand_field_assignment, make_extraction): Likewise.
15895 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15896 (get_last_value): Likewise.
15897 * convert.c (convert_to_integer_1): Likewise.
15898 * cse.c (cse_insn): Likewise.
15899 * expr.c (expand_expr_real_1): Likewise.
15900 * lra-constraints.c (simplify_operand_subreg): Likewise.
15901 * optabs-query.c (can_atomic_load_p): Likewise.
15902 * optabs.c (expand_atomic_load): Likewise.
15903 (expand_atomic_store): Likewise.
15904 * ree.c (combine_reaching_defs): Likewise.
15905 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15906 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15907 * tree.h (type_has_mode_precision_p): Likewise.
15908 * ubsan.c (instrument_si_overflow): Likewise.
15910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15911 Alan Hayward <alan.hayward@arm.com>
15912 David Sherwood <david.sherwood@arm.com>
15914 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15915 polynomial numbers of units.
15916 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15917 (valid_vector_subparts_p): New function.
15918 (build_vector_type): Remove temporary shim and take the number
15919 of units as a poly_uint64 rather than an int.
15920 (build_opaque_vector_type): Take the number of units as a
15921 poly_uint64 rather than an int.
15922 * tree.c (build_vector_from_ctor): Handle polynomial
15923 TYPE_VECTOR_SUBPARTS.
15924 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15925 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15926 (build_vector_from_val): If the number of units is variable,
15927 use build_vec_duplicate_cst for constant operands and
15928 VEC_DUPLICATE_EXPR otherwise.
15929 (make_vector_type): Remove temporary is_constant ().
15930 (build_vector_type, build_opaque_vector_type): Take the number of
15931 units as a poly_uint64 rather than an int.
15932 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15934 * cfgexpand.c (expand_debug_expr): Likewise.
15935 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15936 (store_constructor, expand_expr_real_1): Likewise.
15937 (const_scalar_mask_from_tree): Likewise.
15938 * fold-const-call.c (fold_const_reduction): Likewise.
15939 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15940 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15941 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15942 (fold_relational_const): Likewise.
15943 (native_interpret_vector): Likewise. Change the size from an
15944 int to an unsigned int.
15945 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15946 TYPE_VECTOR_SUBPARTS.
15947 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15948 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15949 duplicating a non-constant operand into a variable-length vector.
15950 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15951 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15952 * ipa-icf.c (sem_variable::equals): Likewise.
15953 * match.pd: Likewise.
15954 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15955 * print-tree.c (print_node): Likewise.
15956 * stor-layout.c (layout_type): Likewise.
15957 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15958 * tree-cfg.c (verify_gimple_comparison): Likewise.
15959 (verify_gimple_assign_binary): Likewise.
15960 (verify_gimple_assign_ternary): Likewise.
15961 (verify_gimple_assign_single): Likewise.
15962 * tree-pretty-print.c (dump_generic_node): Likewise.
15963 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15964 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15965 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15966 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15967 (vect_shift_permute_load_chain): Likewise.
15968 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15969 (expand_vector_condition, optimize_vector_constructor): Likewise.
15970 (lower_vec_perm, get_compute_type): Likewise.
15971 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15972 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15973 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15974 (vect_recog_mask_conversion_pattern): Likewise.
15975 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15976 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15977 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15978 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15979 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15980 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15981 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15982 (supportable_widening_operation): Likewise.
15983 (supportable_narrowing_operation): Likewise.
15984 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15986 * varasm.c (output_constant): Likewise.
15988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15989 Alan Hayward <alan.hayward@arm.com>
15990 David Sherwood <david.sherwood@arm.com>
15992 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15993 so that both the length == 3 and length != 3 cases set up their
15994 own permute vectors. Add comments explaining why we know the
15995 number of elements is constant.
15996 (vect_permute_load_chain): Likewise.
15998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15999 Alan Hayward <alan.hayward@arm.com>
16000 David Sherwood <david.sherwood@arm.com>
16002 * machmode.h (mode_nunits): Change from unsigned char to
16004 (ONLY_FIXED_SIZE_MODES): New macro.
16005 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16006 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16007 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16009 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16010 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16011 or if measurement_type is not polynomial.
16012 * genmodes.c (ZERO_COEFFS): New macro.
16013 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16015 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16016 Use ZERO_COEFFS when emitting initializers.
16017 * data-streamer.h (bp_pack_poly_value): New function.
16018 (bp_unpack_poly_value): Likewise.
16019 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16020 for GET_MODE_NUNITS.
16021 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16022 for GET_MODE_NUNITS.
16023 * tree.c (make_vector_type): Remove temporary shim and make
16024 the real function take the number of units as a poly_uint64
16025 rather than an int.
16026 (build_vector_type_for_mode): Handle polynomial nunits.
16027 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16028 * emit-rtl.c (const_vec_series_p_1): Likewise.
16029 (gen_rtx_CONST_VECTOR): Likewise.
16030 * fold-const.c (test_vec_duplicate_folding): Likewise.
16031 * genrecog.c (validate_pattern): Likewise.
16032 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16033 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16034 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16035 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16036 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16037 * rtlanal.c (subreg_get_info): Likewise.
16038 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16039 (vect_grouped_load_supported): Likewise.
16040 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16041 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16042 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16043 (simplify_const_unary_operation, simplify_binary_operation_1)
16044 (simplify_const_binary_operation, simplify_ternary_operation)
16045 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16046 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16047 instead of CONST_VECTOR_NUNITS.
16048 * varasm.c (output_constant_pool_2): Likewise.
16049 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16050 explicit-encoded elements in the XVEC for variable-length vectors.
16052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16054 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16057 Alan Hayward <alan.hayward@arm.com>
16058 David Sherwood <david.sherwood@arm.com>
16060 * coretypes.h (fixed_size_mode): Declare.
16061 (fixed_size_mode_pod): New typedef.
16062 * builtins.h (target_builtins::x_apply_args_mode)
16063 (target_builtins::x_apply_result_mode): Change type to
16064 fixed_size_mode_pod.
16065 * builtins.c (apply_args_size, apply_result_size, result_vector)
16066 (expand_builtin_apply_args_1, expand_builtin_apply)
16067 (expand_builtin_return): Update accordingly.
16069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16071 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16072 * cselib.c (cselib_hash_rtx): Likewise.
16073 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16074 CONST_VECTOR encoding.
16076 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16077 Jeff Law <law@redhat.com>
16080 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16081 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16082 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16083 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16086 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16087 explicitly probe *sp in a noreturn function if there were any callee
16088 register saves or frame pointer is needed.
16090 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16093 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16094 BLKmode for ternary, binary or unary expressions.
16097 * var-tracking.c (delete_vta_debug_insn): New inline function.
16098 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16099 insns from get_insns () to NULL instead of each bb separately.
16100 Use delete_vta_debug_insn. No longer static.
16101 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16102 delete_vta_debug_insns callers.
16103 * rtl.h (delete_vta_debug_insns): Declare.
16104 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16105 instead of variable_tracking_main.
16107 2018-01-03 Martin Sebor <msebor@redhat.com>
16109 PR tree-optimization/83603
16110 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16111 arguments past the endof the argument list in functions declared
16112 without a prototype.
16113 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16114 Avoid checking when arguments are null.
16116 2018-01-03 Martin Sebor <msebor@redhat.com>
16119 * doc/extend.texi (attribute const): Fix a typo.
16120 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16121 issuing -Wsuggest-attribute for void functions.
16123 2018-01-03 Martin Sebor <msebor@redhat.com>
16125 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16126 offset_int::from instead of wide_int::to_shwi.
16127 (maybe_diag_overlap): Remove assertion.
16128 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16129 * gimple-ssa-sprintf.c (format_directive): Same.
16130 (parse_directive): Same.
16131 (sprintf_dom_walker::compute_format_length): Same.
16132 (try_substitute_return_value): Same.
16134 2018-01-03 Jeff Law <law@redhat.com>
16136 PR middle-end/83654
16137 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16138 non-constant residual for zero at runtime and avoid probing in
16139 that case. Reorganize code for trailing problem to mirror handling
16142 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16144 PR tree-optimization/83501
16145 * tree-ssa-strlen.c (get_string_cst): New.
16146 (handle_char_store): Call get_string_cst.
16148 2018-01-03 Martin Liska <mliska@suse.cz>
16150 PR tree-optimization/83593
16151 * tree-ssa-strlen.c: Include tree-cfg.h.
16152 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16153 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16154 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16156 (strlen_dom_walker::before_dom_children): Call
16157 gimple_purge_dead_eh_edges. Dump tranformation with details
16159 (strlen_dom_walker::before_dom_children): Update call by adding
16160 new argument cleanup_eh.
16161 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16163 2018-01-03 Martin Liska <mliska@suse.cz>
16166 * cif-code.def (VARIADIC_THUNK): New enum value.
16167 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16170 2018-01-03 Jan Beulich <jbeulich@suse.com>
16172 * sse.md (mov<mode>_internal): Tighten condition for when to use
16173 vmovdqu<ssescalarsize> for TI and OI modes.
16175 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16177 Update copyright years.
16179 2018-01-03 Martin Liska <mliska@suse.cz>
16182 * ipa-visibility.c (function_and_variable_visibility): Skip
16183 functions with noipa attribure.
16185 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16187 * gcc.c (process_command): Update copyright notice dates.
16188 * gcov-dump.c (print_version): Ditto.
16189 * gcov.c (print_version): Ditto.
16190 * gcov-tool.c (print_version): Ditto.
16191 * gengtype.c (create_file): Ditto.
16192 * doc/cpp.texi: Bump @copying's copyright year.
16193 * doc/cppinternals.texi: Ditto.
16194 * doc/gcc.texi: Ditto.
16195 * doc/gccint.texi: Ditto.
16196 * doc/gcov.texi: Ditto.
16197 * doc/install.texi: Ditto.
16198 * doc/invoke.texi: Ditto.
16200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16202 * vector-builder.h (vector_builder::m_full_nelts): Change from
16203 unsigned int to poly_uint64.
16204 (vector_builder::full_nelts): Update prototype accordingly.
16205 (vector_builder::new_vector): Likewise.
16206 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16207 (vector_builder::operator ==): Likewise.
16208 (vector_builder::finalize): Likewise.
16209 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16210 Take the number of elements as a poly_uint64 rather than an
16212 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16213 from unsigned int to poly_uint64.
16214 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16215 (vec_perm_indices::new_vector): Likewise.
16216 (vec_perm_indices::length): Likewise.
16217 (vec_perm_indices::nelts_per_input): Likewise.
16218 (vec_perm_indices::input_nelts): Likewise.
16219 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16220 number of elements per input as a poly_uint64 rather than an
16221 unsigned int. Use the original encoding for variable-length
16222 vectors, rather than clamping each individual element.
16223 For the second and subsequent elements in each pattern,
16224 clamp the step and base before clamping their sum.
16225 (vec_perm_indices::series_p): Handle polynomial element counts.
16226 (vec_perm_indices::all_in_range_p): Likewise.
16227 (vec_perm_indices_to_tree): Likewise.
16228 (vec_perm_indices_to_rtx): Likewise.
16229 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16230 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16231 (tree_vector_builder::new_binary_operation): Handle polynomial
16232 element counts. Return false if we need to know the number
16233 of elements at compile time.
16234 * fold-const.c (fold_vec_perm): Punt if the number of elements
16235 isn't known at compile time.
16237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16239 * vec-perm-indices.h (vec_perm_builder): Change element type
16240 from HOST_WIDE_INT to poly_int64.
16241 (vec_perm_indices::element_type): Update accordingly.
16242 (vec_perm_indices::clamp): Handle polynomial element_types.
16243 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16244 (vec_perm_indices::all_in_range_p): Likewise.
16245 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16247 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16248 polynomial vec_perm_indices element types.
16249 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16250 * fold-const.c (fold_vec_perm): Likewise.
16251 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16252 * tree-vect-generic.c (lower_vec_perm): Likewise.
16253 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16254 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16255 element type to HOST_WIDE_INT.
16257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16258 Alan Hayward <alan.hayward@arm.com>
16259 David Sherwood <david.sherwood@arm.com>
16261 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16262 rather than an int. Use plus_constant.
16263 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16264 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16267 Alan Hayward <alan.hayward@arm.com>
16268 David Sherwood <david.sherwood@arm.com>
16270 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16271 a HOST_WIDE_INT to a poly_int64.
16273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16274 Alan Hayward <alan.hayward@arm.com>
16275 David Sherwood <david.sherwood@arm.com>
16277 * calls.c (load_register_parameters): Cope with polynomial
16278 mode sizes. Require a constant size for BLKmode parameters
16279 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16280 forces a parameter to be padded at the lsb end in order to
16281 fill a complete number of words, require the parameter size
16282 to be ordered wrt UNITS_PER_WORD.
16284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16285 Alan Hayward <alan.hayward@arm.com>
16286 David Sherwood <david.sherwood@arm.com>
16288 * reload1.c (spill_stack_slot_width): Change element type
16289 from unsigned int to poly_uint64_pod.
16290 (alter_reg): Treat mode sizes as polynomial.
16292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16293 Alan Hayward <alan.hayward@arm.com>
16294 David Sherwood <david.sherwood@arm.com>
16296 * reload.c (complex_word_subreg_p): New function.
16297 (reload_inner_reg_of_subreg, push_reload): Use it.
16299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16300 Alan Hayward <alan.hayward@arm.com>
16301 David Sherwood <david.sherwood@arm.com>
16303 * lra-constraints.c (process_alt_operands): Reject matched
16304 operands whose sizes aren't ordered.
16305 (match_reload): Refer to this check here.
16307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16308 Alan Hayward <alan.hayward@arm.com>
16309 David Sherwood <david.sherwood@arm.com>
16311 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16312 that the mode size is in the set {1, 2, 4, 8, 16}.
16314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16315 Alan Hayward <alan.hayward@arm.com>
16316 David Sherwood <david.sherwood@arm.com>
16318 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16319 Use plus_constant instead of gen_rtx_PLUS.
16321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16322 Alan Hayward <alan.hayward@arm.com>
16323 David Sherwood <david.sherwood@arm.com>
16325 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16326 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16327 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16328 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16329 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16330 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16331 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16332 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16333 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16334 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16336 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16337 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16338 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16339 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16340 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16341 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16342 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16343 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16344 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16345 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16347 * expr.c (emit_move_resolve_push): Treat the input and result
16348 of PUSH_ROUNDING as a poly_int64.
16349 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16350 (emit_push_insn): Likewise.
16351 * lra-eliminations.c (mark_not_eliminable): Likewise.
16352 * recog.c (push_operand): Likewise.
16353 * reload1.c (elimination_effects): Likewise.
16354 * rtlanal.c (nonzero_bits1): Likewise.
16355 * calls.c (store_one_arg): Likewise. Require the padding to be
16356 known at compile time.
16358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16359 Alan Hayward <alan.hayward@arm.com>
16360 David Sherwood <david.sherwood@arm.com>
16362 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16363 Use plus_constant instead of gen_rtx_PLUS.
16365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16366 Alan Hayward <alan.hayward@arm.com>
16367 David Sherwood <david.sherwood@arm.com>
16369 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16370 rather than an int.
16372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16373 Alan Hayward <alan.hayward@arm.com>
16374 David Sherwood <david.sherwood@arm.com>
16376 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16377 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16378 via stack temporaries. Treat the mode size as polynomial too.
16380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16381 Alan Hayward <alan.hayward@arm.com>
16382 David Sherwood <david.sherwood@arm.com>
16384 * expr.c (expand_expr_real_2): When handling conversions involving
16385 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16386 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16387 as a poly_uint64 too.
16389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16390 Alan Hayward <alan.hayward@arm.com>
16391 David Sherwood <david.sherwood@arm.com>
16393 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16396 Alan Hayward <alan.hayward@arm.com>
16397 David Sherwood <david.sherwood@arm.com>
16399 * combine.c (can_change_dest_mode): Handle polynomial
16400 REGMODE_NATURAL_SIZE.
16401 * expmed.c (store_bit_field_1): Likewise.
16402 * expr.c (store_constructor): Likewise.
16403 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16404 and polynomial REGMODE_NATURAL_SIZE.
16405 (gen_lowpart_common): Likewise.
16406 * reginfo.c (record_subregs_of_mode): Likewise.
16407 * rtlanal.c (read_modify_subreg_p): Likewise.
16409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16410 Alan Hayward <alan.hayward@arm.com>
16411 David Sherwood <david.sherwood@arm.com>
16413 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16414 numbers of elements.
16416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16417 Alan Hayward <alan.hayward@arm.com>
16418 David Sherwood <david.sherwood@arm.com>
16420 * match.pd: Cope with polynomial numbers of vector elements.
16422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16423 Alan Hayward <alan.hayward@arm.com>
16424 David Sherwood <david.sherwood@arm.com>
16426 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16427 in a POINTER_PLUS_EXPR.
16429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16430 Alan Hayward <alan.hayward@arm.com>
16431 David Sherwood <david.sherwood@arm.com>
16433 * omp-simd-clone.c (simd_clone_subparts): New function.
16434 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16435 (ipa_simd_modify_function_body): Likewise.
16437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16438 Alan Hayward <alan.hayward@arm.com>
16439 David Sherwood <david.sherwood@arm.com>
16441 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16442 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16443 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16444 (expand_vector_condition, vector_element): Likewise.
16445 (subparts_gt): New function.
16446 (get_compute_type): Use subparts_gt.
16447 (count_type_subparts): Delete.
16448 (expand_vector_operations_1): Use subparts_gt instead of
16449 count_type_subparts.
16451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16452 Alan Hayward <alan.hayward@arm.com>
16453 David Sherwood <david.sherwood@arm.com>
16455 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16456 (vect_compile_time_alias): ...this new function. Do the calculation
16457 on poly_ints rather than trees.
16458 (vect_prune_runtime_alias_test_list): Update call accordingly.
16460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16461 Alan Hayward <alan.hayward@arm.com>
16462 David Sherwood <david.sherwood@arm.com>
16464 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16466 (vect_schedule_slp_instance): Likewise.
16468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16469 Alan Hayward <alan.hayward@arm.com>
16470 David Sherwood <david.sherwood@arm.com>
16472 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16473 constant and extern definitions for variable-length vectors.
16474 (vect_get_constant_vectors): Note that the number of units
16475 is known to be constant.
16477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16478 Alan Hayward <alan.hayward@arm.com>
16479 David Sherwood <david.sherwood@arm.com>
16481 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16482 of units as polynomial. Choose between WIDE and NARROW based
16485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16486 Alan Hayward <alan.hayward@arm.com>
16487 David Sherwood <david.sherwood@arm.com>
16489 * tree-vect-stmts.c (simd_clone_subparts): New function.
16490 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16493 Alan Hayward <alan.hayward@arm.com>
16494 David Sherwood <david.sherwood@arm.com>
16496 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16497 vectors as polynomial. Use build_index_vector for
16498 IFN_GOMP_SIMD_LANE.
16500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16501 Alan Hayward <alan.hayward@arm.com>
16502 David Sherwood <david.sherwood@arm.com>
16504 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16505 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16506 for variable-length vectors.
16507 (vectorizable_mask_load_store): Treat the number of units as
16508 polynomial, asserting that it is constant if the condition has
16509 already been enforced.
16510 (vectorizable_store, vectorizable_load): Likewise.
16512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16513 Alan Hayward <alan.hayward@arm.com>
16514 David Sherwood <david.sherwood@arm.com>
16516 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16517 of units as polynomial. Punt if we can't tell at compile time
16518 which vector contains the final result.
16520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16521 Alan Hayward <alan.hayward@arm.com>
16522 David Sherwood <david.sherwood@arm.com>
16524 * tree-vect-loop.c (vectorizable_induction): Treat the number
16525 of units as polynomial. Punt on SLP inductions. Use an integer
16526 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16527 cast of such a series for variable-length floating-point
16530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16531 Alan Hayward <alan.hayward@arm.com>
16532 David Sherwood <david.sherwood@arm.com>
16534 * tree.h (build_index_vector): Declare.
16535 * tree.c (build_index_vector): New function.
16536 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16537 of units as polynomial, forcibly converting it to a constant if
16538 vectorizable_reduction has already enforced the condition.
16539 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16540 to create a {1,2,3,...} vector.
16541 (vectorizable_reduction): Treat the number of units as polynomial.
16542 Choose vectype_in based on the largest scalar element size rather
16543 than the smallest number of units. Enforce the restrictions
16546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16547 Alan Hayward <alan.hayward@arm.com>
16548 David Sherwood <david.sherwood@arm.com>
16550 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16551 number of units as polynomial.
16553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16554 Alan Hayward <alan.hayward@arm.com>
16555 David Sherwood <david.sherwood@arm.com>
16557 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16558 * target.def (autovectorize_vector_sizes): Return the vector sizes
16559 by pointer, using vector_sizes rather than a bitmask.
16560 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16561 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16562 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16564 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16565 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16566 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16567 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16568 * omp-general.c (omp_max_vf): Likewise.
16569 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16570 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16571 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16572 * tree-vect-slp.c (vect_slp_bb): Likewise.
16573 * doc/tm.texi: Regenerate.
16574 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16576 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16577 the vector size as a poly_uint64 rather than an unsigned int.
16578 (current_vector_size): Change from an unsigned int to a poly_uint64.
16579 (get_vectype_for_scalar_type): Update accordingly.
16580 * tree.h (build_truth_vector_type): Take the size and number of
16581 units as a poly_uint64 rather than an unsigned int.
16582 (build_vector_type): Add a temporary overload that takes
16583 the number of units as a poly_uint64 rather than an unsigned int.
16584 * tree.c (make_vector_type): Likewise.
16585 (build_truth_vector_type): Take the number of units as a poly_uint64
16586 rather than an unsigned int.
16588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16589 Alan Hayward <alan.hayward@arm.com>
16590 David Sherwood <david.sherwood@arm.com>
16592 * target.def (get_mask_mode): Take the number of units and length
16593 as poly_uint64s rather than unsigned ints.
16594 * targhooks.h (default_get_mask_mode): Update accordingly.
16595 * targhooks.c (default_get_mask_mode): Likewise.
16596 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16597 * doc/tm.texi: Regenerate.
16599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16600 Alan Hayward <alan.hayward@arm.com>
16601 David Sherwood <david.sherwood@arm.com>
16603 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16604 * omp-general.c (omp_max_vf): Likewise.
16605 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16606 (expand_omp_simd): Handle polynomial safelen.
16607 * omp-low.c (omplow_simd_context): Add a default constructor.
16608 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16609 (lower_rec_simd_input_clauses): Update accordingly.
16610 (lower_rec_input_clauses): Likewise.
16612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16613 Alan Hayward <alan.hayward@arm.com>
16614 David Sherwood <david.sherwood@arm.com>
16616 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16617 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16618 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16619 (vect_analyze_slp_cost): Likewise.
16620 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16621 (vect_model_load_cost): Likewise.
16623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16624 Alan Hayward <alan.hayward@arm.com>
16625 David Sherwood <david.sherwood@arm.com>
16627 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16628 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16629 from an unsigned int * to a poly_uint64_pod *.
16630 (calculate_unrolling_factor): New function.
16631 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16634 Alan Hayward <alan.hayward@arm.com>
16635 David Sherwood <david.sherwood@arm.com>
16637 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16638 from an unsigned int to a poly_uint64.
16639 (_loop_vec_info::slp_unrolling_factor): Likewise.
16640 (_loop_vec_info::vectorization_factor): Change from an int
16642 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16643 (vect_get_num_vectors): New function.
16644 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16645 (vect_get_num_copies): Use vect_get_num_vectors.
16646 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16647 to an unsigned int *.
16648 (vect_analyze_data_refs): Change min_vf from an int * to a
16650 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16651 than an unsigned HOST_WIDE_INT.
16652 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16653 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16654 to an unsigned int *.
16655 (vect_analyze_data_ref_dependences): Likewise.
16656 (vect_compute_data_ref_alignment): Handle polynomial vf.
16657 (vect_enhance_data_refs_alignment): Likewise.
16658 (vect_prune_runtime_alias_test_list): Likewise.
16659 (vect_shift_permute_load_chain): Likewise.
16660 (vect_supportable_dr_alignment): Likewise.
16661 (dependence_distance_ge_vf): Take the vectorization factor as a
16662 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16663 (vect_analyze_data_refs): Change min_vf from an int * to a
16665 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16666 vfm1 as a poly_uint64 rather than an int. Make the same change
16667 for the returned bound_scalar.
16668 (vect_gen_vector_loop_niters): Handle polynomial vf.
16669 (vect_do_peeling): Likewise. Update call to
16670 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16671 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16673 * tree-vect-loop.c (vect_determine_vectorization_factor)
16674 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16675 (vect_get_known_peeling_cost): Likewise.
16676 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16677 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16678 (vect_transform_loop): Likewise. Use the lowest possible VF when
16679 updating the upper bounds of the loop.
16680 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16681 rather than an int.
16682 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16683 polynomial unroll factors.
16684 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16685 (vect_make_slp_decision): Likewise.
16686 (vect_supported_load_permutation_p): Likewise, and polynomial
16688 (vect_analyze_slp_cost): Handle polynomial vf.
16689 (vect_slp_analyze_node_operations): Likewise.
16690 (vect_slp_analyze_bb_1): Likewise.
16691 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16692 than an unsigned HOST_WIDE_INT.
16693 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16694 (vectorizable_load): Handle polynomial vf.
16695 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16697 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16700 Alan Hayward <alan.hayward@arm.com>
16701 David Sherwood <david.sherwood@arm.com>
16703 * match.pd: Handle bit operations involving three constants
16704 and try to fold one pair.
16706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16708 * tree-vect-loop-manip.c: Include gimple-fold.h.
16709 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16710 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16711 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16712 Add a path that uses a step of VF instead of 1, but disable it
16714 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16715 and niters_no_overflow parameters. Update calls to
16716 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16717 Create a new SSA name if the latter choses to use a ste other
16718 than zero, and return it via niters_vector_mult_vf_var.
16719 * tree-vect-loop.c (vect_transform_loop): Update calls to
16720 vect_do_peeling, vect_gen_vector_loop_niters and
16721 slpeel_make_loop_iterate_ntimes.
16722 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16723 (vect_gen_vector_loop_niters): Update declarations after above changes.
16725 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16727 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16728 128-bit round to integer instructions.
16729 (ceil<mode>2): Likewise.
16730 (btrunc<mode>2): Likewise.
16731 (round<mode>2): Likewise.
16733 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16735 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16736 unaligned VSX load/store on P8/P9.
16737 (expand_block_clear): Allow the use of unaligned VSX
16738 load/store on P8/P9.
16740 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16742 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16744 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16745 swap associated with both a load and a store.
16747 2018-01-02 Andrew Waterman <andrew@sifive.com>
16749 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16750 * config/riscv/riscv.md (clear_cache): Use it.
16752 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16754 * web.c: Remove out-of-date comment.
16756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16758 * expr.c (fixup_args_size_notes): Check that any existing
16759 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16760 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16761 (emit_single_push_insn): ...here.
16763 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16765 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16766 (const_vector_encoded_nelts): New function.
16767 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16768 (const_vector_int_elt, const_vector_elt): Declare.
16769 * emit-rtl.c (const_vector_int_elt_1): New function.
16770 (const_vector_elt): Likewise.
16771 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16772 of CONST_VECTOR_ELT.
16774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16776 * expr.c: Include rtx-vector-builder.h.
16777 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16778 directly on the tree encoding.
16779 (const_vector_from_tree): Likewise.
16780 * optabs.c: Include rtx-vector-builder.h.
16781 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16782 sequence of "u" values.
16783 * vec-perm-indices.c: Include rtx-vector-builder.h.
16784 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16785 directly on the vec_perm_indices encoding.
16787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16789 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16790 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16791 * rtx-vector-builder.h: New file.
16792 * rtx-vector-builder.c: Likewise.
16793 * rtl.h (rtx_def::u2): Add a const_vector field.
16794 (CONST_VECTOR_NPATTERNS): New macro.
16795 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16796 (CONST_VECTOR_DUPLICATE_P): Likewise.
16797 (CONST_VECTOR_STEPPED_P): Likewise.
16798 (CONST_VECTOR_ENCODED_ELT): Likewise.
16799 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16800 (unwrap_const_vec_duplicate): Likewise.
16801 (const_vec_series_p): Check for a non-duplicated vector encoding.
16802 Say that the function only returns true for integer vectors.
16803 * emit-rtl.c: Include rtx-vector-builder.h.
16804 (gen_const_vec_duplicate_1): Delete.
16805 (gen_const_vector): Call gen_const_vec_duplicate instead of
16806 gen_const_vec_duplicate_1.
16807 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16808 (gen_const_vec_duplicate): Use rtx_vector_builder.
16809 (gen_const_vec_series): Likewise.
16810 (gen_rtx_CONST_VECTOR): Likewise.
16811 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16812 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16813 Build a new vector rather than modifying a CONST_VECTOR in-place.
16814 (handle_special_swappables): Update call accordingly.
16815 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16816 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16817 Build a new vector rather than modifying a CONST_VECTOR in-place.
16818 (handle_special_swappables): Update call accordingly.
16820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16822 * simplify-rtx.c (simplify_const_binary_operation): Use
16823 CONST_VECTOR_ELT instead of XVECEXP.
16825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16827 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16828 the selector elements to be different from the data elements
16829 if the selector is a VECTOR_CST.
16830 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16831 ssizetype for the selector.
16833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16835 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16836 before testing each element individually.
16837 * tree-vect-generic.c (lower_vec_perm): Likewise.
16839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16841 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16842 * selftest-run-tests.c (selftest::run_tests): Call it.
16843 * vector-builder.h (vector_builder::operator ==): New function.
16844 (vector_builder::operator !=): Likewise.
16845 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16846 (vec_perm_indices::all_from_input_p): New function.
16847 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16848 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16849 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16850 instead of reading the VECTOR_CST directly. Detect whether both
16851 vector inputs are the same before constructing the vec_perm_indices,
16852 and update the number of inputs argument accordingly. Use the
16853 utility functions added above. Only construct sel2 if we need to.
16855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16857 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16858 the broadcast of the low byte.
16859 (expand_mult_highpart): Use an explicit encoding for the permutes.
16860 * optabs-query.c (can_mult_highpart_p): Likewise.
16861 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16862 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16863 (vectorizable_bswap): Likewise.
16864 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16865 explicit encoding for the power-of-2 permutes.
16866 (vect_permute_store_chain): Likewise.
16867 (vect_grouped_load_supported): Likewise.
16868 (vect_permute_load_chain): Likewise.
16870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16872 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16873 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16874 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16875 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16876 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16877 (vect_gen_perm_mask_any): Likewise.
16879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16881 * int-vector-builder.h: New file.
16882 * vec-perm-indices.h: Include int-vector-builder.h.
16883 (vec_perm_indices): Redefine as an int_vector_builder.
16884 (auto_vec_perm_indices): Delete.
16885 (vec_perm_builder): Redefine as a stand-alone class.
16886 (vec_perm_indices::vec_perm_indices): New function.
16887 (vec_perm_indices::clamp): Likewise.
16888 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16889 (vec_perm_indices::new_vector): New function.
16890 (vec_perm_indices::new_expanded_vector): Update for new
16891 vec_perm_indices class.
16892 (vec_perm_indices::rotate_inputs): New function.
16893 (vec_perm_indices::all_in_range_p): Operate directly on the
16894 encoded form, without computing elided elements.
16895 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16896 encoding. Update for new vec_perm_indices class.
16897 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16898 the given vec_perm_builder.
16899 (expand_vec_perm_var): Update vec_perm_builder constructor.
16900 (expand_mult_highpart): Use vec_perm_builder instead of
16901 auto_vec_perm_indices.
16902 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16903 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16904 or double series encoding as appropriate.
16905 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16906 vec_perm_indices instead of auto_vec_perm_indices.
16907 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16908 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16909 (vect_permute_store_chain): Likewise.
16910 (vect_grouped_load_supported): Likewise.
16911 (vect_permute_load_chain): Likewise.
16912 (vect_shift_permute_load_chain): Likewise.
16913 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16914 (vect_transform_slp_perm_load): Likewise.
16915 (vect_schedule_slp_instance): Likewise.
16916 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16917 (vectorizable_mask_load_store): Likewise.
16918 (vectorizable_bswap): Likewise.
16919 (vectorizable_store): Likewise.
16920 (vectorizable_load): Likewise.
16921 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16922 vec_perm_indices instead of auto_vec_perm_indices. Use
16923 tree_to_vec_perm_builder to read the vector from a tree.
16924 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16925 vec_perm_builder instead of a vec_perm_indices.
16926 (have_whole_vector_shift): Use vec_perm_builder and
16927 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16928 truncation to calc_vec_perm_mask_for_shift.
16929 (vect_create_epilog_for_reduction): Likewise.
16930 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16931 from auto_vec_perm_indices to vec_perm_indices.
16932 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16933 instead of changing individual elements.
16934 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16935 the vector in d.perm.
16936 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16937 from auto_vec_perm_indices to vec_perm_indices.
16938 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16939 instead of changing individual elements.
16940 (arm_vectorize_vec_perm_const): Use new_vector to install
16941 the vector in d.perm.
16942 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16943 Update vec_perm_builder constructor.
16944 (rs6000_expand_interleave): Likewise.
16945 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16946 (rs6000_expand_interleave): Likewise.
16948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16950 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16951 to qimode could truncate the indices.
16952 * optabs.c (expand_vec_perm_var): Likewise.
16954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16956 * Makefile.in (OBJS): Add vec-perm-indices.o.
16957 * vec-perm-indices.h: New file.
16958 * vec-perm-indices.c: Likewise.
16959 * target.h (vec_perm_indices): Replace with a forward class
16961 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16962 * optabs.h: Include vec-perm-indices.h.
16963 (expand_vec_perm): Delete.
16964 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16965 (expand_vec_perm_const): Declare.
16966 * target.def (vec_perm_const_ok): Replace with...
16967 (vec_perm_const): ...this new hook.
16968 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16969 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16970 * doc/tm.texi: Regenerate.
16971 * optabs.def (vec_perm_const): Delete.
16972 * doc/md.texi (vec_perm_const): Likewise.
16973 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16974 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16975 expand_vec_perm for constant permutation vectors. Assert that
16976 the mode of variable permutation vectors is the integer equivalent
16977 of the mode that is being permuted.
16978 * optabs-query.h (selector_fits_mode_p): Declare.
16979 * optabs-query.c: Include vec-perm-indices.h.
16980 (selector_fits_mode_p): New function.
16981 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16982 is defined, instead of checking whether the vec_perm_const_optab
16983 exists. Use targetm.vectorize.vec_perm_const instead of
16984 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16985 fit in the vector mode before using a variable permute.
16986 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16987 vec_perm_indices instead of an rtx.
16988 (expand_vec_perm): Replace with...
16989 (expand_vec_perm_const): ...this new function. Take the selector
16990 as a vec_perm_indices rather than an rtx. Also take the mode of
16991 the selector. Update call to shift_amt_for_vec_perm_mask.
16992 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16993 Use vec_perm_indices::new_expanded_vector to expand the original
16994 selector into bytes. Check whether the indices fit in the vector
16995 mode before using a variable permute.
16996 (expand_vec_perm_var): Make global.
16997 (expand_mult_highpart): Use expand_vec_perm_const.
16998 * fold-const.c: Includes vec-perm-indices.h.
16999 * tree-ssa-forwprop.c: Likewise.
17000 * tree-vect-data-refs.c: Likewise.
17001 * tree-vect-generic.c: Likewise.
17002 * tree-vect-loop.c: Likewise.
17003 * tree-vect-slp.c: Likewise.
17004 * tree-vect-stmts.c: Likewise.
17005 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17007 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17008 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17009 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17010 (aarch64_vectorize_vec_perm_const): ...this new function.
17011 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17012 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17013 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17014 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17015 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17016 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17017 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17019 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17020 check for NEON modes.
17021 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17022 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17023 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17024 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17026 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17027 the old VEC_PERM_CONST conditions.
17028 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17029 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17030 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17031 (ia64_vectorize_vec_perm_const_ok): Merge into...
17032 (ia64_vectorize_vec_perm_const): ...this new function.
17033 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17034 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17035 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17036 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17037 * config/mips/mips.c (mips_expand_vec_perm_const)
17038 (mips_vectorize_vec_perm_const_ok): Merge into...
17039 (mips_vectorize_vec_perm_const): ...this new function.
17040 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17041 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17042 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17043 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17044 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17045 (rs6000_expand_vec_perm_const): Delete.
17046 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17048 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17049 (altivec_expand_vec_perm_const_le): Take each operand individually.
17050 Operate on constant selectors rather than rtxes.
17051 (altivec_expand_vec_perm_const): Likewise. Update call to
17052 altivec_expand_vec_perm_const_le.
17053 (rs6000_expand_vec_perm_const): Delete.
17054 (rs6000_vectorize_vec_perm_const_ok): Delete.
17055 (rs6000_vectorize_vec_perm_const): New function.
17056 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17057 an element count and rtx array.
17058 (rs6000_expand_extract_even): Update call accordingly.
17059 (rs6000_expand_interleave): Likewise.
17060 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17061 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17062 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17063 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17064 (rs6000_expand_vec_perm_const): Delete.
17065 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17066 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17067 (altivec_expand_vec_perm_const_le): Take each operand individually.
17068 Operate on constant selectors rather than rtxes.
17069 (altivec_expand_vec_perm_const): Likewise. Update call to
17070 altivec_expand_vec_perm_const_le.
17071 (rs6000_expand_vec_perm_const): Delete.
17072 (rs6000_vectorize_vec_perm_const_ok): Delete.
17073 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17074 reference to the SPE evmerge intructions.
17075 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17076 an element count and rtx array.
17077 (rs6000_expand_extract_even): Update call accordingly.
17078 (rs6000_expand_interleave): Likewise.
17079 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17080 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17082 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17086 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17087 vector mode and that that mode matches the mode of the data
17089 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17090 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17091 directly using expand_vec_perm_1 when forcing selectors into
17093 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17097 * optabs-query.h (can_vec_perm_p): Delete.
17098 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17099 * optabs-query.c (can_vec_perm_p): Split into...
17100 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17101 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17102 particular selector is valid.
17103 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17104 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17105 (vect_grouped_load_supported): Likewise.
17106 (vect_shift_permute_load_chain): Likewise.
17107 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17108 (vect_transform_slp_perm_load): Likewise.
17109 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17110 (vectorizable_bswap): Likewise.
17111 (vect_gen_perm_mask_checked): Likewise.
17112 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17113 implementations of variable permutation vectors into account
17114 when deciding which selector to use.
17115 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17116 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17117 with a false third argument.
17118 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17119 to test whether the constant selector is valid and can_vec_perm_var_p
17120 to test whether a variable selector is valid.
17122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17124 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17125 * optabs-query.c (can_vec_perm_p): Likewise.
17126 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17127 instead of vec_perm_indices.
17128 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17129 (vect_gen_perm_mask_checked): Likewise,
17130 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17131 (vect_gen_perm_mask_checked): Likewise,
17133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17135 * optabs-query.h (qimode_for_vec_perm): Declare.
17136 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17137 (qimode_for_vec_perm): ...this new function.
17138 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17140 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17142 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17143 does not have a conditional at the top.
17145 2018-01-02 Richard Biener <rguenther@suse.de>
17147 * ipa-inline.c (big_speedup_p): Fix expression.
17149 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17152 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17155 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17159 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17160 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17161 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17162 cond_taken_branch_cost 3->4.
17164 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17166 PR tree-optimization/83581
17167 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17168 TODO_cleanup_cfg if any changes have been made.
17170 PR middle-end/83608
17171 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17172 convert_modes if target mode has the right side, but different mode
17175 PR middle-end/83609
17176 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17177 last argument when extracting from CONCAT. If either from_real or
17178 from_imag is NULL, use expansion through memory. If result is not
17179 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17180 the parts directly to inner mode, if even that fails, use expansion
17183 PR middle-end/83623
17184 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17185 check for bswap in mode rather than HImode and use that in expand_unop
17188 Copyright (C) 2018 Free Software Foundation, Inc.
17190 Copying and distribution of this file, with or without modification,
17191 are permitted in any medium without royalty provided the copyright
17192 notice and this notice are preserved.