1 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4 definition statements before the early exit for statements that aren't
6 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8 (vect_transform_loop): ...here. Process pattern definition
9 statements without first checking whether the main pattern
10 statement is live or relevant.
12 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
14 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
17 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
19 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
20 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
21 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
22 (expand_block_compare): Change select_block_compare_mode call.
23 (expand_strncmp_align_check): Use new functions, fix comment.
24 (emit_final_str_compare_gpr): New function.
25 (expand_strn_compare): Refactor and clean up code.
26 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
28 2018-06-19 Tony Reix <tony.reix@atos.com>
29 Damien Bergamini <damien.bergamini@atos.com>
30 David Edelsohn <dje.gcc@gmail.com>
32 * collect2.c (static_obj): New variable.
33 (static_libs): New variable.
34 (is_in_list): Uncomment declaration.
35 (main): Track AIX libraries linked statically.
36 (is_in_list): Uncomment definition.
37 (scan_prog_file): Don't add AIX shared libraries initializer
38 to constructor list if linking statically.
40 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
42 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
44 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
46 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
48 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
51 2018-06-19 Martin Liska <mliska@suse.cz>
53 * config/i386/i386.c (ix86_can_inline_p): Do not use
54 ipa_fn_summaries::get_create.
55 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
57 (devirtualization_time_bonus): Likewise.
58 (ipcp_propagate_stage): Likewise.
59 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
60 (edge_set_predicate): Likewise.
61 (evaluate_conditions_for_known_args): Likewise.
62 (evaluate_properties_for_edge): Likewise.
63 (ipa_call_summary::reset): Tranform to ...
64 (ipa_call_summary::~ipa_call_summary): ... this.
65 (ipa_fn_summary::reset): Transform to ...
66 (ipa_fn_summary::~ipa_fn_summary): ... this.
67 (ipa_fn_summary_t::remove): Rename to ...
68 (ipa_fn_summary_t::remove_callees): ... this.
69 (ipa_fn_summary_t::duplicate): Use placement new
70 instead of memory copy.
71 (ipa_call_summary_t::duplicate): Likewise.
72 (ipa_call_summary_t::remove): Remove.
73 (dump_ipa_call_summary): Change get_create to get.
74 (ipa_dump_fn_summary): Dump only when summary exists.
75 (analyze_function_body): Use symbol_summary::get instead
77 (compute_fn_summary): Likewise.
78 (estimate_edge_devirt_benefit): Likewise.
79 (estimate_edge_size_and_time): Likewise.
80 (inline_update_callee_summaries): Likewise.
81 (remap_edge_change_prob): Likewise.
82 (remap_edge_summaries): Likewise.
83 (ipa_merge_fn_summary_after_inlining): Likewise.
84 (write_ipa_call_summary): Likewise.
85 (ipa_fn_summary_write): Likewise.
86 (ipa_free_fn_summary): Likewise.
87 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
88 (struct ipa_call_summary): Likewise.
89 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
91 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
92 (estimate_size_after_inlining): Likewise.
93 (estimate_growth): Likewise.
94 (growth_likely_positive): Likewise.
95 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
96 (inline_call): Likewise.
97 * ipa-inline.c (caller_growth_limits): Likewise.
98 (can_inline_edge_p): Likewise.
99 (can_inline_edge_by_limits_p): Likewise.
100 (compute_uninlined_call_time): Likewise.
101 (compute_inlined_call_time): Likewise.
102 (want_inline_small_function_p): Likewise.
103 (edge_badness): Likewise.
104 (update_caller_keys): Likewise.
105 (update_callee_keys): Likewise.
106 (inline_small_functions): Likewise.
107 (inline_to_all_callers_1): Likewise.
108 (dump_overall_stats): Likewise.
109 (early_inline_small_functions): Likewise.
110 (early_inliner): Likewise.
111 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
112 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
113 * ipa-pure-const.c (malloc_candidate_p): Likewise.
114 * ipa-split.c (execute_split_functions): Likewise.
115 * symbol-summary.h: Likewise.
116 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
118 2018-06-19 Richard Biener <rguenther@suse.de>
120 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
121 (vectorize_loops): ... here. Fix dbgcnt handling.
122 (try_vectorize_loop): Wrap try_vectorize_loop_1.
124 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
127 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
128 ieee128 argument takes up only one (vector) register, not two (floating
131 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
133 * gimplify.c (gimplify_init_constructor): Really never clear for an
134 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
136 2018-06-19 Richard Biener <rguenther@suse.de>
138 PR tree-optimization/86179
139 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
140 after failed recognition.
142 2018-06-18 Martin Sebor <msebor@redhat.com>
145 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
146 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
147 Handle integer subtraction.
148 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
149 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
151 2018-06-18 David Malcolm <dmalcolm@redhat.com>
153 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
154 param from rtx to rtx_insn *.
155 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
157 (frv_ifcvt_modify_insn): Likwise.
158 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
159 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
160 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
161 as_a <rtx_insn *> cast to local "unprotected_region" once
162 it's been established that it's not NULL or pc_rtx.
163 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
164 param "sethi" from rtx to rtx_insn *.
165 (nds32_group_float_insns): Likewise for param "insn".
166 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
168 (vax_output_int_subtract): Likewise.
169 * config/vax/vax.c (vax_output_int_add): Likewise for param
171 (vax_output_int_subtract): Likewise.
172 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
173 (emit_pattern_after): Likewise for param "after".
174 (emit_insn_after): Likewise.
175 (emit_jump_insn_after): Likewise.
176 (emit_call_insn_after): Likewise.
177 (emit_debug_insn_after): Likewise.
178 (emit_pattern_before): Likewise for param "before".
179 (emit_insn_before): Likewise.
180 (emit_jump_insn_before): Likewise.
181 * final.c (get_insn_template): Likewise for param "insn", removing
183 * output.h (get_insn_template): Likewise for 2nd param.
184 * rtl.h (emit_insn_before): Likewise.
185 (emit_jump_insn_before): Likewise.
186 (emit_debug_insn_before_noloc): Likewise.
187 (emit_insn_after): Likewise.
188 (emit_jump_insn_after): Likewise.
189 (emit_call_insn_after): Likewise.
190 (emit_debug_insn_after): Likewise.
191 (set_insn_deleted): Likewise for param.
193 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
196 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
197 floating point modes, so that IFmode is numerically greater than
198 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
199 to declare the ordering. This prevents IFmode from being
200 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
201 machine. Include rs6000-modes.h to share the fractional values
202 between genmodes* and the rest of the compiler.
206 * config/rs6000/rs6000-modes.h: New file.
207 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
208 meaning of rs6000_long_double_size so that 126..128 selects an
209 appropriate 128-bit floating point type.
210 (rs6000_option_override_internal): Likewise.
211 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
212 (TARGET_LONG_DOUBLE_128): Change the meaning of
213 rs6000_long_double_size so that 126..128 selects an appropriate
214 128-bit floating point type.
215 (LONG_DOUBLE_TYPE_SIZE): Update comment.
216 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
217 source and destination to match the standard usage.
218 (truncifkf2): Likewise.
219 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
220 ISA 2.07 to use an explicit clobber, instead of passing in a
222 (copysign<mode>3_soft): Likewise.
224 2018-06-18 David Malcolm <dmalcolm@redhat.com>
226 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
227 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
228 (vect_slp_analyze_instance_dependence): Likewise.
229 (vect_enhance_data_refs_alignment): Likewise.
230 (vect_analyze_data_refs_alignment): Likewise.
231 (vect_slp_analyze_and_verify_instance_alignment
232 (vect_analyze_data_ref_accesses): Likewise.
233 (vect_prune_runtime_alias_test_list): Likewise.
234 (vect_analyze_data_refs): Likewise.
235 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
236 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
237 (vect_analyze_scalar_cycles_1): Likewise.
238 (vect_get_loop_niters): Likewise.
239 (vect_analyze_loop_form_1): Likewise.
240 (vect_update_vf_for_slp): Likewise.
241 (vect_analyze_loop_operations): Likewise.
242 (vect_analyze_loop): Likewise.
243 (vectorizable_induction): Likewise.
244 (vect_transform_loop): Likewise.
245 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
246 * tree-vect-slp.c (vect_analyze_slp): Likewise.
247 (vect_make_slp_decision): Likewise.
248 (vect_detect_hybrid_slp): Likewise.
249 (vect_slp_analyze_operations): Likewise.
250 (vect_slp_bb): Likewise.
251 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
252 (vectorizable_bswap): Likewise.
253 (vectorizable_call): Likewise.
254 (vectorizable_simd_clone_call): Likewise.
255 (vectorizable_conversion): Likewise.
256 (vectorizable_assignment): Likewise.
257 (vectorizable_shift): Likewise.
258 (vectorizable_operation): Likewise.
259 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
261 2018-06-18 Martin Sebor <msebor@redhat.com>
263 PR tree-optimization/81384
264 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
265 * builtins.c (expand_builtin_strnlen): New function.
266 (expand_builtin): Call it.
267 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
268 * builtins.def (BUILT_IN_STRNLEN): New.
269 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
270 Warn for bounds in excess of maximum object size.
271 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
272 single-value ranges. Handle strnlen.
273 (handle_builtin_strlen): Handle strnlen.
274 (strlen_check_and_optimize_stmt): Same.
275 * doc/extend.texi (Other Builtins): Document strnlen.
277 2018-06-18 Maya Rashish <coypu@sdf.org>
279 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
280 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
281 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
283 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
285 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
287 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
289 * tree.c (escaped_string::escape): Replace cast to char * by
290 const_cast<char *> (unescaped).
292 2018-06-18 Nick Clifton <nickc@redhat.com>
295 * tree.c (escaped_string): New class. Converts an unescaped
296 string into its escaped equivalent.
297 (warn_deprecated_use): Use the new class to convert the
298 deprecation message, if present.
299 (test_escaped_strings): New self test.
300 (test_c_tests): Add test_escaped_strings.
301 * doc/extend.texi (deprecated): Add a note that the
302 deprecation message is affected by the -fmessage-length
303 option, and that control characters will be escaped.
304 (#pragma GCC error): Document this pragma.
305 (#pragma GCC warning): Likewise.
306 * doc/invoke.texi (-fmessage-length): Document this option's
307 effect on the #warning and #error preprocessor directives and
308 the deprecated attribute.
310 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
312 * tree.c (decl_value_expr_lookup): Revert latest change.
313 (decl_value_expr_insert): Likewise.
315 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
317 * gimplify.c (nonlocal_vlas): Delete.
318 (nonlocal_vla_vars): Likewise.
319 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
321 (gimplify_body): Do not create and destroy nonlocal_vlas.
322 * tree-nested.c: Include diagnostic.h.
323 (use_pointer_in_frame): Tweak.
324 (lookup_field_for_decl): Add assertion and declare the transformation.
325 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
326 internal error when the reference is in a wrong context. Do not
327 create a debug decl by default.
328 (note_nonlocal_block_vlas): Delete.
329 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
330 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
331 create a debug decl by default.
332 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
333 call is in a wrong context.
334 (fixup_vla_decls): New function.
335 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
336 debug variables were created.
337 * tree.c (decl_value_expr_lookup): Add checking assertion.
338 (decl_value_expr_insert): Likewise.
340 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
343 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
344 * tree-scalar-evolution.c (interpret_expr): Likewise.
345 (expression_expensive_p): Likewise.
346 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
347 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
348 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
349 (ssa_defined_by_minus_one_stmt_p): New.
351 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
354 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
355 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
356 * dojump.c (do_jump): Likewise.
357 * expr.c (expand_expr_real_2): Check operand type's sign.
358 * fold-const.c (const_unop): Handle ABSU_EXPR.
359 (fold_abs_const): Likewise.
360 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
361 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
362 (strip_sign_op_1): Likesise.
363 * match.pd: Add new pattern to generate ABSU_EXPR.
364 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
365 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
366 * tree-eh.c (operation_could_trap_helper_p): Likewise.
367 * tree-inline.c (estimate_operator_cost): Likewise.
368 * tree-pretty-print.c (dump_generic_node): Likewise.
369 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
370 * tree.def (ABSU_EXPR): New.
372 2018-06-16 Jakub Jelinek <jakub@redhat.com>
375 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
376 documented as preserved for backward compatibility only.
377 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
379 PR rtl-optimization/86108
380 * bb-reorder.c (create_forwarder_block): Renamed to ...
381 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
382 jump from new landing pad to the second part.
383 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
386 2018-06-15 Jakub Jelinek <jakub@redhat.com>
389 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
390 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
391 Only call store_expr for halves if the mode is the same.
394 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
395 Fix up comment formatting.
397 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
399 * typed-splay-tree.h (typed_splay_tree::remove): New function.
400 (typed_splay_tree::closure,
401 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
402 (typed_splay_tree::typed_splay_tree,
403 typed_splay_tree::operator =): Declared private.
404 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
405 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
406 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
407 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
408 typed_splay_tree::splay_tree_splay,
409 typed_splay_tree::splay_tree_foreach_helper,
410 typed_splay_tree::splay_tree_insert,
411 typed_splay_tree::splay_tree_remove,
412 typed_splay_tree::splay_tree_lookup,
413 typed_splay_tree::splay_tree_predecessor,
414 typed_splay_tree::splay_tree_successor,
415 typed_splay_tree::splay_tree_min,
416 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
417 (typed_splay_tree::root, typed_splay_tree::comp,
418 typed_splay_tree::delete_key,
419 typed_splay_tree::delete_value): New data members.
420 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
421 typed_splay_tree::remove.
423 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
425 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
426 -mginv and -mno-ginv to the assembler.
427 * config/mips/mips.opt (-mcrc): New option.
429 * doc/invoke.text (-mcrc): Document.
432 2018-06-15 Nick Clifton <nickc@redhat.com>
435 * tree.c (escaped_string): New class. Converts an unescaped
436 string into its escaped equivalent.
437 (warn_deprecated_use): Use the new class to convert the
438 deprecation message, if present.
439 (test_escaped_strings): New self test.
440 (test_c_tests): Add test_escaped_strings.
441 * doc/extend.texi (deprecated): Add a note that the
442 deprecation message is affected by the -fmessage-length
443 option, and that control characters will be escaped.
444 (#pragma GCC error): Document this pragma.
445 (#pragma GCC warning): Likewise.
446 * doc/invoke.texi (-fmessage-length): Document this option's
447 effect on the #warning and #error preprocessor directives and
448 the deprecated attribute.
450 2018-06-15 Richard Biener <rguenther@suse.de>
452 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
453 here, also noting vector size used.
454 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
455 size used in MSG_OPTIMIZED_LOCATIONS dump.
456 (pass_slp_vectorize::execute): Adjust.
458 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
461 * config/arc/arc.c (arc_return_address_register): Fix
464 2018-06-15 Richard Biener <rguenther@suse.de>
467 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
468 leave useless conversion stripping to force_gimple_operand_gsi.
469 (gimplify_build2): Likewise.
470 (gimplify_build1): Likewise.
472 2018-06-15 Richard Biener <rguenther@suse.de>
475 * tree-cfg.c (move_stmt_op): unshare invariant addresses
476 before adjusting their block.
478 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
480 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
481 multilibs for *-*-rtems*.
482 * config/riscv/t-rtems: New file.
484 2018-06-14 Jakub Jelinek <jakub@redhat.com>
487 * match.pd ((A +- CST1) +- CST2): Punt if last resort
488 unsigned_type_for returns NULL.
491 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
492 subregs of multi-word pseudos unless the float mode has word size.
494 2018-06-14 Richard Biener <rguenther@suse.de>
497 * tree-vect-generic.c (build_word_mode_vector_type): Remove
498 duplicate and harmful type_hash_canon.
499 * tree.c (type_hash_canon): Assert we didn't find ourselves.
501 2018-06-14 Richard Biener <rguenther@suse.de>
504 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
507 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
509 * config/rtems.h (STDINT_LONG32): Define.
511 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
512 Prachi Godbole <prachi.godbole@imgtec.com>
514 * config/mips/mips-cpus.def: Define P6600.
515 * config/mips/mips-tables.opt: Regenerate.
516 * config/mips/mips.c (mips_ucbranch_type): New enum.
517 (mips_rtx_cost_data): Add support for P6600.
518 (mips_issue_rate): Likewise.
519 (mips_multipass_dfa_lookahead): Likewise.
520 (mips_avoid_hazard): Likewise.
521 (mips_reorg_process_insns): Likewise.
522 (mips_classify_branch_p6600): New function.
523 * config/mips/mips.h (TUNE_P6600): New define.
524 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
525 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
526 * config/mips/mips.md: Include p6600.md.
527 (processor): Add p6600.
528 * config/mips/p6600.md: New file.
529 * doc/invoke.texi: Add p6600 to supported architectures.
531 2018-06-13 Martin Sebor <msebor@redhat.com>
533 PR tree-optimization/86114
534 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
536 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
538 2018-06-13 Richard Biener <rguenther@suse.de>
540 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
541 Properly set vector type of the intermediate stmt.
542 * tree-vect-stmts.c (vectorizable_operation): The destination
543 var always has vectype_out type.
545 2018-06-13 Jeff Law <law@redhat.com>
547 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
548 integer 0 for argument to print_rtl_with_bb.
549 (rl78_reorg): Likewise.
551 2018-06-13 David Malcolm <dmalcolm@redhat.com>
553 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
554 from rtx to rtx_insn *.
555 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
557 (add_sched_insns_for_speculation): Likewise for local "target",
558 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
559 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
560 from rtx_insn ** to rtx_code_label **.
561 (reorg_emit_nops): Likewise.
562 (c6x_reorg): Likewise for local "call_labels".
563 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
565 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
566 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
567 the loops over LABEL_REFS.
568 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
570 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
571 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
572 (split_branches): Strengthen local "olabel" from rtx to
573 rtx_insn *, adding a safe_as_a cast.
574 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
576 (add_insn_after): Likewise for first two params.
577 (add_insn_before): Likewise.
578 (remove_insn): Likewise for param.
579 (emit_pattern_before_noloc): Likewise for second and third params.
580 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
581 (emit_call_insn_before_noloc): Likewise.
582 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
584 (emit_barrier_before): Likewise.
585 (emit_label_before): Strengthen "label" param from "rtx" to
586 "rtx_code_label *". Strengthen "before" param from "rtx" to
588 (emit_insn_after_1): Strengthen "after" param from "rtx" to
590 (emit_pattern_after_noloc): Likewise.
591 (emit_insn_after_noloc): Likewise.
592 (emit_jump_insn_after_noloc): Likewise.
593 (emit_call_insn_after_noloc): Likewise.
594 (emit_debug_insn_after_noloc): Likewise.
595 (emit_barrier_after): Likewise.
596 (emit_label_after): Likewise for both params.
597 (emit_pattern_after_setloc): Likewise for "after" param. Convert
598 "loc" param from "int" to "location_t".
599 (emit_insn_after_setloc): Likewise.
600 (emit_jump_insn_after_setloc): Likewise.
601 (emit_call_insn_after_setloc): Likewise.
602 (emit_debug_insn_after_setloc): Likewise.
603 (emit_pattern_before_setloc): Likewise for "before" param. Convert
604 "loc" param from "int" to "location_t".
605 (emit_pattern_before): Convert NULL_RTX to NULL.
606 (emit_insn_before_setloc): Convert "loc" param from "int" to
608 (emit_jump_insn_before_setloc): Likewise.
609 (emit_call_insn_before_setloc): Likewise.
610 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
611 rtx_insn *. Convert "loc" param from "int" to "location_t".
612 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
613 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
614 Convert 3rd param from "int" to "location_t".
615 (emit_barrier_before, emit_barrier_after, next_real_insn):
616 Strengthen param from rtx to rtx_insn *.
617 (emit_label_before): Strengthen 1st param from "rtx" to
618 "rtx_code_label *". Strengthen 2nd param from "rtx" to
620 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
621 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
622 Strengthen 2nd param from "rtx" to "rtx_insn *".
623 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
624 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
625 Likewise. Convert 3rd param from "int" to "location_t".
626 (emit_label_after): Strengthen 1st param from "rtx" to
628 (next_real_insn, remove_insn): Strengthen param from "rtx" to
630 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
631 from "rtx" to "rtx_insn *".
633 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
635 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
636 bodies streamed in with -Q.
637 * dumpfile.c (dump_files): Add lto-stream-out dump file.
638 * dumpfile.h (tree_dump_index): Add lto_stream_out.
639 * gimple-streamer-out.c: Include gimple-pretty-print.h
640 (output_bb): Dump stmts streamed.
641 * lto-section-out.c: Include print-tree.h
642 (lto_begin_section): Dump sections created.
643 (lto_output_decl_index): Dump decl encoded.
644 * lto-streamer-out.c: Include print-tree.h
645 (create_output_block): Dump output block created.
646 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
647 (output_function): Dump function output.
648 (output_constructor): Dump constructor streamed.
649 (write_global_stream): Output indexes encoded.
650 (produce_asm_for_decls): Dump streams encoded.
651 * lto-streamer.c (streamer_dump_file): New global var.
652 * lto-streamer.h (streamer_dump_file): Declare.
653 * passes.c (ipa_write_summaries): Initialize streamer dump.
654 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
657 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
660 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
661 offsets for register save directives. Emit a second batch of save
662 directives, if need be, when the function accesses prior frames.
664 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
666 * config/arc/fpu.md (fmasf4): Force operand to register.
669 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
671 * config/arc/arc-protos.h (arc_pad_return): Remove.
672 * config/arc/arc.c (machine_function): Remove force_short_suffix
674 (arc_print_operand): Adjust printing of '&'.
675 (arc_verify_short): Remove conditional printing of short suffix.
676 (arc_final_prescan_insn): Remove reference to size_reason.
677 (pad_return): New function.
678 (arc_reorg): Call pad_return.
679 (arc_pad_return): Remove.
680 (arc_init_machine_status): Remove reference to force_short_suffix.
681 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
682 (attr length): When attribute iscompact is true force to 2
683 regardless; in the case of maybe check if we want to force the
684 instruction to have 4 bytes length.
685 (nopv): Change it to generate 4 byte long nop as well.
686 (blockage): New pattern.
687 (simple_return): Remove call to arc_pad_return.
688 (p_return_i): Likewise.
690 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
692 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
694 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
696 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
699 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
701 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
702 for ARC700 and ARCv2.
704 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
707 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
708 operands[2] instead of operands[1].
711 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
713 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
714 case, check whether the outer register overlaps an unallocatable
715 register, not just whether it fits the required class.
717 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
719 * poly-int.h (can_div_trunc_p): Add new overload in which all values
721 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
722 (memrefs_conflict_p): Likewise.
723 (init_alias_analysis): Likewise.
724 * cfgexpand.c (expand_debug_expr): Likewise.
725 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
726 * cse.c (fold_rtx): Likewise.
727 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
728 * expr.c (emit_block_move_hints): Likewise.
729 (clear_storage_hints, push_block, emit_push_insn): Likewise.
730 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
731 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
732 (emit_group_store): Likewise.
733 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
734 to read the PRE/POST_MODIFY increment.
735 * calls.c (store_one_arg): Use strip_offset.
736 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
738 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
740 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
741 (simplify_binary_operation_1): Extend CONST_INT handling to
743 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
744 than a HOST_WIDE_INT.
745 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
747 (adjust_mems, add_stores): Update accodingly.
748 (vt_canonicalize_addr): Track polynomial offsets.
749 (emit_note_insn_var_location): Likewise.
750 (vt_add_function_parameter): Likewise.
751 (vt_initialize): Likewise.
753 2018-06-12 Jeff Law <law@redhat.com>
755 * config.gcc (alpha*-*-freebsd*): Remove.
756 * config/alpha/freebsd.h: Remove.
758 2018-06-12 David Malcolm <dmalcolm@redhat.com>
761 * spellcheck-tree.c (levenshtein_distance): Rename to...
762 (get_edit_distance): ...this, and update for underlying renaming.
763 * spellcheck-tree.h (levenshtein_distance): Rename to...
764 (get_edit_distance): ...this.
765 * spellcheck.c (levenshtein_distance): Rename to...
766 (get_edit_distance): ...this. Convert from Levenshtein distance
767 to Damerau-Levenshtein distance by supporting transpositions of
768 adjacent characters. Rename "v1" to "v_next" and "v0" to
770 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
771 (selftest::test_edit_distance_unit_test_oneway): ...this, and
772 update for underlying renaming.
773 (selftest::levenshtein_distance_unit_test): Rename to...
774 (selftest::test_get_edit_distance_unit): ...this, and update for
776 (selftest::test_find_closest_string): Add example from PR 69968
777 where transposition helps
778 (selftest::test_metric_conditions): Update for renaming.
779 (selftest::test_metric_conditions): Likewise.
780 (selftest::spellcheck_c_tests): Likewise.
781 * spellcheck.h (levenshtein_distance): Rename both overloads to...
782 (get_edit_distance): ...this.
783 (best_match::consider): Update for renaming.
785 2018-06-12 Martin Sebor <msebor@redhat.com>
787 PR tree-optimization/85259
788 * builtins.c (compute_objsize): Handle constant offsets.
789 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
790 true iff a warning has been issued.
791 * gimple.h (gimple_nonartificial_location): New function.
792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
793 gimple_nonartificial_location and handle -Wno-system-headers.
794 (handle_builtin_stxncpy): Same.
796 2018-06-12 Martin Sebor <msebor@redhat.com>
799 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
801 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
803 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
804 BUILTIN_VEC_XST entries for pointer to double and long long.
806 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
809 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
811 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
814 2018-06-12 Martin Liska <mliska@suse.cz>
816 * doc/options.texi: Document IntegerRange.
818 2018-06-12 Martin Liska <mliska@suse.cz>
820 * config/i386/i386.opt: Make MPX-related options as Deprecated.
821 * opt-functions.awk: Handle Deprecated flag.
822 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
824 (read_cmdline_option): Report warning for a deprecated option.
825 * opts.h (struct cl_option): Add new field cl_deprecated.
826 (CL_ERR_DEPRECATED): New.
828 2018-06-12 Martin Liska <mliska@suse.cz>
830 * doc/options.texi: Document Deprecated option flag.
832 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
834 * config/arc/arc-arch.h (arc_extras): New enum.
835 (arc_cpu_t):Add field extra.
836 (arc_cpu_types): Consider the extras.
837 * config/arc/arc-cpus.def: Add extras info.
838 * config/arc/arc-opts.h (processor_type): Consider extra field.
839 * config/arc/arc.c (arc_override_options): Handle extra field.
841 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
843 * config/arc/arc-arch.h: Update ARC_OPTX macro.
844 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
846 * config/arc/arc.c (arc_init): Update pic warning.
847 (irq_range): Update irq range parsing warnings.
848 (arc_override_options): Update various warning messages.
849 (arc_handle_aux_attribute): Likewise.
851 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
853 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
855 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
857 * doc/sourcebuild.texi: Document usage of line number 0 in verify
858 compiler messages directives.
860 2018-06-12 Matthew Fortune <mfortune@gmail.com>
862 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
863 * config/mips/mips-tables.opt: Regenerate.
864 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
866 * doc/invoke.texi: Document -march=i6500.
868 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
870 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
871 (i6400_gpmul): Add cpu_unit.
872 (i6400_gpdiv): Likewise.
873 (i6400_msa_add_d): Update reservations.
874 (i6400_msa_int_add) Likewise.
875 (i6400_msa_short_logic3) Likewise.
876 (i6400_msa_short_logic2) Likewise.
877 (i6400_msa_short_logic) Likewise.
878 (i6400_msa_move) Likewise.
879 (i6400_msa_cmp) Likewise.
880 (i6400_msa_short_float2) Likewise.
881 (i6400_msa_div_d) Likewise.
882 (i6400_msa_long_logic1) Likewise.
883 (i6400_msa_long_logic2) Likewise.
884 (i6400_msa_mult) Likewise.
885 (i6400_msa_long_float2) Likewise.
886 (i6400_msa_long_float4) Likewise.
887 (i6400_msa_long_float5) Likewise.
888 (i6400_msa_long_float8) Likewise.
889 (i6400_fpu_fadd): Include frint type.
890 (i6400_fpu_store): New define_insn_reservation.
891 (i6400_fpu_load): Likewise.
892 (i6400_fpu_move): Likewise.
893 (i6400_fpu_fcmp): Likewise.
894 (i6400_fpu_fmadd): Likewise.
895 (i6400_int_mult): Include imul3nc type and update reservation.
896 (i6400_int_div): Include idiv3 type and update reservation.
897 (i6400_int_load): Update to check type not move_type.
898 (i6400_int_store): Likewise.
899 (i6400_int_prefetch): Set zero latency.
901 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
903 * gcc.c: Document new %@{...} sequence.
904 (LINK_COMMAND_SPEC): Use it for the -L switches.
905 (cpp_unique_options): Use it for the -I switches.
906 (at_file_argbuf): New global variable.
907 (in_at_file): Likewise.
908 (alloc_args): Create at_file_argbuf.
909 (clear_args): Truncate at_file_argbuf.
910 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
911 (open_at_file): New function.
912 (close_at_file): Likewise.
913 (create_at_file): Delete.
914 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
917 (validate_switches_from_spec): Deal with %@{...} sequence.
918 (validate_switches): Likewise.
919 (driver::finalize): Call clear_args.
921 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
923 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
925 2018-06-11 Martin Sebor <msebor@redhat.com>
927 * doc/invoke.texi (-Wall): List -Wc++17-compat.
928 (Wno-class-memaccess): Add @opindex.
929 (Wno-templates, Wno-multiple-inheritance): Same.
930 (Wno-virtual-inheritance, Wno-namespaces): Same.
931 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
932 (Wno-format-overflow, Wno-format-truncation): Same.
933 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
934 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
935 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
936 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
937 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
938 (Wno-misspelled-isr): Same.
940 2018-06-11 Martin Sebor <msebor@redhat.com>
942 * PR tree-optimization/86083
943 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
945 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
947 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
949 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
952 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
953 on the correct operand.
954 (*movdi_internal64): Ditto.
956 2018-06-11 Martin Liska <mliska@suse.cz>
958 PR tree-optimization/86089
959 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
961 2018-06-11 Julia Koval <julia.koval@intel.com>
963 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
964 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
965 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
967 2018-06-11 Olivier Hainque <hainque@adacore.com>
969 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
970 for Ada with strict dwarf2.
972 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
975 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
978 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
980 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
982 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
984 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
987 2018-06-08 Martin Liska <mliska@suse.cz>
989 * tree-cfg.h (debug_function): Fix argument type to match
992 2018-06-08 Martin Liska <mliska@suse.cz>
994 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
995 Remove usage of MPX-related (and removed) fields.
996 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
998 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1000 * cfg.c (debug): Use TDF_NONE rather than 0.
1001 * cfghooks.c (debug): Likewise.
1002 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1003 (struct dump_option_value_info): Convert to...
1004 (struct kv_pair): ...this template type.
1005 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1007 (optinfo_verbosity_options): Likewise.
1008 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1010 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1011 than int for "optgroup_flags" param.
1012 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1014 (dump_dec): Likewise.
1015 (dump_finish): Use TDF_NONE rather than 0.
1016 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1017 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1018 than 0. Update for change to option_ptr.
1019 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1020 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1021 0. Update for changes to optinfo_verbosity_options and
1023 (opt_info_switch_p): Convert optgroup_flags from int to
1025 (dump_basic_block): Use dump_flags_t rather than int
1026 for "dump_kind" param.
1027 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1028 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1029 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1030 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1031 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1032 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1033 TDF_NONE): Convert from macros to...
1034 (enum dump_flag): ...this new enum.
1035 (dump_flags_t): Update to use enum.
1036 (operator|, operator&, operator~, operator|=, operator&=):
1037 Implement for dump_flags_t.
1038 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1039 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1040 Convert from macros to...
1041 (enum optgroup_flag): ...this new enum.
1042 (optgroup_flags_t): New typedef.
1043 (operator|, operator|=): Implement for optgroup_flags_t.
1044 (struct dump_file_info): Convert field "alt_flags" to
1045 dump_flags_t. Convert field "optgroup_flags" to
1047 (dump_basic_block): Use dump_flags_t rather than int for param.
1048 (dump_generic_expr_loc): Likewise.
1049 (dump_dec): Likewise.
1050 (dump_register): Convert param "optgroup_flags" to
1052 (opt_info_enable_passes): Likewise.
1053 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1055 * gimple-pretty-print.c (debug): Likewise.
1056 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1057 (merged_store_group::apply_stores): Likewise.
1058 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1059 * gimple.c (verify_gimple_pp): Likewise.
1060 * graphite-poly.c (print_pbb_body): Likewise.
1061 * passes.c (pass_manager::register_one_dump_file): Convert
1062 local "optgroup_flags" to optgroup_flags_t.
1063 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1065 (debug_body): Likewise.
1066 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1067 to optgroup_flags_t.
1068 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1070 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1071 (convert_mult_to_fma): Likewise.
1072 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1073 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1074 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1075 "dump_kind" to dump_flags_t.
1077 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1079 * config/rs6000/rs6000.c (min, max): Delete.
1081 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1083 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1084 -mabi=spe and -mabi=no-spe.
1086 2018-06-08 Martin Liska <mliska@suse.cz>
1088 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1089 where we expect an existing summary.
1091 2018-06-08 Martin Liska <mliska@suse.cz>
1093 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1094 * ipa-inline.h (estimate_edge_growth): Likewise.
1096 2018-06-08 Martin Liska <mliska@suse.cz>
1098 * cgraph.c (function_version_hasher::hash): Use
1099 cgraph_node::get_uid ().
1100 (function_version_hasher::equal):
1101 * cgraph.h (cgraph_node::get_uid): New method.
1102 * ipa-inline.c (update_caller_keys): Use
1103 cgraph_node::get_uid ().
1104 (update_callee_keys): Likewise.
1105 * ipa-utils.c (searchc): Likewise.
1106 (ipa_reduced_postorder): Likewise.
1107 * lto-cgraph.c (input_node): Likewise.
1108 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1109 * symbol-summary.h (symtab_insertion): Likewise.
1110 (symtab_removal): Likewise.
1111 (symtab_duplication): Likewise.
1112 * tree-pretty-print.c (dump_function_header): Likewise.
1113 * tree-sra.c (convert_callers_for_node): Likewise.
1115 2018-06-08 Martin Liska <mliska@suse.cz>
1117 * cgraph.c (symbol_table::create_edge): Always assign a new
1119 (symbol_table::free_edge): Do not recycle numbers.
1120 * cgraph.h (cgraph_edge::get): New method.
1121 * symbol-summary.h (symtab_removal): Use it.
1122 (symtab_duplication): Likewise.
1123 (call_summary::hashable_uid): Remove.
1125 2018-06-08 Martin Liska <mliska@suse.cz>
1127 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1128 (initialize_growth_caches): Remove.
1129 (free_growth_caches): Likewise.
1130 (do_estimate_edge_time): Use edge_growth_cache.
1131 (do_estimate_edge_size): Likewise.
1132 (do_estimate_edge_hints): Likewise.
1133 * ipa-inline.c (reset_edge_caches): Likewise.
1134 (recursive_inlining): Likewise.
1135 (inline_small_functions): Likewise.
1136 * ipa-inline.h (initialize_growth_caches): Remove.
1137 (estimate_edge_size): Likewise.
1138 (estimate_edge_time): Likewise.
1139 (estimate_edge_hints): Likewise.
1140 (reset_edge_growth_cache): Likewise.
1141 * symbol-summary.h (call_summary::remove): New method.
1143 2018-06-08 Martin Liska <mliska@suse.cz>
1145 * ipa-cp.c (class edge_clone_summary): New summary.
1146 (grow_edge_clone_vectors): Remove.
1147 (ipcp_edge_duplication_hook): Remove.
1148 (class edge_clone_summary_t): New call_summary class.
1149 (ipcp_edge_removal_hook): Remove.
1150 (edge_clone_summary_t::duplicate): New function.
1151 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1152 (create_specialized_node): Likewise.
1153 (ipcp_driver): Initialize edge_clone_summaries and do not
1156 2018-06-08 Martin Liska <mliska@suse.cz>
1158 * symbol-summary.h (get): New function.
1159 (call_summary::m_initialize_when_cloning): New class member.
1161 2018-06-08 Martin Liska <mliska@suse.cz>
1163 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1164 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1165 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1166 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1167 (remove_cgraph_node_from_order): Use the removed_nodes set.
1168 (do_per_function_toporder): Likwise.
1169 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1170 instead of summary_uid.
1171 (symtab_removal): Likewise.
1172 (symtab_duplication): Likewise.
1174 2018-06-08 Martin Liska <mliska@suse.cz>
1176 * ipa-cp.c (ipcp_store_bits_results): Use
1177 ipcp_transformation_sum.
1178 (ipcp_store_vr_results): Likewise.
1179 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1181 (ipcp_transformation_initialize): ... this.
1182 (ipa_set_node_agg_value_chain):
1183 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1184 (write_ipcp_transformation_info): Likewise.
1185 (read_ipcp_transformation_info): Likewise.
1186 (ipcp_update_bits): Likewise.
1187 (ipcp_update_vr): Likewise.
1188 (ipcp_transform_function): Likewise.
1189 * ipa-prop.h: Rename ipcp_transformation_summary to
1190 ipcp_transformation.
1191 (class ipcp_transformation_t): New function summary.
1192 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1193 (ipa_get_agg_replacements_for_node): Likewise.
1195 2018-06-08 Martin Liska <mliska@suse.cz>
1197 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1199 (class funct_state_summary_t): New function_summary class.
1200 (has_function_state): Remove.
1201 (get_function_state): Likewise.
1202 (set_function_state): Likewise.
1203 (add_new_function): Likewise.
1204 (funct_state_summary_t::insert): New function.
1205 (duplicate_node_data): Remove.
1206 (remove_node_data): Remove.
1207 (funct_state_summary_t::duplicate): New function.
1208 (register_hooks): Create new funct_state_summaries.
1209 (pure_const_generate_summary): Use it.
1210 (pure_const_write_summary): Likewise.
1211 (pure_const_read_summary): Likewise.
1212 (propagate_pure_const): Likewise.
1213 (propagate_nothrow): Likewise.
1214 (dump_malloc_lattice): Likewise.
1215 (propagate_malloc): Likewise.
1216 (execute): Do not register hooks, just remove summary
1218 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1221 2018-06-08 Martin Liska <mliska@suse.cz>
1223 * ipa-reference.c (remove_node_data): Remove.
1224 (duplicate_node_data): Likewise.
1225 (class ipa_ref_var_info_summary_t): New class.
1226 (class ipa_ref_opt_summary_t): Likewise.
1227 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1228 (get_reference_optimization_summary): Use
1229 ipa_ref_opt_sum_summaries.
1230 (set_reference_vars_info): Remove.
1231 (set_reference_optimization_summary): Likewise.
1232 (ipa_init): Create summaries.
1233 (init_function_info): Use function summary.
1234 (ipa_ref_opt_summary_t::duplicate): New function.
1235 (ipa_ref_opt_summary_t::remove): New function.
1236 (get_read_write_all_from_node): Fix GNU coding style.
1237 (propagate): Use function summary.
1238 (write_node_summary_p): Fix GNU coding style.
1239 (stream_out_bitmap): Likewise.
1240 (ipa_reference_read_optimization_summary): Use function summary.
1241 (ipa_reference_c_finalize): Do not release hooks.
1243 2018-06-08 Martin Liska <mliska@suse.cz>
1245 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1246 (analyze_function_body): Extract multiple calls of get_create.
1247 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1248 * ipa-inline.c (recursive_inlining): Use ::get method.
1249 * ipa-inline.h (estimate_edge_growth): Likewise.
1251 2018-06-08 Martin Liska <mliska@suse.cz>
1253 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1255 (hsa_function_summary::hsa_function_summary): Use the new enum
1257 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1258 * hsa-gen.c (hsa_get_host_function): Likewise.
1259 (get_brig_function_name): Likewise.
1260 * ipa-hsa.c (process_hsa_functions): Likewise.
1261 (ipa_hsa_write_summary): Likewise.
1262 * symbol-summary.h (symtab_duplication): Use ::get function/
1263 (get): New function.
1265 2018-06-08 Martin Liska <mliska@suse.cz>
1267 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1269 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1270 (hsa_register_kernel): Likewise.
1271 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1272 * hsa-gen.c (hsa_get_host_function): Likewise.
1273 (get_brig_function_name): Likewise.
1274 (generate_hsa): Likewise.
1275 (pass_gen_hsail::execute): Likewise.
1276 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1277 (devirtualization_time_bonus): Likewise.
1278 (ipcp_propagate_stage): Likewise.
1279 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1280 (edge_set_predicate): Likewise.
1281 (evaluate_conditions_for_known_args): Likewise.
1282 (evaluate_properties_for_edge): Likewise.
1283 (ipa_fn_summary::reset): Likewise.
1284 (ipa_fn_summary_t::duplicate): Likewise.
1285 (dump_ipa_call_summary): Likewise.
1286 (ipa_dump_fn_summary): Likewise.
1287 (analyze_function_body): Likewise.
1288 (compute_fn_summary): Likewise.
1289 (estimate_edge_devirt_benefit): Likewise.
1290 (estimate_edge_size_and_time): Likewise.
1291 (estimate_calls_size_and_time): Likewise.
1292 (estimate_node_size_and_time): Likewise.
1293 (inline_update_callee_summaries): Likewise.
1294 (remap_edge_change_prob): Likewise.
1295 (remap_edge_summaries): Likewise.
1296 (ipa_merge_fn_summary_after_inlining): Likewise.
1297 (ipa_update_overall_fn_summary): Likewise.
1298 (read_ipa_call_summary): Likewise.
1299 (inline_read_section): Likewise.
1300 (write_ipa_call_summary): Likewise.
1301 (ipa_fn_summary_write): Likewise.
1302 (ipa_free_fn_summary): Likewise.
1303 * ipa-hsa.c (process_hsa_functions): Likewise.
1304 (ipa_hsa_write_summary): Likewise.
1305 (ipa_hsa_read_section): Likewise.
1306 * ipa-icf.c (sem_function::merge): Likewise.
1307 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1308 (do_estimate_edge_time): Likewise.
1309 (estimate_size_after_inlining): Likewise.
1310 (estimate_growth): Likewise.
1311 (growth_likely_positive): Likewise.
1312 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1313 (inline_call): Likewise.
1314 * ipa-inline.c (caller_growth_limits): Likewise.
1315 (can_inline_edge_p): Likewise.
1316 (can_inline_edge_by_limits_p): Likewise.
1317 (compute_uninlined_call_time): Likewise.
1318 (compute_inlined_call_time): Likewise.
1319 (want_inline_small_function_p): Likewise.
1320 (edge_badness): Likewise.
1321 (update_caller_keys): Likewise.
1322 (update_callee_keys): Likewise.
1323 (recursive_inlining): Likewise.
1324 (inline_small_functions): Likewise.
1325 (inline_to_all_callers_1): Likewise.
1326 (dump_overall_stats): Likewise.
1327 (early_inline_small_functions): Likewise.
1328 (early_inliner): Likewise.
1329 * ipa-inline.h (estimate_edge_growth): Likewise.
1330 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1331 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1332 * ipa-prop.h (IPA_NODE_REF): Likewise.
1333 (IPA_EDGE_REF): Likewise.
1334 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1335 (propagate_malloc): Likewise.
1336 * ipa-split.c (execute_split_functions): Likewise.
1337 * symbol-summary.h: Rename get to get_create.
1339 (get_create): Likewise.
1340 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1342 2018-06-08 Martin Liska <mliska@suse.cz>
1344 * symbol-summary.h (release): Move definition out of class
1346 (symtab_removal): Likewise.
1347 (symtab_duplication): Likewise.
1349 2018-06-08 Martin Liska <mliska@suse.cz>
1351 * symbol-summary.h (function_summary): Move constructor
1352 implementation out of class declaration.
1353 (release): Likewise.
1354 (symtab_insertion): Likewise.
1355 (symtab_removal): Likewise.
1356 (symtab_duplication): Likewise.
1359 2018-06-08 Martin Liska <mliska@suse.cz>
1361 * Makefile.in: Remove support for MPX (macros, related functions,
1362 fields in cgraph_node, ...).
1363 * builtin-types.def (BT_BND): Likewise.
1364 (BT_FN_BND_CONST_PTR): Likewise.
1365 (BT_FN_CONST_PTR_BND): Likewise.
1366 (BT_FN_VOID_PTR_BND): Likewise.
1367 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1368 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1369 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1370 (expand_builtin_mempcpy_with_bounds): Likewise.
1371 (expand_builtin_memset_with_bounds): Likewise.
1372 (expand_builtin_memset_args): Likewise.
1373 (std_expand_builtin_va_start): Likewise.
1374 (expand_builtin): Likewise.
1375 (expand_builtin_with_bounds): Likewise.
1376 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1377 (DEF_LIB_BUILTIN_CHKP): Likewise.
1378 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1379 (DEF_CHKP_BUILTIN): Likewise.
1380 (BUILT_IN_MEMCPY): Likewise.
1381 (BUILT_IN_MEMMOVE): Likewise.
1382 (BUILT_IN_MEMPCPY): Likewise.
1383 (BUILT_IN_MEMSET): Likewise.
1384 (BUILT_IN_STPCPY): Likewise.
1385 (BUILT_IN_STRCAT): Likewise.
1386 (BUILT_IN_STRCHR): Likewise.
1387 (BUILT_IN_STRCPY): Likewise.
1388 (BUILT_IN_STRLEN): Likewise.
1389 (BUILT_IN_MEMCPY_CHK): Likewise.
1390 (BUILT_IN_MEMMOVE_CHK): Likewise.
1391 (BUILT_IN_MEMPCPY_CHK): Likewise.
1392 (BUILT_IN_MEMSET_CHK): Likewise.
1393 (BUILT_IN_STPCPY_CHK): Likewise.
1394 (BUILT_IN_STRCAT_CHK): Likewise.
1395 (BUILT_IN_STRCPY_CHK): Likewise.
1396 * calls.c (store_bounds): Likewise.
1397 (emit_call_1): Likewise.
1398 (special_function_p): Likewise.
1399 (maybe_warn_nonstring_arg): Likewise.
1400 (initialize_argument_information): Likewise.
1401 (finalize_must_preallocate): Likewise.
1402 (compute_argument_addresses): Likewise.
1403 (expand_call): Likewise.
1404 * cfgexpand.c (expand_call_stmt): Likewise.
1405 (expand_return): Likewise.
1406 (expand_gimple_stmt_1): Likewise.
1407 (pass_expand::execute): Likewise.
1408 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1409 (cgraph_node::remove): Likewise.
1410 (cgraph_node::dump): Likewise.
1411 (cgraph_node::verify_node): Likewise.
1412 * cgraph.h (chkp_function_instrumented_p): Likewise.
1413 (symtab_node::get_alias_target): Likewise.
1414 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1415 (cgraph_local_p): Likewise.
1416 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1417 (cgraph_edge::rebuild_references): Likewise.
1418 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1419 (walk_polymorphic_call_targets): Likewise.
1420 (cgraph_node::expand_thunk): Likewise.
1421 (symbol_table::output_weakrefs): Likewise.
1422 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1423 (ix86_handle_option): Likewise.
1424 * config/i386/constraints.md: Likewise.
1425 * config/i386/i386-builtin-types.def (BND): Likewise.
1429 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1430 (BDESC_FIRST): Likewise.
1432 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1433 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1434 * config/i386/i386.c (enum reg_class): Likewise.
1435 (ix86_target_string): Likewise.
1436 (ix86_option_override_internal): Likewise.
1437 (ix86_conditional_register_usage): Likewise.
1438 (ix86_valid_target_attribute_inner_p): Likewise.
1439 (ix86_set_indirect_branch_type): Likewise.
1440 (ix86_set_current_function): Likewise.
1441 (ix86_function_arg_regno_p): Likewise.
1442 (init_cumulative_args): Likewise.
1443 (ix86_function_arg_advance): Likewise.
1444 (ix86_function_arg): Likewise.
1445 (ix86_pass_by_reference): Likewise.
1446 (ix86_function_value_regno_p): Likewise.
1447 (ix86_function_value_1): Likewise.
1448 (ix86_function_value_bounds): Likewise.
1449 (ix86_return_in_memory): Likewise.
1450 (ix86_setup_incoming_vararg_bounds): Likewise.
1451 (ix86_va_start): Likewise.
1452 (indirect_thunk_need_prefix): Likewise.
1453 (print_reg): Likewise.
1454 (ix86_print_operand): Likewise.
1455 (ix86_expand_call): Likewise.
1456 (ix86_output_function_return): Likewise.
1457 (reg_encoded_number): Likewise.
1458 (BDESC_VERIFYS): Likewise.
1459 (ix86_init_mpx_builtins): Likewise.
1460 (ix86_init_builtins): Likewise.
1461 (ix86_emit_cmove): Likewise.
1462 (ix86_emit_move_max): Likewise.
1463 (ix86_expand_builtin): Likewise.
1464 (ix86_builtin_mpx_function): Likewise.
1465 (ix86_get_arg_address_for_bt): Likewise.
1466 (ix86_load_bounds): Likewise.
1467 (ix86_store_bounds): Likewise.
1468 (ix86_load_returned_bounds): Likewise.
1469 (ix86_store_returned_bounds): Likewise.
1470 (ix86_class_likely_spilled_p): Likewise.
1471 (ix86_hard_regno_mode_ok): Likewise.
1472 (x86_order_regs_for_local_alloc): Likewise.
1473 (ix86_mitigate_rop): Likewise.
1474 (ix86_bnd_prefixed_insn_p): Likewise.
1475 (ix86_mpx_bound_mode): Likewise.
1476 (ix86_make_bounds_constant): Likewise.
1477 (ix86_initialize_bounds): Likewise.
1478 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1479 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1480 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1481 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1482 (TARGET_CHKP_BOUND_MODE): Likewise.
1483 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1484 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1485 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1486 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1487 * config/i386/i386.h (TARGET_MPX): Likewise.
1488 (TARGET_MPX_P): Likewise.
1489 (VALID_BND_REG_MODE): Likewise.
1490 (FIRST_BND_REG): Likewise.
1491 (LAST_BND_REG): Likewise.
1492 (enum reg_class): Likewise.
1493 (BND_REG_P): Likewise.
1494 (BND_REGNO_P): Likewise.
1495 (BNDmode): Likewise.
1496 (ADJUST_INSN_LENGTH): Likewise.
1497 * config/i386/i386.md: Likewise.
1498 * config/i386/i386.opt: Likewise.
1499 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1500 (defined): Likewise.
1501 (LINK_MPX): Likewise.
1502 (MPX_SPEC): Likewise.
1503 (LIBMPX_SPEC): Likewise.
1504 (LIBMPXWRAPPERS_SPEC): Likewise.
1505 (CHKP_SPEC): Likewise.
1506 * config/i386/predicates.md: Likewise.
1507 * dbxout.c (dbxout_type): Likewise.
1508 * doc/extend.texi: Likewise.
1509 * doc/invoke.texi: Likewise.
1510 * doc/md.texi: Likewise.
1511 * doc/tm.texi: Likewise.
1512 * doc/tm.texi.in: Likewise.
1513 * dwarf2out.c (is_base_type): Likewise.
1514 (gen_formal_types_die): Likewise.
1515 (gen_subprogram_die): Likewise.
1516 (gen_type_die_with_usage): Likewise.
1517 (gen_decl_die): Likewise.
1518 (dwarf2out_late_global_decl): Likewise.
1519 * expr.c (expand_assignment): Likewise.
1520 (emit_storent_insn): Likewise.
1521 (store_expr_with_bounds): Likewise.
1522 (store_expr): Likewise.
1523 (expand_expr_real_1): Likewise.
1524 * expr.h (store_expr_with_bounds): Likewise.
1525 * function.c (use_register_for_decl): Likewise.
1526 (struct bounds_parm_data): Likewise.
1527 (assign_parms_augmented_arg_list): Likewise.
1528 (assign_parm_find_entry_rtl): Likewise.
1529 (assign_parm_is_stack_parm): Likewise.
1530 (assign_parm_load_bounds): Likewise.
1531 (assign_bounds): Likewise.
1532 (assign_parms): Likewise.
1533 (expand_function_start): Likewise.
1534 * gcc.c (CHKP_SPEC): Likewise.
1535 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1536 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1537 (wrestrict_dom_walker::check_call): Likewise.
1538 * gimple.c (gimple_build_call_from_tree): Likewise.
1539 * gimple.h (enum gf_mask): Likewise.
1540 (gimple_call_with_bounds_p): Likewise.
1541 (gimple_call_set_with_bounds): Likewise.
1542 * gimplify.c (gimplify_init_constructor): Likewise.
1543 * ipa-cp.c (initialize_node_lattices): Likewise.
1544 (propagate_constants_across_call): Likewise.
1545 (find_more_scalar_values_for_callers_subset): Likewise.
1546 * ipa-hsa.c (process_hsa_functions): Likewise.
1547 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1548 * ipa-icf.c (sem_function::merge): Likewise.
1549 * ipa-inline.c (early_inliner): Likewise.
1550 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1551 (warn_function_cold): Likewise.
1552 (propagate_pure_const): Likewise.
1553 * ipa-ref.h (enum GTY): Likewise.
1554 * ipa-split.c (find_retbnd): Likewise.
1555 (consider_split): Likewise.
1556 (split_function): Likewise.
1557 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1558 * ipa.c (walk_polymorphic_call_targets): Likewise.
1559 (symbol_table::remove_unreachable_nodes): Likewise.
1560 (process_references): Likewise.
1561 (cgraph_build_static_cdtor_1): Likewise.
1562 * lto-cgraph.c (lto_output_node): Likewise.
1563 (output_refs): Likewise.
1564 (compute_ltrans_boundary): Likewise.
1565 (input_overwrite_node): Likewise.
1566 (input_node): Likewise.
1567 (input_cgraph_1): Likewise.
1568 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1569 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1570 (class pass_chkp_instrumentation_passes): Likewise.
1571 (make_pass_chkp_instrumentation_passes): Likewise.
1572 * passes.def: Likewise.
1573 * rtl.h (struct GTY): Likewise.
1574 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1575 * stor-layout.c (layout_type): Likewise.
1576 * symtab.c: Likewise.
1577 * target.def: Likewise.
1578 * targhooks.c (default_chkp_bound_type): Likewise.
1579 (default_chkp_bound_mode): Likewise.
1580 (default_builtin_chkp_function): Likewise.
1581 (default_chkp_function_value_bounds): Likewise.
1582 (default_chkp_make_bounds_constant): Likewise.
1583 (default_chkp_initialize_bounds): Likewise.
1584 * targhooks.h (default_chkp_bound_type): Likewise.
1585 (default_chkp_bound_mode): Likewise.
1586 (default_builtin_chkp_function): Likewise.
1587 (default_chkp_function_value_bounds): Likewise.
1588 (default_chkp_make_bounds_constant): Likewise.
1589 (default_chkp_initialize_bounds): Likewise.
1590 * toplev.c (compile_file): Likewise.
1591 (process_options): Likewise.
1592 * tree-core.h (DEF_BUILTIN): Likewise.
1593 (DEF_BUILTIN_CHKP): Likewise.
1594 * tree-inline.c (declare_return_variable): Likewise.
1595 (remap_gimple_stmt): Likewise.
1596 (copy_bb): Likewise.
1597 (initialize_inlined_parameters): Likewise.
1598 (expand_call_inline): Likewise.
1599 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1600 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1601 (make_pass_ipa_chkp_produce_thunks): Likewise.
1602 (make_pass_chkp): Likewise.
1603 (make_pass_chkp_opt): Likewise.
1604 (make_pass_chkp_instrumentation_passes): Likewise.
1605 * tree-pretty-print.c (dump_generic_node): Likewise.
1606 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1607 * tree-ssa-dce.c (propagate_necessity): Likewise.
1608 (eliminate_unnecessary_stmts): Likewise.
1609 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1610 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1611 * tree-ssa-sccvn.h: Likewise.
1612 * tree-ssa-strlen.c (get_string_length): Likewise.
1613 (valid_builtin_call): Likewise.
1614 (adjust_last_stmt): Likewise.
1615 (handle_builtin_strchr): Likewise.
1616 (handle_builtin_strcpy): Likewise.
1617 (handle_builtin_stxncpy): Likewise.
1618 (handle_builtin_memcpy): Likewise.
1619 (handle_builtin_strcat): Likewise.
1620 (strlen_check_and_optimize_stmt): Likewise.
1621 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1622 * tree-streamer-in.c: Likewise.
1623 * tree-streamer.c (record_common_node): Likewise.
1624 * tree.c (tree_code_size): Likewise.
1625 (wide_int_to_tree_1): Likewise.
1626 (type_contains_placeholder_1): Likewise.
1627 (build_common_tree_nodes): Likewise.
1628 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1629 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1630 (POINTER_BOUNDS_P): Likewise.
1631 (BOUNDED_TYPE_P): Likewise.
1632 (BOUNDED_P): Likewise.
1633 (CALL_WITH_BOUNDS_P): Likewise.
1634 (pointer_bounds_type_node): Likewise.
1635 * value-prof.c (gimple_ic): Likewise.
1636 * var-tracking.c (vt_add_function_parameters): Likewise.
1637 * varasm.c (make_decl_rtl): Likewise.
1638 (assemble_start_function): Likewise.
1639 (output_constant): Likewise.
1640 (maybe_assemble_visibility): Likewise.
1641 * varpool.c (ctor_for_folding): Likewise.
1642 * chkp-builtins.def: Remove.
1643 * ipa-chkp.c: Remove.
1644 * ipa-chkp.h: Remove.
1645 * rtl-chkp.c: Remove.
1646 * rtl-chkp.h: Remove.
1647 * tree-chkp-opt.c: Remove.
1648 * tree-chkp.c: Remove.
1649 * tree-chkp.h: Remove.
1651 2018-06-07 Carl Love <cel@us.ibm.com>
1653 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1654 vextract_fp_from_shortl): Add BE support.
1656 2018-06-07 Paul Koning <ni1d@arrl.net>
1658 * compare-elim.c (try_merge_compare): Don't merge compare if
1659 address contains a side effect.
1660 (try_eliminate_compare): Likewise.
1662 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1664 * config.gcc: Support "tremont".
1665 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1666 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1668 * config/i386/i386.c (m_TREMONT): Define.
1669 (processor_target_table): Add "tremont".
1670 (PTA_TREMONT): Define.
1671 (ix86_lea_outperforms): Add TARGET_TREMONT.
1672 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1673 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1674 and M_INTEL_GOLDMONT_PLUS.
1675 (fold_builtin_cpu): Add "tremont".
1676 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1677 (ix86_option_override_internal): Add "tremont".
1678 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1679 (processor_type): Add PROCESSOR_TREMONT.
1680 * config/i386/x86-tune.def: Add m_TREMONT.
1681 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1683 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1685 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1686 symbol defined for msp430i* devices to be lower case.
1688 gcc/testsuite/gcc.target/msp430/
1689 * msp430i-device-symbol.c: New test.
1690 * msp430f-device-symbol.c: New test.
1691 * msp430.h: New test header file.
1693 2018-06-07 Richard Biener <rguenther@suse.de>
1695 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1696 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1697 Properly wrap signed arithmetic if overflow wraps.
1699 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1701 PR tree-optimization/69615
1702 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1703 of a cast from a same precision integral SSA_NAME in a bb dominated
1704 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1705 cast to utype if rhs2 has already a compatible type.
1707 2018-06-07 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/85935
1710 * graphite-scop-detection.c (find_params_in_bb): Analyze
1711 condition operands with respect to the correct loop. Assert
1712 the analysis doesn't fail.
1714 2018-06-04 Carl Love <cel@us.ibm.com>
1716 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1717 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1718 as it is slightly cheaper.
1719 (first_match_or_eos_index_<mode>):
1720 Calculate index using natural element order.
1721 (first_match_index_<mode>):
1722 Calculate index using natural element order.
1723 (first_match_or_eos_index_<mode>):
1724 Calculate index using natural order.
1725 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1726 for BE and LE modes.
1727 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1728 P9V_BUILTIN_VCLZLSBB_V16QI.
1729 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1732 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1734 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1735 indentation and line wrap for many prototypes. Add missing
1736 @smallexample directives around block of prototypes for vec_xl and
1739 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1741 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1742 track if we pass or return IEEE 128-bit floating point.
1743 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1744 C++ mangling that is compatible with GCC 8.1.
1745 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1746 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1747 floating point types.
1748 (rs6000_function_arg_advance_1): Likewise.
1749 (rs6000_mangle_type): Optionally generate mangled names that match
1750 what GCC 8.1 generated for IEEE 128-bit floating point types.
1751 (rs6000_globalize_decl_name): If we have an external function that
1752 passes or returns IEEE 128-bit types, generate a weak reference
1753 from the mangled name used in GCC 8.1 to the current mangled
1755 (rs6000_init_builtins): Make __ibm128 use the long double type if
1756 long double is IBM extended double. Make __float128 use the long
1757 double type if long double is IEEE 128-bit.
1760 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1761 macro for __ibm128 built-in functions.
1762 (PACK_IF): Add __ibm128 pack/unpack functions.
1763 (UNPACK_IF): Likewise.
1764 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1765 enable long double built-in functions if long double is IEEE
1766 128-bit floating point.
1767 (rs6000_invalid_builtin): Update long double built-in function
1769 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1770 functions, adjust the built-in function to use the long double
1771 built-in function if __ibm128 and long double are the same type.
1772 * doc/extend.texi (PowerPC builtins): Update documention for
1773 __builtin_{,un}pack_longdouble. Add documentation for
1774 __builtin_{,un}pack_ibm128.
1776 2018-06-06 Jim Wilson <jimw@sifive.com>
1778 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1779 (struct machine_function): New field interrupt_mode.
1780 (riscv_handle_type_attribute): New function. Add forward declaration.
1781 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1782 (riscv_expand_epilogue): Check interrupt_mode field.
1783 (riscv_set_current_function): Check interrupt attribute args and
1784 set interrupt_mode field.
1785 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1786 (riscv_sret, riscv_uret): New.
1787 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1788 new arguments to interrupt attribute.
1790 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1793 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1794 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1796 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1798 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1799 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1800 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1801 several redundant entries.
1803 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1805 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1806 type from "rtx" to "rtx_insn *".
1807 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1808 for local "call_insn", removing cast.
1809 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1811 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1813 PR tree-optimization/86066
1814 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1815 for BIT_INSERT_EXPR stores.
1817 2018-06-06 Richard Biener <rguenther@suse.de>
1819 PR tree-optimization/86062
1820 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1821 component refs ontop
1822 of to be offsetted base.
1824 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1826 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1827 to be static and remove check on interrupt attribute name.
1829 gcc/testsuite/gcc.target/msp430/
1830 * function-attributes-4.c: New test.
1831 * static-interrupts.c: New test.
1833 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1835 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1836 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1838 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1841 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1843 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1844 Remove second argument, change how error is called.
1845 (aarch64_layout_arg): Remove second argument from
1846 aarch64_err_no_fpadvsimd call.
1847 (aarch64_init_cumulative_args): Ditto.
1848 (aarch64_gimplify_va_arg_expr): Ditto.
1849 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1851 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1853 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1854 (*simple_return_indirect_internal<mode>): Rename from
1855 simple_return_indirect_internal. Use W mode iterator.
1856 (rstorssp): New expander.
1857 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1858 (clrssbsy): New expander.
1859 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1861 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1863 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1865 (cmse_check_pointed_object): Likewise.
1867 2018-06-05 Martin Liska <mliska@suse.cz>
1869 PR gcov-profile/47618
1870 * doc/invoke.texi: Document how -fprofile-dir format
1873 2018-06-05 Richard Biener <rguenther@suse.de>
1875 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1876 removal pretend DOM info isn't available so we do not update
1877 it and only remove edges, not dominated blocks. Actually free
1878 DOM info in case we removed something. Remove unreachable blocks.
1879 (mfb_keep_latches): Work with either DOM info or marked backedges.
1880 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1881 first. Mark backedges if DOM info isn't available.
1882 (Re-)compute DOM info after cleanup_control_flow_pre.
1884 2018-06-05 Richard Biener <rguenther@suse.de>
1886 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1887 (locus_discrim_hasher::hash): Adjust.
1888 (locus_discrim_hasher::equal): Likewise.
1889 (next_discriminator_for_locus): Work on line directly.
1890 (same_line_p): Pass in expanded locus1 as well.
1891 (assign_discriminators): Avoid redundant location expansions.
1893 2018-06-05 Richard Biener <rguenther@suse.de>
1895 PR tree-optimization/86046
1896 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1897 if required after clearing TREE_ADDRESSABLE.
1899 2018-06-05 Richard Biener <rguenther@suse.de>
1901 PR tree-optimization/86047
1902 * tree-ssa-loop.c (for_each_index): Glob handling of all
1903 decls and constants and really handle all of them.
1905 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1908 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1909 qualifier_void_pointer and qualifier_const_void_pointer.
1910 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1911 (arm_init_builtins): Handle the above.
1912 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1913 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1914 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1917 2018-06-05 Martin Liska <mliska@suse.cz>
1919 * auto-profile.c (read_autofdo_file): Do not use
1920 gcov_ctr_summary struct.
1921 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1922 * coverage.c (struct counts_entry): Likewise.
1923 (read_counts_file): Read just single summary entry.
1924 (get_coverage_counts): Use gcov_summary struct.
1925 * coverage.h (get_coverage_counts): Likewise.
1926 * gcov-dump.c (dump_working_sets): Likewise.
1927 (tag_summary): Dump just single summary.
1928 * gcov-io.c (gcov_write_summary): Write just histogram
1930 (gcov_read_summary): Read just single summary.
1931 (compute_working_sets): Use gcov_summary struct.
1932 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1933 of GCOV_COUNTERS_SUMMABLE.
1934 (GCOV_COUNTERS_SUMMABLE): Remove.
1935 (GCOV_FIRST_VALUE_COUNTER): Replace with
1936 GCOV_COUNTER_V_INTERVAL.
1937 (struct gcov_ctr_summary): Remove.
1938 (struct gcov_summary): Directly use fields of former
1940 (compute_working_sets): Use gcov_summary struct.
1941 * gcov.c (read_count_file): Do not use ctrs fields.
1942 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1944 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1946 * profile.c: Likewise.
1947 * profile.h: Likewise.
1949 2018-06-05 Martin Liska <mliska@suse.cz>
1951 PR gcov-profile/84846
1952 * gcov.c (output_lines): Print working directory only
1953 in intermediate format.
1955 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1957 * config/s390/s390-builtin-types.def: Add void function type.
1958 * config/s390/s390-builtins.def: Use the function type for the
1961 2018-06-04 Jim Wilson <jimw@sifive.com>
1963 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1965 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1966 and maybe_eh_return. Change regno to unsigned int. Use new args to
1967 handle EH_RETURN_DATA_REGNO registers properly.
1968 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1969 (riscv_expand_epilogue): Update comment. Change argument name and
1970 type. Update code to use new name and type. Pass new args to
1971 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1973 * config/riscv/riscv.md (NORMAL_RETURN): New.
1974 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1975 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1976 (eh_return): Call gen_eh_return_internal and emit barrier.
1977 (eh_return_internal): Call riscv_expand_epilogue.
1979 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1981 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1982 bit_insertion field and declare can_be_merged_into method.
1983 (merged_store_group::can_be_merged_into): New method.
1984 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1985 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1986 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1988 2018-06-04 Richard Biener <rguenther@suse.de>
1990 PR tree-optimization/85955
1991 * builtins.c (fold_builtin_sincos): Convert pointers to
1992 destination to appropriate type before dereferencing.
1994 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1996 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1998 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2000 * expr.c (expand_expr_real_1): Force the operand into memory if
2001 its TYPE_MODE is BLKmode and if there is no integer mode for
2002 the number of bits being extracted.
2004 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2008 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2009 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2011 2018-06-04 Richard Biener <rguenther@suse.de>
2013 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2014 (cleanup_tree_cfg_noloop): ... single caller. Do
2015 start_recording_case_labels later.
2017 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2019 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2020 to _IMMINTRIN_H_INCLUDED.
2021 * config/i386/pconfigintrin.h: Ditto.
2022 * config/i386/waitpkgintrin.h: Ditto.
2023 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2024 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2025 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2026 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2027 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2028 waitpkgintrin.h and cldemoteintrin.h.
2030 2018-06-04 Richard Biener <rguenther@suse.de>
2032 PR tree-optimization/86038
2033 * tracer.c (find_best_successor): Check probability for
2034 being initialized, bail out if not.
2036 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2039 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2040 of bits to ignore when comparing architectures.
2042 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2044 PR tree-optimization/69615
2045 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2046 maximum or minimum of the type, try to merge it also as if
2047 range1 is + [-, x - 1] or + [x + 1, -].
2050 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2052 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2054 PR tree-optimization/86034
2055 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2056 the unsigned bitfield type in a bit insertion sequence if it does not
2057 have a larger precision than the bitfield size.
2058 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2060 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2062 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2064 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2066 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2067 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2068 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2069 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2071 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2073 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2074 Disable -fdelete-null-pointer-checks for ELF toolchain.
2076 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2077 Kito Cheng <kito.cheng@gmail.com>
2079 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2080 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2081 (nds32*-*-*): Add glibc and uclibc conditions.
2082 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2083 (TARGET_EXCEPT_UNWIND_INFO): Define.
2084 * config/nds32/elf.h: New file.
2085 * config/nds32/linux.h: New file.
2086 * config/nds32/nds32-elf.opt: New file.
2087 * config/nds32/nds32-linux.opt: New file.
2088 * config/nds32/nds32-fp-as-gp.c
2089 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2090 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2092 (nds32_asm_file_end): Ditto.
2093 (nds32_print_operand): Ditto.
2094 (nds32_insert_attributes): Ditto.
2095 (nds32_init_libfuncs): New function.
2096 (TARGET_HAVE_TLS): Define.
2097 (TARGET_INIT_LIBFUNCS): Define.
2098 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2100 (TARGET_ELF): Apply different mcmodel setting.
2101 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2102 been migrated into elf.h and linux.h files.
2103 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2104 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2105 (mcmodel): The content has been migrated into nds32-elf.opt and
2106 nds32-linux.opt files.
2107 * config/nds32/t-elf: New file.
2108 * config/nds32/t-linux: New file.
2110 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2111 Shiva Chen <shiva0217@gmail.com>
2113 * config/nds32/constants.md (unspec_volatile_element): Add
2114 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2115 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2117 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2118 (make_pass_nds32_fp_as_gp): Declare.
2119 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2121 (nds32_asm_function_end_prologue): Remove unused asm output.
2122 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2123 (nds32_asm_file_start): Output necessary fp_as_gp information.
2124 (nds32_option_override): Adjust register usage.
2125 (nds32_expand_prologue): Consider fp_as_gp situation.
2126 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2127 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2130 (simple_return): Ditto.
2131 (omit_fp_begin): Output special directive for fp_as_gp.
2132 (omit_fp_end): Output special directive for fp_as_gp.
2133 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2134 mforbid-fp-as-gp): New options.
2136 2018-06-01 Mark Wielaard <mark@klomp.org>
2138 * dwarf2out.c (dwarf2out_finish): Remove generation of
2139 DW_AT_loclists_base.
2141 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2143 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2144 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2145 (struct merged_store_group): Add bit_insertion field.
2146 (dump_char_array): Use standard hexadecimal format.
2147 (merged_store_group::merged_store_group): Set bit_insertion to false.
2148 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2149 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2150 also print the mask in the dump file.
2151 (pass_store_merging::gate): Minor tweak.
2152 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2153 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2154 stores with INTEGER_CST stores.
2155 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2156 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2157 and use it throughout. Generate bit insertion sequences if need be.
2158 (pass_store_merging::process_store): Remove redundant condition.
2159 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2161 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2163 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2164 the 128-bit floating point types. Fix function comment.
2166 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2168 * config/aarch64/aarch64-simd.md
2169 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2171 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2174 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2176 PR tree-optimization/85989
2177 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2179 (backprop::intersect_uses): Check it when deciding whether this
2180 is a backedge reference.
2181 (backprop::process_block): Add each phi to m_visited_phis
2182 after visiting it, then clear it at the end.
2184 2018-06-01 Richard Biener <rguenther@suse.de>
2186 * tree-vectorizer.h (vect_dr_stmt): New function.
2187 (vect_get_load_cost): Adjust.
2188 (vect_get_store_cost): Likewise.
2189 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2190 Use vect_dr_stmt instead of DR_SMTT.
2191 (vect_record_base_alignments): Likewise.
2192 (vect_calculate_target_alignment): Likewise.
2193 (vect_compute_data_ref_alignment): Likewise and make static.
2194 (vect_update_misalignment_for_peel): Likewise.
2195 (vect_verify_datarefs_alignment): Likewise.
2196 (vector_alignment_reachable_p): Likewise.
2197 (vect_get_data_access_cost): Likewise. Pass down
2198 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2199 (vect_get_peeling_costs_all_drs): Likewise.
2200 (vect_peeling_hash_get_lowest_cost): Likewise.
2201 (vect_enhance_data_refs_alignment): Likewise.
2202 (vect_find_same_alignment_drs): Likewise.
2203 (vect_analyze_data_refs_alignment): Likewise.
2204 (vect_analyze_group_access_1): Likewise.
2205 (vect_analyze_group_access): Likewise.
2206 (vect_analyze_data_ref_access): Likewise.
2207 (vect_analyze_data_ref_accesses): Likewise.
2208 (vect_vfa_segment_size): Likewise.
2209 (vect_small_gap_p): Likewise.
2210 (vectorizable_with_step_bound_p): Likewise.
2211 (vect_prune_runtime_alias_test_list): Likewise.
2212 (vect_analyze_data_refs): Likewise.
2213 (vect_supportable_dr_alignment): Likewise.
2214 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2215 (vect_gen_prolog_loop_niters): Likewise.
2216 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2217 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2219 (vect_recog_mask_conversion_pattern): Likewise.
2220 (vect_try_gather_scatter_pattern): Likewise.
2221 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2222 to vect_get_store_cost.
2223 (vect_get_store_cost): Get stmt_info instead of DR.
2224 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2225 (vect_get_load_cost): Get stmt_info instead of DR.
2227 2018-06-01 Richard Biener <rguenther@suse.de>
2230 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2231 vars wrapped in MEM_REFs.
2233 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2235 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2236 Fix subreg tests so that we only return a choice between
2237 GENERAL_REGS and FP_REGS if the original classes included both.
2239 2018-06-01 Richard Biener <rguenther@suse.de>
2242 * tree-ssa-structalias.c (get_function_part_constraint):
2243 Handle NULL fi->decl.
2244 (find_func_aliases_for_call): Properly handle indirect
2245 fi from direct call.
2246 (find_func_clobbers): Likewise.
2247 (ipa_pta_execute): Likewise.
2248 (create_variable_info_for): For functions that are ifunc_resolver
2249 resolve to a varinfo that contains the result of the resolver call.
2250 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2253 2018-05-31 Michael Collison <michael.collison@arm.com>
2255 * config/aarch64/aarch64.md:
2256 (*fix_to_zero_extenddfdi2): New pattern.
2257 * gcc.target/aarch64/fix_extend1.c: New testcase.
2259 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2263 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2264 and BUILT_IN_STRNCMP_EQ.
2265 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2266 BUILT_IN_STRNCMP_EQ.
2267 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2268 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2269 (gimple_fold_builtin): Likewise.
2270 * tree-ssa-strlen.c (compute_string_length): New function.
2271 (determine_min_obsize): New function.
2272 (handle_builtin_string_cmp): New function to handle calls to
2273 string compare functions.
2274 (strlen_optimize_stmt): Add handling to builtin string compare
2276 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2277 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2278 * tree.c (build_common_builtin_nodes): Add new defines of
2279 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2281 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2284 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2285 functions with naked attribute.
2287 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2289 * config/i386/sse.md (avx_vec_concat<mode>):
2290 Substitute concat_tg_mode mode attribute with xtg_mode.
2291 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2292 (concat_tg_mode): Remove mode attribute.
2294 2018-05-31 Martin Sebor <msebor@redhat.com>
2297 * calls.c (alloc_max_size): Correct a logic error/typo.
2298 Treat excessive arguments as infinite. Warn for invalid arguments.
2299 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2301 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2304 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2305 and movx for Haswell.
2307 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2308 Cesar Philippidis <cesar@codesourcery.com>
2311 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2312 when emitting error on private/firstprivate reductions.
2313 * omp-low.c (lower_omp_target): Avoid reference-type processing
2314 on pointers for firstprivate clause.
2316 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2318 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2321 * config/aarch64/aarch64-simd.md
2322 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2323 (aarch64_ld1_x3_<mode>): Likewise
2324 (aarch64_st1x2<VALLDIF:mode>): Likewise
2325 (aarch64_st1_x2_<mode>): Likewise
2326 (aarch64_st1x3<VALLDIF:mode>): Likewise
2327 (aarch64_st1_x3_<mode>): Likewise
2328 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2329 (vld1_s8_x3): Likewise.
2330 (vld1_u16_x3): Likewise.
2331 (vld1_s16_x3): Likewise.
2332 (vld1_u32_x3): Likewise.
2333 (vld1_s32_x3): Likewise.
2334 (vld1_u64_x3): Likewise.
2335 (vld1_s64_x3): Likewise.
2336 (vld1_f16_x3): Likewise.
2337 (vld1_f32_x3): Likewise.
2338 (vld1_f64_x3): Likewise.
2339 (vld1_p8_x3): Likewise.
2340 (vld1_p16_x3): Likewise.
2341 (vld1_p64_x3): Likewise.
2342 (vld1q_u8_x3): Likewise.
2343 (vld1q_s8_x3): Likewise.
2344 (vld1q_u16_x3): Likewise.
2345 (vld1q_s16_x3): Likewise.
2346 (vld1q_u32_x3): Likewise.
2347 (vld1q_s32_x3): Likewise.
2348 (vld1q_u64_x3): Likewise.
2349 (vld1q_s64_x3): Likewise.
2350 (vld1q_f16_x3): Likewise.
2351 (vld1q_f32_x3): Likewise.
2352 (vld1q_f64_x3): Likewise.
2353 (vld1q_p8_x3): Likewise.
2354 (vld1q_p16_x3): Likewise.
2355 (vld1q_p64_x3): Likewise.
2356 (vst1_s64_x2): Likewise.
2357 (vst1_u64_x2): Likewise.
2358 (vst1_f64_x2): Likewise.
2359 (vst1_s8_x2): Likewise.
2360 (vst1_p8_x2): Likewise.
2361 (vst1_s16_x2): Likewise.
2362 (vst1_p16_x2): Likewise.
2363 (vst1_s32_x2): Likewise.
2364 (vst1_u8_x2): Likewise.
2365 (vst1_u16_x2): Likewise.
2366 (vst1_u32_x2): Likewise.
2367 (vst1_f16_x2): Likewise.
2368 (vst1_f32_x2): Likewise.
2369 (vst1_p64_x2): Likewise.
2370 (vst1q_s8_x2): Likewise.
2371 (vst1q_p8_x2): Likewise.
2372 (vst1q_s16_x2): Likewise.
2373 (vst1q_p16_x2): Likewise.
2374 (vst1q_s32_x2): Likewise.
2375 (vst1q_s64_x2): Likewise.
2376 (vst1q_u8_x2): Likewise.
2377 (vst1q_u16_x2): Likewise.
2378 (vst1q_u32_x2): Likewise.
2379 (vst1q_u64_x2): Likewise.
2380 (vst1q_f16_x2): Likewise.
2381 (vst1q_f32_x2): Likewise.
2382 (vst1q_f64_x2): Likewise.
2383 (vst1q_p64_x2): Likewise.
2384 (vst1_s64_x3): Likewise.
2385 (vst1_u64_x3): Likewise.
2386 (vst1_f64_x3): Likewise.
2387 (vst1_s8_x3): Likewise.
2388 (vst1_p8_x3): Likewise.
2389 (vst1_s16_x3): Likewise.
2390 (vst1_p16_x3): Likewise.
2391 (vst1_s32_x3): Likewise.
2392 (vst1_u8_x3): Likewise.
2393 (vst1_u16_x3): Likewise.
2394 (vst1_u32_x3): Likewise.
2395 (vst1_f16_x3): Likewise.
2396 (vst1_f32_x3): Likewise.
2397 (vst1_p64_x3): Likewise.
2398 (vst1q_s8_x3): Likewise.
2399 (vst1q_p8_x3): Likewise.
2400 (vst1q_s16_x3): Likewise.
2401 (vst1q_p16_x3): Likewise.
2402 (vst1q_s32_x3): Likewise.
2403 (vst1q_s64_x3): Likewise.
2404 (vst1q_u8_x3): Likewise.
2405 (vst1q_u16_x3): Likewise.
2406 (vst1q_u32_x3): Likewise.
2407 (vst1q_u64_x3): Likewise.
2408 (vst1q_f16_x3): Likewise.
2409 (vst1q_f32_x3): Likewise.
2410 (vst1q_f64_x3): Likewise.
2411 (vst1q_p64_x3): Likewise.
2413 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2415 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2416 user_label_prefix to name.
2418 * tree-core.h: Update comment about the format of NAME string
2419 passed to handler in attribute_spec.
2421 * config/msp430/msp430.md: Remove erroneous subreg expression from
2422 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2423 zero_extend{q,h}isi2.
2425 2018-05-30 Borislav Petkov <bp@suse.de>
2427 * doc/extend.texi: Document some architecture specific
2428 constraints and sort entries.
2430 2018-05-30 Martin Sebor <msebor@redhat.com>
2433 * builtins.c (expand_builtin_stpcpy_1): New function.
2434 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2435 only if the former succeeds.
2437 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2439 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2442 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2444 * doc/invoke.texi (-flinker-output): Document
2446 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2448 * passes.c (ipa_write_summaries): Only modify statements if body
2450 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2451 incrementally linking.
2452 (ipa_passes): Likewise.
2453 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2454 pass down resolution info.
2455 * common.opt (flag_incremental_link): Update info.
2456 * gcc.c (plugin specs): Turn flinker-output=* to
2457 -plugin-opt=-linker-output-known
2458 * toplev.c (compile_file): Also cut compilation when doing incremental
2460 * flag-types. (enum lto_partition_model): Add
2461 LTO_LINKER_OUTPUT_NOLTOREL.
2462 (invoke.texi): Add -flinker-output docs.
2463 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2464 link same way as WPA; do not stream in dead initializers.
2466 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2467 darf2out_register_external_decl): Support incremental link.
2469 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2471 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2473 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2475 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2476 it down to simple_object_copy_lto_debug_sections.
2477 (run_gcc): Determine incremental LTO link time and configure
2478 lto1 into non-wpa mode, disable renaming of debug sections.
2480 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2482 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2483 descriptions of various incorrectly documented functions.
2485 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2488 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2489 address check not strict.
2491 2018-05-30 Richard Biener <rguenther@suse.de>
2493 PR tree-optimization/85964
2494 * tracer.c (better_p): Drop initialized count check, we only
2495 call the function with initialized counts now.
2496 (find_best_successor): Do find a best edge if one
2497 has uninitialized count.
2498 (find_best_predecessor): Likewise. Do BB frequency check only
2499 if count is initialized.
2501 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2503 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2504 (aarch64_ldrstr_offset_compare): New.
2505 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2506 load/store orderings.
2507 (aarch64_gen_adjusted_ldpstp): Likewise.
2509 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2511 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2512 Check for subset of GENERAL_REGS and FP_REGS.
2513 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2516 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2518 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2519 and wi::to_poly_offset. Add the current offset and then check
2520 whether the sum fits, rather than using an unchecked addition of
2521 a checked term. Check for a shwi rather than a uhwi.
2522 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2523 (store_constructor): Use poly_int_tree_p.
2524 (expand_expr_real_1): Likewise.
2525 * function.c (assign_temp): Likewise.
2526 * fold-const.c (const_binop): Use poly_int_tree_p and
2528 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2530 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2531 to_poly_offset for MEM offsets.
2532 * ipa-icf.c (sem_variable::equals): Likewise.
2533 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2534 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2535 wi::to_poly_offset for BIT_FIELD_REF offsets.
2536 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2538 * var-tracking.c (emit_note_insn_var_location): Use
2539 tree_to_poly_uint64.
2541 2018-05-29 Jim Wilson <jimw@sifive.com>
2543 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2545 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2548 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2549 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2551 (sse4_1_round<mode>2): Use nonimmediate_operand
2552 for operand 1 predicate.
2554 2018-05-29 Martin Sebor <msebor@redhat.com>
2555 Richard Biener <rguenther@suse.de>
2558 * calls.c (get_size_range): Call determine_value_range instead
2559 of get_value_range..
2560 * tree-vrp.h (determine_value_range): Declared new function.
2561 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2563 2018-05-29 Richard Biener <rguenther@suse.de>
2565 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2566 sure to use non-pattern stmts for get_earlier_stmt arguments.
2567 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2568 called on pattern stmts.
2569 (get_later_stmt): Likewise.
2571 2018-05-29 Martin Liska <mliska@suse.cz>
2573 PR gcov-profile/85759
2574 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2577 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2579 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2581 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2585 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2586 VEC_PACK_FLOAT_EXPR): New tree codes.
2587 * tree-pretty-print.c (op_code_prio): Handle
2588 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2589 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2590 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2591 * tree-inline.c (estimate_operator_cost): Likewise.
2592 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2593 * fold-const.c (const_binop): Likewise.
2594 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2595 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2596 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2597 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2598 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2599 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2600 * expr.c (expand_expr_real_2): Likewise.
2601 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2602 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2603 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2605 * optabs.c (expand_widen_pattern_expr): For
2606 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2607 sign from result type rather than operand's type.
2608 (expand_binop_directly): For vec_packu_float_optab and
2609 vec_packs_float_optab allow result type to be different from operand's
2611 * optabs-tree.c (optab_for_tree_code): Handle
2612 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2613 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2614 * tree-vect-generic.c (expand_vector_operations_1): Handle
2615 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2616 VEC_PACK_FLOAT_EXPR.
2617 * tree-vect-stmts.c (supportable_widening_operation): Handle
2619 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2620 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2621 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2622 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2623 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2625 (vec_pack<floatprefix>_float_<mode>): New expander.
2626 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2628 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2629 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2630 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2631 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2632 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2634 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2635 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2636 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2637 VEC_PACK_FLOAT_EXPR): Document.
2639 2018-05-29 Richard Biener <rguenther@suse.de>
2641 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2643 (stmt_vec_info_vec): Make pointer.
2644 (init_stmt_vec_info_vec): Remove.
2645 (free_stmt_vec_info_vec): Likewise.
2646 (set_stmt_vec_info_vec): New function.
2647 (free_stmt_vec_infos): Likewise.
2648 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2649 (set_vinfo_for_stmt): Likewise.
2650 (get_earlier_stmt): Likewise.
2651 (get_later_stmt): Likewise.
2652 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2653 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2654 (vec_info::~vec_info): Free stmt_vec_infos.
2655 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2656 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2657 (pass_slp_vectorize::execute): Likewise.
2658 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2659 (free_stmt_vec_info_vec): Likewise.
2660 (set_stmt_vec_info_vec): New function.
2661 (free_stmt_vec_infos): Likewise.
2662 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2663 the global stmt_vec_info_vec.
2664 * tree-parloops.c (gather_scalar_reductions): Use
2665 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2668 2018-05-29 Richard Biener <rguenther@suse.de>
2670 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2672 2018-05-29 Martin Liska <mliska@suse.cz>
2673 David Malcolm <dmalcolm@redhat.com>
2675 * vec.c (test_reverse): New.
2676 (vec_c_tests): Add new test.
2677 * vec.h (vl_ptr>::reverse): New function.
2679 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2681 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2683 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2686 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2688 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2690 2018-05-28 Richard Biener <rguenther@suse.de>
2692 PR tree-optimization/85933
2693 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2694 look at stmts marked as vectorizable.
2696 2018-05-28 Richard Biener <rguenther@suse.de>
2698 PR tree-optimization/85934
2699 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2700 vector boolean check before scalar optimization.
2702 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2704 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2707 2018-05-28 Mark Wielaard <mark@klomp.org>
2709 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2710 if it is an expression containing a minus sign.
2712 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2714 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2716 2018-05-27 Paul Koning <ni1d@arrl.net>
2718 * config/pdp11/pdp11.md (truncsihi2): Remove.
2720 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2721 Chung-Ju Wu <jasonwucj@gmail.com>
2723 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2725 (unaligned_store_dw): Ditto.
2726 * config/nds32/nds32-memory-manipulation.c
2727 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2728 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2729 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2730 (emit_setmem_word_loop): Rename to ...
2731 (emit_setmem_doubleword_loop): ... this.
2732 (nds32_gen_dup_4_byte_to_word_value): New function.
2733 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2734 (nds32_expand_setmem_loop): Refine implementation.
2735 (nds32_expand_setmem_loop_v3m): Ditto.
2736 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2739 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2741 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2743 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2745 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2746 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2747 machine->attr_no_prologue_p.
2748 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2749 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2750 (nds32_expand_epilogue): Consider attr_naked_p.
2751 (nds32_expand_epilogue_v3pop): Likewise.
2752 (nds32_can_use_return_insn): Likewise.
2753 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2754 attr_no_prologue_p fields.
2755 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2757 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2760 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2762 * config/i386/sse.md
2763 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2765 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2767 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2769 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2771 (*<floatsuffix>floatv2div2sf2): Rename to ...
2772 (*float<floatunssuffix>v2div2sf2): ... this.
2773 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2774 (float<floatunssuffix>v2div2sf2_mask): ... this.
2775 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2776 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2777 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2779 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2781 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2783 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2785 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2787 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2789 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2790 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2791 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2792 gen_ufix_truncv8dfv8si2.
2793 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2794 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2795 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2796 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2797 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2798 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2799 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2800 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2802 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2806 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2808 2018-05-25 Jim Wilson <jimw@sifive.com>
2810 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2811 * config/riscv/riscv.c (struct machine_function): Add
2812 interrupt_handler_p and attribute_checked_p fields.
2813 (riscv_attribute_table): Add interrupt.
2814 (riscv_interrupt_type_p): New.
2815 (riscv_save_reg_p): Save extra regs for interrupt handler.
2816 (riscv_use_save_libcall): Return false for interrupt handler.
2817 (riscv_first_stack_step): Add forward declaration.
2818 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2819 for interrupt handler with large frame. Use it for saved reg list.
2820 (riscv_expand_prologue): Move flag_stack_usage_info support to
2821 eliminate duplication.
2822 (riscv_expand_epilogue): Generate mret for interrupt handler.
2823 (riscv_epilogue_uses): New.
2824 (riscv_can_use_return_insn): Return false for interrupt handler.
2825 (riscv_function_ok_for_sibcall): Likewise.
2826 (riscv_set_current_function): Add interrupt handler support.
2827 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2828 * config/riscv/riscv.md (UNSPECV_MRET): New.
2830 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2832 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2834 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2836 PR tree-optimization/85712
2837 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2838 this candidate has already been replaced in-situ by a copy.
2840 2018-05-25 Jason Merrill <jason@redhat.com>
2842 PR c++/80485 - inline function non-zero address.
2843 * symtab.c (nonzero_address): Check DECL_COMDAT.
2845 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2848 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2849 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2850 extension of SImode operation. Use const123_operand predicate.
2851 (*saddsi_1): Remove.
2852 (*saddl_se_1): Ditto.
2854 (*ssubl_se_1): Ditto.
2855 * config/alpha/predicates.md (const123_operand): New predicate.
2856 * config/alpha/constraints.md (P): Use IN_RANGE.
2858 2018-05-25 Richard Biener <rguenther@suse.de>
2860 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2862 (ref_maybe_used_by_stmt_p): Likewise.
2863 (stmt_may_clobber_ref_p): Likewise.
2864 (stmt_may_clobber_ref_p_1): Likewise.
2865 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2867 (ref_maybe_used_by_stmt_p): Likewise.
2868 (stmt_may_clobber_ref_p): Likewise.
2869 (stmt_may_clobber_ref_p_1): Likewise.
2870 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2871 the alias oracle to disambiguate DRs with stmts DR analysis
2873 (vect_analyze_data_refs): Do not give up on not analyzable
2874 DRs for BB vectorization. Remove code truncating the dataref
2877 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2880 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2881 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2882 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2884 2018-05-25 Richard Biener <rguenther@suse.de>
2886 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2887 function, combining stmt data ref gathering and fatal analysis
2889 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2890 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2891 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2892 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2893 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2895 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2897 PR tree-optimization/85720
2898 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2899 SCC if all partitions are builtins.
2900 (version_loop_by_alias_check): New parameter. Generate cancelable
2901 runtime alias check if all partitions are builtins.
2902 (distribute_loop): Update call to above function.
2904 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2906 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2907 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2908 (parm_default_def_partition_arg): Ditto.
2909 (set_parm_default_def_partition): Ditto.
2910 (get_parm_default_def_partitions): Ditto and make it static.
2911 (get_undefined_value_partitions): Ditto and make it static.
2912 (remove_ssa_form): Refactor call to init_var_map here.
2913 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2914 computation for loop region.
2915 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2916 (register_default_def): Delete.
2917 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2918 (parm_default_def_partition_arg): Ditto.
2919 (set_parm_default_def_partition): Ditto.
2920 (get_parm_default_def_partitions): Ditto and make it static.
2921 (get_undefined_value_partitions): Ditto and make it static.
2922 (coalesce_with_default, coalesce_with_default): Update comment.
2923 (create_coalesce_list_for_region): New func factored out from
2924 create_outofssa_var_map.
2925 (populate_coalesce_list_for_outofssa): New func factored out from
2926 create_outofssa_var_map and coalesce_ssa_name.
2927 (create_outofssa_var_map): Delete.
2928 (coalesce_ssa_name): Refactor to support live range computation.
2929 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2930 (get_parm_default_def_partitions): Delete.
2931 (get_undefined_value_partitions): Ditto.
2932 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2933 computation for loop region.
2934 (new_tree_live_info, loe_visit_block): Ditto.
2935 (live_worklist, set_var_live_on_entry): Ditto.
2936 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2937 * tree-ssa-live.h (struct _var_map): New fields.
2938 (init_var_map): Change decl.
2939 (region_contains_p): New.
2941 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2943 * tree-ssa-live.h (live_merge_and_clear): Delete.
2945 2018-05-25 Richard Biener <rguenther@suse.de>
2948 * tree-dump.c (dequeue_and_dump): Remove access to removed
2949 operand 2 of a SWITCH_EXPR.
2951 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2953 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2954 multiplication and division.
2955 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2956 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2957 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2958 (cond_udiv_optab, cond_umod_optab): New optabs.
2959 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2960 (IFN_COND_RDIV): New internal functions.
2961 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2962 TRUNC_MOD_EXPR and RDIV_EXPR.
2963 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2964 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2966 (SVE_INT_BINARY): Include mult.
2967 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2968 (optab, sve_int_op): Handle mult.
2969 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2971 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2972 for SVE_INT_BINARY_SD.
2974 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2976 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2977 (optab, sve_int_op): Handle div and udiv.
2978 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2979 for SVE_INT_BINARY_SD.
2980 (*<optab><mode>3): New insn for the same.
2982 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2984 * tree-vect-patterns.c: Include predict.h.
2985 (vect_recog_divmod_pattern): Restrict check for division support
2986 to when optimizing for size.
2988 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2990 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2991 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2992 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2993 (gimple_match_op::set_op): Likewise.
2994 (gimple_resimplify4): Declare.
2995 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2996 (expr::gen_transform): Likewise.
2997 (decision_tree::gen): Generate a simplification routine for 4 operands.
2998 * gimple-match-head.c (gimple_simplify): Add an overload for
2999 4 operands. In the top-level function, handle up to 4 call
3000 arguments and call gimple_resimplify4.
3001 (gimple_resimplify4): New function.
3002 (build_call_internal): Pass a fourth operand.
3003 (maybe_push_to_seq): Likewise.
3004 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3005 Fold VEC_COND_EXPRs of an operation and a default value into
3006 an IFN_COND_* function if possible.
3007 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3009 (SVE_COND_FP_BINARY): Include them.
3010 (optab, sve_fp_op): Handle them.
3011 (SVE_INT_BINARY_REV): New code iterator.
3012 (SVE_COND_FP_BINARY_REV): New int iterator.
3013 (commutative): New int attribute.
3014 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3016 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3018 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3019 (*cond_<optab><mode>): New patterns for reversed operands.
3021 2018-05-25 Richard Biener <rguenther@suse.de>
3023 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3024 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3025 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3026 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3027 * tree-vect-data-refs.c (everywhere): Adjust users.
3028 * tree-vect-loop.c (everywhere): Likewise.
3029 * tree-vect-slp.c (everywhere): Likewise.
3030 * tree-vect-stmts.c (everywhere): Likewise.
3031 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3033 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3035 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3037 (gcc_cv_as_section_exclude): ... this.
3038 Try Solaris as #exclude syntax.
3039 * configure: Regenerate.
3040 * config.in: Regenerate.
3041 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3043 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3044 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3046 * varasm.c (default_elf_asm_named_section): Don't check if
3047 HAVE_GAS_SECTION_EXCLUDE is defined.
3049 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3051 * doc/md.texi: Update the documentation of the cond_* optabs
3052 to mention the new final operand. Fix GET_MODE_NUNITS call.
3053 Describe the scalar case too.
3054 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3055 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3057 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3058 (get_conditional_internal_fn): Update comment.
3059 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3060 accumulator value as a final argument to conditional functions.
3061 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3062 a define_expand and add an "else" operand. Assert for now that
3063 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3064 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3065 (*cond_<optab><mode>): New patterns.
3066 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3067 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3068 (UNSPEC_COND_EOR): Delete.
3069 (optab): Remove associated mappings.
3070 (SVE_INT_BINARY): New code iterator.
3071 (sve_int_op): Remove int attribute and add "minus" to the code
3073 (SVE_COND_INT_OP): Delete.
3074 (SVE_COND_FP_OP): Rename to...
3075 (SVE_COND_FP_BINARY): ...this.
3077 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3079 * optabs.c (can_reuse_operands_p): New function.
3080 (maybe_legitimize_operands): Try to reuse the results for
3083 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3085 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3086 Add {q} suffix to insn mnemonic.
3088 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3090 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3091 (msp430_warn_func_return): New.
3093 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3095 * fold-const.c (tree_nonzero_bits): New function.
3096 * fold-const.h (tree_nonzero_bits): Likewise.
3097 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3098 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3100 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3104 * varasm.c (assemble_alias): Check ifunc_resolver only on
3107 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3110 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3111 when memory input operand is handled.
3113 2018-05-24 Luis Machado <luis.machado@linaro.org>
3115 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3117 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3119 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3121 * match.pd: Delay FMA folds until after vectorization.
3123 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3126 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3127 address check not strict.
3129 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3131 * gimple-match.h (gimple_match_op): New class.
3132 (mprts_hook): Replace parameters with a gimple_match_op *.
3133 (maybe_build_generic_op): Likewise.
3134 (gimple_simplified_result_is_gimple_val): Replace parameters with
3135 a const gimple_match_op *.
3136 (gimple_simplify): Replace code_helper * and tree * parameters with
3137 a gimple_match_op * parameter.
3138 (gimple_resimplify1): Replace code_helper *, tree and tree *
3139 parameters with a gimple_match_op * parameter.
3140 (gimple_resimplify2): Likewise.
3141 (gimple_resimplify3): Likewise.
3142 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3143 parameters with a gimple_match_op * parameter.
3144 * gimple-match-head.c (gimple_simplify): Change prototypes of
3145 auto-generated functions to take a gimple_match_op * instead of
3146 separate code_helper * and tree * parameters. Make the same
3147 change in the top-level overload and update calls to the
3148 gimple_resimplify routines. Update calls to the auto-generated
3149 functions and to maybe_push_res_to_seq in the publicly-facing
3150 operation-specific gimple_simplify overloads.
3151 (gimple_match_op::MAX_NUM_OPS): Define.
3152 (gimple_resimplify1): Replace rcode and ops with a single res_op
3153 parameter. Update call to gimple_simplify.
3154 (gimple_resimplify2): Likewise.
3155 (gimple_resimplify3): Likewise.
3156 (mprts_hook): Replace parameters with a gimple_match_op *.
3157 (maybe_build_generic_op): Likewise.
3158 (build_call_internal): Replace type, nargs and ops with
3159 a gimple_match_op *.
3160 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3161 with a single gimple_match_op *. Update calls to mprts_hook,
3162 build_call_internal and gimple_simplified_result_is_gimple_val.
3163 Factor out code that is common to the tree_code and combined_fn cases.
3164 * genmatch.c (expr::gen_transform): Replace tem_code and
3165 tem_ops with a gimple_match_op called tem_op. Update calls
3166 to the gimple_resimplify functions and maybe_push_res_to_seq.
3167 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3168 res_ops. Update call to the gimple_resimplify functions.
3169 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3170 (decision_tree::gen): Make the functions take a gimple_match_op *
3171 called res_op instead of separate res_code and res_ops parameters.
3172 Update call accordingly.
3173 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3174 and ops with a single res_op parameter. Update calls to
3175 maybe_build_generic_op and maybe_push_res_to_seq.
3176 (fold_stmt_1): Update calls to gimple_simplify and
3177 replace_stmt_with_simplification.
3178 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3179 and gimple_simplified_result_is_gimple_val.
3180 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3182 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3183 with a gimple_match_op *.
3184 (vn_nary_build_or_lookup): Likewise. Update call to
3185 vn_nary_build_or_lookup_1.
3186 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3187 gimple_match_op *. Update calls to the gimple_resimplify routines
3188 and to gimple_simplified_result_is_gimple_val.
3189 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3190 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3191 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3192 (visit_nary_op): Likewise.
3193 (visit_reference_op_load): Likewise.
3195 2018-05-23 Luis Machado <luis.machado@linaro.org>
3197 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3198 modifier for printing the step amount.
3200 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3203 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3206 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3208 * doc/sourcebuild.texi (Endianness): New subsubsection.
3210 2018-05-23 Luis Machado <luis.machado@linaro.org>
3212 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3213 <prefetch_dynamic_strides>: New const bool field.
3214 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3215 prefetch_dynamic_strides.
3216 (exynosm1_prefetch_tune): Likewise.
3217 (thunderxt88_prefetch_tune): Likewise.
3218 (thunderx_prefetch_tune): Likewise.
3219 (thunderx2t99_prefetch_tune): Likewise.
3220 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3222 (aarch64_override_options_internal): Update to set
3223 PARAM_PREFETCH_DYNAMIC_STRIDES.
3224 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3225 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3226 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3227 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3228 prefetch-dynamic-strides setting.
3230 2018-05-23 Luis Machado <luis.machado@linaro.org>
3232 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3233 <minimum_stride>: New const int field.
3234 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3235 minimum_stride field defaulting to -1.
3236 (exynosm1_prefetch_tune): Likewise.
3237 (thunderxt88_prefetch_tune): Likewise.
3238 (thunderx_prefetch_tune): Likewise.
3239 (thunderx2t99_prefetch_tune): Likewise.
3240 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3241 <default_opt_level>: Set to 3.
3242 (aarch64_override_options_internal): Update to set
3243 PARAM_PREFETCH_MINIMUM_STRIDE.
3244 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3245 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3246 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3247 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3248 stride is constant and is below the minimum stride threshold.
3250 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3252 * config/arm/arm-cpus.in (mode26): Delete.
3253 (armv4): Delete mode26 reference.
3254 * config/arm/arm.c (arm_configure_build_target): Delete use of
3257 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3259 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3261 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3262 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3264 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3265 (floatunsdidf2): Ditto.
3267 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3269 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3270 (fixuns_trunc<mode>si2_avx512f): Ditto.
3271 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3272 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3273 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3275 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3277 PR rtl-optimization/79985
3278 * df-scan.c (df_insn_refs_collect): Remove special case for
3279 global registers and asm statements.
3281 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3283 * extend.texi (Global Register Variables): Rewrite the bullet list.
3284 Note that the register is available for allocation. Note that access
3285 via inline asm must use constraints. Add note about async-signal
3286 handlers. Remove paragraph about automagic register selection.
3288 2018-05-23 Richard Biener <rguenther@suse.de>
3290 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3291 of fixed offset from memset VN.
3293 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3295 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3297 (alloc_cand_and_find_basis): Initialize first_interp field.
3298 (slsr_process_mul): Modify first_interp field.
3299 (slsr_process_add): Likewise.
3300 (slsr_process_cast): Modify first_interp field for each new
3302 (slsr_process_copy): Likewise.
3303 (dump_candidate): Dump first_interp field.
3304 (replace_mult_candidate): Process all interpretations, not just
3306 (replace_rhs_if_not_dup): Likewise.
3307 (replace_one_candidate): Likewise.
3309 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3311 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3313 (aarch64_needs_frame_chain): New function.
3314 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3316 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3319 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3320 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3321 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3322 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3323 as true for strict-align.
3324 (aarch64_can_inline_p): Perform checks even when callee has no
3325 attributes to check for strict alignment.
3326 * doc/extend.texi (AArch64 Function Attributes): Document
3328 * doc/invoke.texi: (AArch64 Options): Likewise.
3330 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3332 PR tree-optimization/85853
3333 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3334 the handling of the root of the node to...
3335 (vect_slp_analyze_node_operations_1): ...this new function,
3336 and run the whole thing with the child nodes' def types
3337 set according to their SLP node's def type.
3339 2018-05-23 Richard Biener <rguenther@suse.de>
3342 * tree-data-ref.c (create_runtime_alias_checks): Defer
3343 and ignore overflow warnings.
3345 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3347 PR tree-optimization/85822
3348 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3351 2018-05-23 Richard Biener <rguenther@suse.de>
3353 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3354 memset constants via native_interpret_expr.
3356 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3359 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3361 (cgraph_node::create_alias): Likewise.
3362 (cgraph_node::get_availability): Check ifunc_resolver instead
3363 of looking up ifunc attribute.
3364 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3365 * varasm.c (do_assemble_alias): Likewise.
3366 (assemble_alias): Likewise.
3367 (default_binds_local_p_3): Likewise.
3368 * cgraph.h (cgraph_node): Add ifunc_resolver.
3369 (cgraph_node::only_called_directly_or_aliased_p): Return false
3371 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3373 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3374 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3375 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3376 instead of looking up ifunc attribute.
3378 2018-05-22 Luis Machado <luis.machado@linaro.org>
3380 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3382 2018-05-22 Martin Sebor <msebor@redhat.com>
3385 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3386 only when expasion succeeds.
3387 (expand_builtin_strcmp): Same.
3388 (expand_builtin_strncmp): Same.
3390 2018-05-22 Martin Sebor <msebor@redhat.com>
3392 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3394 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3395 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3397 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3398 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3400 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3401 Add check for legitimate_address.
3402 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3403 (aarch64_swap_ldrstr_operands): New.
3404 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3407 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3408 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3410 * config/aarch64/aarch64.md: New patterns to generate stp
3412 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3413 single words and double words.
3414 (load_pair_sw, load_pair_dw): Likewise.
3415 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3417 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3419 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3420 for different mode ldpstp and add peephole for merged zero stores.
3422 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3424 (aarch64_gen_store_pair): Rename calls to match new patterns.
3425 (aarch64_gen_load_pair): Rename calls to match new patterns.
3426 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3427 (load_pair<DREG:mode><DREG2:mode>): ... This.
3428 (store_pair<mode>): Rename to...
3429 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3430 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3432 (V_INT_EQUIV): Handle SImode.
3433 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3436 2018-05-22 Martin Sebor <msebor@redhat.com>
3439 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3440 or ajust the presumed bound on an operation to avoid unnecessary
3443 2018-05-22 Martin Sebor <msebor@redhat.com>
3445 PR tree-optimization/85826
3446 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3447 assuming that a DECL necesarily has a constant size.
3449 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3452 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3454 2018-05-22 Richard Biener <rguenther@suse.de>
3456 PR tree-optimization/85834
3457 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3458 non-constant and non-zero memset arguments.
3460 2018-05-22 Martin Liska <mliska@suse.cz>
3463 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3465 2018-05-22 Richard Biener <rguenther@suse.de>
3467 PR tree-optimization/85863
3468 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3469 comparisons when vectype is specified.
3470 (vectorizable_condition): Do not specify vectype for
3471 vect_is_simple_cond when SLP vectorizing.
3473 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3476 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3477 define __ibm128 as long double.
3478 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3479 as a distinct type when IEEE 128-bit support is enabled.
3480 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3481 128-bit types to use the correct functions.
3482 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3483 convert between 128-bit floating point types that have different
3484 modes but the same representation, instead of using gen_lowpart to
3486 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3488 (IFKF_reg): New attributes to give the register constraints for
3490 (extend<mode>tf2_internal): New insns to mark an explicit
3491 conversion between 128-bit floating point types that have a
3492 different mode but share the same representation.
3494 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3496 PR tree-optimization/85814
3497 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3498 a null return from get_strinfo when unsharing the next
3499 strinfo in the chain.
3501 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3504 * varasm.c (weak_finish): Clean up weak_decls.
3506 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3508 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3509 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3510 UNSPEC_UADALP values.
3511 * config/aarch64/iterators.md (ABAL): New int iterator.
3514 (sur): Add mappings for the above.
3515 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3517 (aarch64_<sur>abal<mode>_4): Likewise.
3518 (aarch64_<sur>adalp<mode>_3): Likewise.
3519 (<sur>sadv16qi): New define_expand.
3521 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3523 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3524 (*movdf_internal): Ditto.
3525 (*rcpsf2_sse): Ditto.
3526 (*rsqrtsf2_sse): Ditto.
3527 (*sqrt<mode>2_sse): Ditto.
3529 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3531 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3533 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3534 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3535 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3536 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3538 * config/aarch64/arm_neon.h: Likewise.
3539 * config/aarch64/iterators.md (VQ_I): New.
3541 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3543 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3545 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3548 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3550 * config/nds32/constraints.md (S): New constraint.
3551 * config/nds32/nds32.md (call_internal): Use constraint S.
3552 (call_value_internal): Likewise.
3553 (sibcall_internal): Likewise.
3554 (sibcall_value_internal): Likewise.
3556 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3557 Chung-Ju Wu <jasonwucj@gmail.com>
3559 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3562 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3563 Chung-Ju Wu <jasonwucj@gmail.com>
3565 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3566 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3567 (nds32_rtx_costs_impl): Simplify.
3568 (nds32_address_cost_impl): Simplify.
3569 (nds32_init_rtx_costs): New function.
3570 (nds32_rtx_costs_speed_prefer): Likewise.
3571 (nds32_rtx_costs_size_prefer): Likewise.
3572 (nds32_address_cost_speed_prefer): Likewise.
3573 (nds32_address_cost_speed_fwprop): Likewise.
3574 (nds32_address_cost_size_prefer): Likewise.
3575 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3576 * config/nds32/nds32.c (nds32_option_override): Use
3577 nds32_init_rtx_costs function.
3579 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3581 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3582 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3583 (TARGET_PIPELINE_N8): Likewise.
3584 (TARGET_PIPELINE_N10): Likewise.
3585 (TARGET_PIPELINE_N13): Likewise.
3586 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3588 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3590 * config/nds32/nds32-fpu.md: Update copyright year.
3592 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3596 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.c
3599 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3600 * config/nds32/nds32.opt (minline-asm-r15): New option.
3602 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3604 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3606 * config/nds32/nds32.md (abssi2): New pattern.
3608 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3610 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3611 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3612 Merge insn pattern from sse_cvtsi2ss<round_name> and
3613 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3614 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3615 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3616 using SWI48 mode iterator.
3617 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3618 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3619 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3620 pattern from sse_cvttss2si<round_saeonly_name>
3621 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3622 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3623 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3624 using SWI48 mode iterator.
3625 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3626 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3627 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3628 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3629 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3630 using SWI48 mode iterator.
3631 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3632 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3633 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3634 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3635 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3636 SWI48 mode iterator.
3637 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3638 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3639 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3640 pattern from sse_cvttsd2si<round_saeonly_name>
3641 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3643 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3645 * config/nds32/nds32-md-auxiliary.c
3646 (nds32_valid_smw_lwm_base_p): Refine.
3647 (nds32_output_smw_single_word): Refine.
3648 (nds32_output_smw_double_word): New.
3649 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3651 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3653 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3654 (nds32_output_stack_pop): Refine.
3655 (nds32_expand_unaligned_load): Refine.
3656 (nds32_expand_unaligned_store): Refine.
3658 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3659 Chung-Ju Wu <jasonwucj@gmail.com>
3661 * config/nds32/constants.md: Add TP_REGNUM constant.
3662 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3663 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3665 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3666 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3667 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3668 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3669 and PIC code generation.
3670 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3672 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3674 * config/nds32/nds32.md: Support TLS and PIC.
3675 * config/nds32/nds32.c: Support TLS and PIC.
3676 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3677 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3680 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3682 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3683 mode with E_ prefix.
3685 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3686 Chung-Ju Wu <jasonwucj@gmail.com>
3688 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3689 * config/nds32/nds32-md-auxiliary.c
3690 (symbolic_reference_mentioned_p): New.
3691 (nds32_legitimize_ict_address): New.
3692 (nds32_expand_ict_move): New.
3693 (nds32_indirect_call_referenced_p): New.
3694 (nds32_symbol_binds_local_p): Delete.
3695 (nds32_long_call_p): Modify.
3696 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3697 * config/nds32/nds32-protos.h
3698 (symbolic_reference_mentioned_p): Declare.
3699 (nds32_legitimize_ict_address): Declare.
3700 (nds32_expand_ict_move): Declare.
3701 (nds32_indirect_call_referenced_p): Declare.
3702 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3703 (nds32_relax_group): Use nds32_ict_const_p as condition.
3704 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3705 (nds32_asm_file_start): Output ict_model directive in asm code.
3706 (nds32_legitimate_address_p): Consider indirect call.
3707 (nds32_print_operand): Consider indirect call.
3708 (nds32_print_operand_address): Consider indirect call.
3709 (nds32_insert_attributes): Handle "indirect_call" attribute.
3710 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3711 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3712 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3713 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3714 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3715 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3716 (TARGET_ICT_MODEL_SMALL): Define.
3717 (TARGET_ICT_MODEL_LARGE): Define.
3718 * config/nds32/nds32.md (movsi): Consider ict model.
3719 (call, call_value): Consider ict model.
3720 (sibcall, sibcall_value): Consider ict model.
3721 * config/nds32/nds32.opt (mict-model): New option.
3722 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3725 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3726 Monk Chiang <sh.chiang04@gmail.com>
3727 Jim Wilson <jimw@sifive.com>
3729 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3730 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3731 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3732 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3733 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3734 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3735 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3736 compute save_libcall_adjustment properly.
3737 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3738 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3739 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3740 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3741 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3742 (ABI_SPEC): Handle mabi=ilp32e.
3743 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3744 (RVE): Add RVE mask.
3745 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3746 <-march>: Add rv32e as an example.
3748 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3751 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3752 (intra_create_variable_infos): Handle C++ constructors.
3754 2018-05-18 Martin Liska <mliska@suse.cz>
3756 * passes.def: Remove a redundant pass.
3758 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3761 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3763 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3765 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3767 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3769 (armv2, armv2a, armv3, armv3m): Delete architectures.
3770 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3771 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3772 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3774 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3775 (*mulsidi3adddi): Likewise.
3776 (mulsidi3): Likewise.
3777 (*mulsidi3_nov6): Likewise.
3778 (umulsidi3): Likewise.
3779 (umulsidi3_nov6): Likewise.
3780 (umaddsidi4): Likewise.
3781 (*umulsidi3adddi): Likewise.
3782 (smulsi3_highpart): Likewise.
3783 (*smulsi3_highpart_nov6): Likewise.
3784 (umulsi3_highpart): Likewise.
3785 (*umulsi3_highpart_nov6): Likewise.
3786 * config/arm/arm.h (arm_arch3m): Delete.
3787 * config/arm/arm.c (arm_arch3m): Delete.
3788 (arm_option_override_internal): Update armv3-related comment.
3789 (arm_configure_build_target): Delete use of isa_bit_mode32.
3790 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3791 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3792 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3794 (mulusa3): Likewise.
3795 * config/arm/arm-protos.h (arm_arch3m): Delete.
3796 * config/arm/arm-tables.opt: Regenerate.
3797 * config/arm/arm-tune.md: Likewise.
3798 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3799 deleted architectures.
3801 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3803 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3804 (armv5t, armv5te): New features.
3805 (ARMv5, ARMv5e): Delete fgroups.
3806 (ARMv5t, ARMv5te): Adjust for above changes.
3808 (armv5, armv5e): Delete arches.
3809 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3811 (*call_reg_arm): Likewise.
3812 (*call_value_reg_armv5): Likewise.
3813 (*call_value_reg_arm): Likewise.
3814 (*call_symbol): Likewise.
3815 (*call_value_symbol): Likewise.
3816 (*sibcall_insn): Likewise.
3817 (*sibcall_value_insn): Likewise.
3819 (prefetch): Likewise.
3820 (define_split and define_peephole2 dependent on arm_arch5):
3822 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3824 (TARGET_ARM_QBIT): Likewise.
3825 (TARGET_DSP_MULTIPLY): Likewise.
3826 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3827 (arm_arch5, arm_arch5e): Delete.
3828 (arm_arch5t, arm_arch5te): Declare.
3829 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3830 (arm_arch5t): Declare.
3831 (arm_option_reconfigure_globals): Update for the above.
3832 (arm_options_perform_arch_sanity_checks): Update comment, replace
3833 use of arm_arch5 with arm_arch5t.
3834 (use_return_insn): Likewise.
3835 (arm_emit_call_insn): Likewise.
3836 (output_return_instruction): Likewise.
3837 (arm_final_prescan_insn): Likewise.
3838 (arm_coproc_builtin_available): Likewise.
3839 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3840 arm_arch5e with arm_arch5t and arm_arch5te.
3841 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3842 (arm_arch5t, arm_arch5te): Declare.
3843 * config/arm/arm-tables.opt: Regenerate.
3844 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3845 * config/arm/t-multilib: Likewise.
3846 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3847 instead of arm_arch5.
3848 (*call_reg_thumb1): Likewise.
3849 (*call_value_reg_thumb1_v5): Likewise.
3850 (*call_value_reg_thumb1): Likewise.
3851 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3853 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3855 2018-05-18 Martin Liska <mliska@suse.cz>
3857 PR gcov-profile/84846
3858 * doc/gcov.texi: Document -t option of gcov tool.
3860 2018-05-18 Martin Liska <mliska@suse.cz>
3862 PR gcov-profile/84846
3863 * gcov.c (print_usage): Add new -t option.
3864 (process_args): Handle the option.
3865 (generate_results): Use stdout as output when requested by
3868 2018-05-18 Martin Liska <mliska@suse.cz>
3870 PR gcov-profile/84846
3871 * coverage.c (coverage_init): Write PWD to .gcno file.
3872 * doc/gcov.texi: Document how working directory is printed.
3873 * gcov-dump.c (dump_gcov_file): Print PWD.
3874 * gcov.c (output_intermediate_file): Likewise.
3875 (read_graph_file): Read PWD string.
3876 (output_lines): Print PWD.
3878 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3881 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3882 for retval and return false if all args to phi are zero.
3884 2018-05-18 Richard Biener <rguenther@suse.de>
3886 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3888 (evrp_dom_walker::before_dom_children): Call it.
3890 2018-05-18 Richard Biener <rguenther@suse.de>
3892 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3893 results when processing array refs with variable index.
3895 2018-05-18 Toon Moene <toon@moene.org>
3897 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3898 directly after that of -floop-interchange. Indicate that both
3899 options are enabled by default when specifying -O3.
3901 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3903 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3904 iterator. Delete separate integer-mode vec_set<mode> expander.
3905 (aarch64_simd_vec_setv2di): Delete.
3906 (vec_setv2di): Delete.
3907 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3908 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3909 the "w, r" alternative.
3911 2018-05-18 Martin Liska <mliska@suse.cz>
3913 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3914 * tree-pass.h (make_pass_lower_switch_O0): New function.
3915 * tree-switch-conversion.c (node_has_low_bound): Remove.
3916 (node_has_high_bound): Likewise.
3917 (node_is_bounded): Likewise.
3918 (class pass_lower_switch): Make it a template type and create
3920 (pass_lower_switch::execute): Add template argument.
3921 (make_pass_lower_switch): New function.
3922 (make_pass_lower_switch_O0): New function.
3923 (do_jump_if_equal): Remove.
3924 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3925 all the hard work to tree optimization passes.
3927 2018-05-18 Martin Liska <mliska@suse.cz>
3929 * dbgcnt.c (limit_low): Renamed from limit.
3930 (limit_high): New variable.
3931 (dbg_cnt_is_enabled): Check for upper limit.
3932 (dbg_cnt): Adjust dumping.
3933 (dbg_cnt_set_limit_by_index): Add new argument for high
3935 (dbg_cnt_set_limit_by_name): Likewise.
3936 (dbg_cnt_process_single_pair): Parse new format.
3937 (dbg_cnt_process_opt): Use strtok.
3938 (dbg_cnt_list_all_counters): Remove 'value' and add
3940 * doc/invoke.texi: Document changes.
3942 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3944 * doc/sourcebuild.texi (scalar_all_fma): Document.
3945 * tree.def (FMA_EXPR): Delete.
3946 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3947 * internal-fn.c (ternary_direct): New macro.
3948 (expand_ternary_optab_fn): Likewise.
3949 (direct_ternary_optab_supported_p): Likewise.
3950 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3951 * builtins.c (fold_builtin_fma): Delete.
3952 (fold_builtin_3): Don't call it.
3953 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3954 * expr.c (expand_expr_real_2): Likewise.
3955 * fold-const.c (operand_equal_p): Likewise.
3956 (fold_ternary_loc): Likewise.
3957 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3958 * gimple.c (DEFTREECODE): Likewise.
3959 * gimplify.c (gimplify_expr): Likewise.
3960 * optabs-tree.c (optab_for_tree_code): Likewise.
3961 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3962 * tree-eh.c (operation_could_trap_p): Likewise.
3963 (stmt_could_throw_1_p): Likewise.
3964 * tree-inline.c (estimate_operator_cost): Likewise.
3965 * tree-pretty-print.c (dump_generic_node): Likewise.
3966 (op_code_prio): Likewise.
3967 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3968 * tree-ssa-operands.c (get_expr_operands): Likewise.
3969 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3970 * fold-const-call.h (fold_fma): Delete.
3971 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3972 CFN_FNMA and CFN_FNMS.
3974 * genmatch.c (combined_fn): New enum.
3975 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3976 (commutative_op): New function.
3977 (commutate): Use it. Handle more than 2 operands.
3978 (dt_operand::gen_gimple_expr): Use commutative_op.
3979 (parser::parse_expr): Allow :c to be used with non-binary
3980 operators if the commutative operand is known.
3981 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3982 CFN_FMS, CFN_FNMA and CFN_FNMS.
3983 (backprop::process_assign_use): Remove FMA_EXPR handling.
3984 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3985 (gen_hsa_fma): New function.
3986 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3987 IFN_FNMA and IFN_FNMS.
3988 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3989 * gimple-fold.h (follow_all_ssa_edges): Declare.
3990 * gimple-fold.c (follow_all_ssa_edges): New function.
3991 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3992 gimple_build interface and use follow_all_ssa_edges to fold the result.
3993 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3994 instead of checking for optabs directly.
3995 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3996 rather than FMA_EXPRs.
3997 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3998 call to IFN_FMA instead of an FMA_EXPR.
4000 2018-05-17 Jim Wilson <jimw@sifive.com>
4002 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4003 sign extended promoted subreg, and the range does not have the sign bit
4004 set, then do a sign extend.
4006 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4007 test, check for sign extended subreg and/or constant operands, and
4008 do a sign extend in that case.
4010 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4012 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4013 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4015 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4016 Change logics_shift_reg to logics_shift_imm.
4017 (thunderx2t99_fp_loadpair_basic): Delete.
4018 (thunderx2t99_fp_storepair_basic): Delete.
4019 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4020 (thunderx2t99_asimd_polynomial): Delete.
4021 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4022 and neon_fp_mul_d_scalar_q.
4023 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4024 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4025 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4026 (thunderx2t99_asimd_lut): Add missing tbl types.
4027 (thunderx2t99_asimd_ext): Delete.
4028 (thunderx2t99_asimd_load1_1_mult): Delete.
4029 (thunderx2t99_asimd_load1_2_mult): Delete.
4030 (thunderx2t99_asimd_load1_ldp): New.
4031 (thunderx2t99_asimd_load1): New.
4032 (thunderx2t99_asimd_load2): Add missing *load2* types.
4033 (thunderx2t99_asimd_load3): New.
4034 (thunderx2t99_asimd_load4): New.
4035 (thunderx2t99_asimd_store1_1_mult): Delete.
4036 (thunderx2t99_asimd_store1_2_mult): Delete.
4037 (thunderx2t99_asimd_store2_mult): Delete.
4038 (thunderx2t99_asimd_store2_onelane): Delete.
4039 (thunderx2t99_asimd_store_stp): New.
4040 (thunderx2t99_asimd_store1): New.
4041 (thunderx2t99_asimd_store2): New.
4042 (thunderx2t99_asimd_store3): New.
4043 (thunderx2t99_asimd_store4): New.
4045 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4047 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4048 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4050 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4051 Segher Boessenkool <segher@kernel.crashing.org>
4054 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4057 2018-05-17 Richard Biener <rguenther@suse.de>
4059 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4060 for pruning loop and prune defs feeding only already visited PHIs.
4062 2018-05-17 Richard Biener <rguenther@suse.de>
4064 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4066 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4067 Richard Biener <rguenther@suse.de>
4069 PR tree-optimization/85793
4070 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4071 for VMAT_ELEMENTWISE.
4073 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4075 * internal-fn.h (lookup_internal_fn): Declare
4076 * internal-fn.c (lookup_internal_fn): New function.
4077 * gimple.c (gimple_build_call_from_tree): Handle calls to
4079 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4080 internal function names.
4081 * tree-pretty-print.c (dump_generic_node): Likewise.
4082 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4084 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4086 * gimple-fold.h (gimple_build): Make the function forms take
4087 combined_fn rather than built_in_function.
4088 (gimple_simplify): Likewise.
4089 * gimple-match-head.c (gimple_simplify): Likewise.
4090 * gimple-fold.c (gimple_build): Likewise.
4091 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4092 rather than gimple_build_call_internal.
4093 (get_initial_defs_for_reduction): Likewise.
4094 (vect_create_epilog_for_reduction): Likewise.
4095 (vectorizable_live_operation): Likewise.
4097 2018-05-17 Martin Liska <mliska@suse.cz>
4099 * gimple-ssa-sprintf.c (format_directive): Do not use
4100 space in between 'G_' and '('.
4102 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4105 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4106 even if the mask is not all ones.
4109 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4111 (ix86_gimple_fold_builtin): Likewise.
4114 * config/i386/i386.c: Include tree-vector-builder.h.
4115 (ix86_vector_shift_count): New function.
4116 (ix86_fold_builtin): Fold shift builtins by scalar count.
4117 (ix86_gimple_fold_builtin): Likewise.
4119 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4120 _mm512_setzero): New intrinsics.
4122 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4123 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4125 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4126 code generation for cases where splatting a value is not useful.
4127 * simplify-rtx.c (simplify_ternary_operation): Simplify
4128 vec_merge across a vec_duplicate and a paradoxical subreg forming
4129 a vector mode to a vec_concat.
4131 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4133 * config.gcc: Support "goldmont-plus".
4134 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4136 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4137 PROCESSOR_GOLDMONT_PLUS.
4138 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4139 (processor_target_table): Add "goldmont-plus".
4140 (PTA_GOLDMONT_PLUS): Define.
4141 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4142 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4143 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4144 (fold_builtin_cpu): Add "goldmont-plus".
4145 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4146 (ix86_option_override_internal): Add "goldmont-plus".
4147 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4148 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4149 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4150 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4152 2018-05-17 Richard Biener <rguenther@suse.de>
4154 PR tree-optimization/85757
4155 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4156 remove defs that only feed that PHI from further processing.
4158 2018-05-16 Jim Wilson <jimw@sifive.com>
4160 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4162 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4164 2018-05-16 Mark Wielaard <mark@klomp.org>
4166 * dwarf2out.c (count_index_strings): New function.
4167 (output_indirect_strings): Call count_index_strings and generate
4168 header for dwarf_version >= 5.
4170 2018-05-16 Mark Wielaard <mark@klomp.org>
4172 * dwarf2out.c (dwarf_FORM): New function.
4173 (set_indirect_string): Use dwarf_FORM.
4174 (reset_indirect_string): Likewise.
4175 (size_of_die): Likewise.
4176 (value_format): Likewise.
4177 (output_die): Likewise.
4178 (add_skeleton_AT_string): Likewise.
4179 (output_macinfo_op): Likewise.
4180 (index_string): Likewise.
4181 (output_index_string_offset): Likewise.
4182 (output_index_string): Likewise.
4183 (count_index_strings): Likewise.
4185 2018-05-16 Carl Love <cel@us.ibm.com>
4187 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4188 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4190 2018-05-16 Martin Jambor <mjambor@suse.cz>
4192 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4193 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4195 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4197 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4198 (fnma<mode>4): Likewise.
4199 (fms<mode>4): Likewise.
4200 (fnms<mode>4): Likewise.
4201 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4202 (aarch64_fnma<mode>4): Likewise.
4203 (aarch64_fms<mode>4): Likewise.
4204 (aarch64_fnms<mode>4): Likewise.
4205 (aarch64_fnmadd<mode>4): Likewise.
4207 2018-05-16 Jason Merrill <jason@redhat.com>
4209 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4211 2018-05-16 Richard Biener <rguenther@suse.de>
4213 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4214 (dump_stmt_cost): Declare.
4215 (add_stmt_cost): Dump cost we add.
4216 (add_stmt_costs): New function.
4217 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4219 (vect_analyze_stmt): Adjust prototype.
4220 (vectorizable_condition): Likewise.
4221 (vectorizable_live_operation): Likewise.
4222 (vectorizable_reduction): Likewise.
4223 (vectorizable_induction): Likewise.
4224 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4225 cost vector to pass to vectorizable_ and record afterwards.
4226 (vect_model_reduction_cost): Take cost vector argument and adjust.
4227 (vect_model_induction_cost): Likewise.
4228 (vectorizable_reduction): Likewise.
4229 (vectorizable_induction): Likewise.
4230 (vectorizable_live_operation): Likewise.
4231 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4232 SLP_TREE_NUMBER_OF_VEC_STMTS.
4233 (vect_analyze_slp_cost_1): Remove.
4234 (vect_analyze_slp_cost): Likewise.
4235 (vect_slp_analyze_node_operations): Take visited args and
4236 a target cost vector. Avoid processing already visited stmt sets.
4237 (vect_slp_analyze_operations): Use a local cost vector to gather
4238 costs and register those of non-discarded instances.
4239 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4240 (vect_schedule_slp_instance): Remove copying of
4241 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4243 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4244 adding cost. Record cost entry location.
4245 (vect_prologue_cost_for_slp_op): Function to compute cost of
4246 a constant or invariant generated for SLP vect in the prologue,
4247 split out from vect_analyze_slp_cost_1.
4248 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4249 (vect_model_promotion_demotion_cost): Likewise.
4250 (vect_model_store_cost): Likewise, make static.
4251 (vect_model_load_cost): Likewise.
4252 (vectorizable_bswap): Add cost vector arg and adjust.
4253 (vectorizable_call): Likewise.
4254 (vectorizable_simd_clone_call): Likewise.
4255 (vectorizable_conversion): Likewise.
4256 (vectorizable_assignment): Likewise.
4257 (vectorizable_shift): Likewise.
4258 (vectorizable_operation): Likewise.
4259 (vectorizable_store): Likewise.
4260 (vectorizable_load): Likewise.
4261 (vectorizable_condition): Likewise.
4262 (vectorizable_comparison): Likewise.
4263 (can_vectorize_live_stmts): Likewise.
4264 (vect_analyze_stmt): Likewise.
4265 (vect_transform_stmt): Adjust calls to vectorizable_*.
4266 * tree-vectorizer.c: Include gimple-pretty-print.h.
4267 (dump_stmt_cost): New function.
4269 2018-05-16 Richard Biener <rguenther@suse.de>
4271 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4272 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4273 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4274 (check_name): New callback.
4275 (dse_classify_store): Track cycles via a visited bitmap of PHI
4276 defs and simplify handling of in-loop and across loop dead stores
4277 and properly fail for loop-variant refs. Handle byte-tracking with
4278 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4281 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4283 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4284 (vect_get_mask_type_for_stmt): Likewise.
4285 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4287 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4288 to determine the statement's vector type and the vector type that
4289 should be used for calculating nunits. Deal with cases in which
4290 the type has to be deferred.
4291 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4292 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4293 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4294 (vect_determine_vf_for_stmt): New functions, split out from...
4295 (vect_determine_vectorization_factor): ...here.
4296 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4297 (vect_get_mask_type_for_stmt): New functions, split out from
4298 vect_determine_vectorization_factor.
4300 2018-05-16 Richard Biener <rguenther@suse.de>
4302 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4303 verify the [VEC_]COND_EXPR embedded comparison.
4305 2018-05-15 Martin Sebor <msebor@redhat.com>
4307 PR tree-optimization/85753
4308 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4309 RECORD_TYPE in addition to ARRAY_TYPE.
4311 2018-05-15 Martin Sebor <msebor@redhat.com>
4314 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4316 2018-05-15 Richard Biener <rguenther@suse.de>
4318 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4319 add by_clobber_p one. Change algorithm to collect all defs
4320 representing uses we need to walk and try reducing them to
4321 a single one before failing.
4322 (dse_dom_walker::dse_optimize_stmt): Adjust.
4324 2018-05-13 Mark Wielaard <mark@klomp.org>
4326 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4327 (size_of_loc_descr): Likewise.
4328 (output_loc_operands): Likewise.
4329 (output_loc_operands_raw): Likewise.
4330 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4331 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4332 (hash_loc_operands): Likewise.
4333 (compare_loc_operands): Likewise.
4335 2018-05-14 Mark Wielaard <mark@klomp.org>
4337 * dwarf2out.c (count_index_addrs): New function.
4338 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4340 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4342 PR tree-optimization/83648
4343 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4344 return value as malloc candidate.
4346 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4349 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4350 param as true in call to suggest_attribute.
4352 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4354 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4355 -mreadonly-in-sdata.
4357 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4359 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4361 (aarch64_crypto_aesd_fused): Likewise.
4363 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4365 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4366 (movsi_aarch64): Likewise.
4367 (load_pairsi): Likewise.
4368 (load_pairdi): Likewise.
4369 (store_pairsi): Likewise.
4370 (store_pairdi): Likewise.
4371 (load_pairsf): Likewise.
4372 (load_pairdf): Likewise.
4373 (store_pairsf): Likewise.
4374 (store_pairdf): Likewise.
4375 (zero_extend): Likewise.
4376 (trunc): Swap alternatives.
4377 (fcvt_target): Add '?' to prefer w over r.
4379 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4382 * config/i386/i386.md: Disallow non-commutative arithmetics in
4383 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4384 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4385 in the peephole2 before it.
4387 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4389 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4390 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4391 (ix86_handle_option): Handle -mcldemote.
4392 * config.gcc: New header.
4393 * config/i386/cldemoteintrin.h: New file.
4394 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4395 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4397 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4398 OPTION_MASK_ISA_CLDEMOTE.
4399 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4400 (ix86_valid_target_attribute_inner_p): Ditto.
4401 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4402 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4403 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4404 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4405 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4407 * config/i386/i386.opt: Add -mcldemote.
4408 * config/i386/x86intrin.h: New header.
4409 * doc/invoke.texi: Add -mcldemote.
4411 2018-05-14 Richard Biener <rguenther@suse.de>
4413 * doc/match-and-simplify.texi: Adjust :s documentation.
4415 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4417 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4418 intended memcpy size.
4419 (REORDER_45): Likewise.
4421 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4423 * sort.cc: New file.
4424 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4425 * vec.c (qsort_chk): Use gcc_qsort.
4426 * Makefile.in (OBJS-libcommon): Add sort.o.
4427 (build/sort.o): New target. Use it...
4428 (BUILD_RTL): ... here, and...
4429 (build/gencfn-macros): ... here, and...
4430 (build/genmatch): ... here.
4432 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4433 Chung-Ju Wu <jasonwucj@gmail.com>
4435 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4436 * config/nds32/nds32-graywolf.md: New file.
4437 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4438 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4440 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4441 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4442 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4443 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4444 * config/nds32/pipelines.md: Include n15 settings.
4446 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4447 Chung-Ju Wu <jasonwucj@gmail.com>
4449 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4450 * config/nds32/nds32-n13.md: New file.
4451 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4452 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4454 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4455 * config/nds32/nds32.md (pipeline_model): Add n13.
4456 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4457 * config/nds32/pipelines.md: Include n13 settings.
4459 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4460 Chung-Ju Wu <jasonwucj@gmail.com>
4462 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4463 * config/nds32/nds32-n10.md: New file.
4464 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4465 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4467 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4468 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4469 * config/nds32/nds32.md (pipeline_model): Add n10.
4470 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4471 * config/nds32/pipelines.md: Include n10 settings.
4473 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4474 Kito Cheng <kito.cheng@gmail.com>
4475 Chung-Ju Wu <jasonwucj@gmail.com>
4477 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4478 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4479 Add enum values for DSP extension instructions.
4480 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4482 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4483 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4485 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4486 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4487 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4488 * config/nds32/nds32-intrinsic.md: Likewise.
4489 * config/nds32/nds32_intrinsic.h: Likewise.
4490 * config/nds32/nds32-md-auxiliary.c: Likewise.
4491 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4492 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4493 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4494 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4495 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4496 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4497 TYPE_DMAC in switch statement.
4498 * config/nds32/nds32.c: New checking and implementation for DSP
4499 extension instructions.
4500 * config/nds32/nds32.h: Likewise.
4501 * config/nds32/nds32.md: Likewise.
4502 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4503 * config/nds32/predicates.md: Implement new predicates for DSP
4506 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4508 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4509 Reformat alternatives and attributes so it is easier to identify
4510 which constraints/attributes go with which instruction.
4511 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4512 (mov<mode>_softfloat32, FMOVE64): Likewise.
4513 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4514 (mov<mode>_softfloat64, FMOVE64): Likewise.
4516 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4518 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4520 (Basic PowerPC Built-in Functions): The new name of the
4521 subsection previously known as "PowerPC Built-in Functions".
4522 (Basic PowerPC Built-in Functions Available on all Configurations):
4524 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4525 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4526 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4527 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4529 2018-05-11 Martin Jambor <mjambor@suse.cz>
4532 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4535 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4538 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4540 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4542 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4543 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4544 (ix86_handle_option): Handle -mwaitpkg.
4545 * config.gcc: New header.
4546 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4547 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4548 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4550 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4551 OPTION_MASK_ISA_WAITPKG.
4552 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4553 (ix86_option_override_internal): Add PTA_WAITPKG.
4554 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4555 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4556 IX86_BUILTIN_TPAUSE.
4557 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4558 __builtin_ia32_umwait and __builtin_ia32_tpause.
4559 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4560 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4561 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4562 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4563 UNSPECV_TPAUSE): New.
4564 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4565 * config/i386/i386.opt: Add -mwaitpkg.
4566 * config/i386/waitpkgintrin.h: New file.
4567 * config/i386/x86intrin.h: New header.
4568 * doc/invoke.texi: Add -mwaitpkg.
4570 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4573 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4575 (cortex-m0): Use armv6s-m isa.
4576 (cortex-m0plus): Likewise.
4577 (cortex-m1): Likewise.
4578 (cortex-m0.small-multiply): Likewise.
4579 (cortex-m0plus.small-multiply): Likewise.
4580 (cortex-m1.small-multiply): Likewise.
4582 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4583 Jakub Jelinek <jakub@redhat.com>
4585 PR tree-optimization/85692
4586 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4587 source permute as well.
4589 2018-05-11 Martin Liska <mliska@suse.cz>
4592 * doc/extend.texi: Document LLVM style format for no_sanitize
4595 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4597 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4598 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4599 (mode_supports_vsx_dform_quad): Likewise.
4600 (mode_supports_vmx_dform): Move these functions to be next to the
4601 other mode_supports functions.
4602 (mode_supports_dq_form): Likewise.
4603 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4604 mode_supports_dq_form.
4605 (reg_offset_addressing_ok_p): Likewise.
4606 (offsettable_ok_by_alignment): Likewise.
4607 (rs6000_legitimate_offset_address_p): Likewise.
4608 (legitimate_lo_sum_address_p): Likewise.
4609 (rs6000_legitimize_address): Likewise.
4610 (rs6000_legitimize_reload_address): Likewise.
4611 (rs6000_secondary_reload_inner): Likewise.
4612 (rs6000_preferred_reload_class): Likewise.
4613 (rs6000_output_move_128bit): Likewise.
4615 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4617 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4618 Generate SImode target register for null target.
4619 <case IX86_BUILTIN_XGETBV>: Ditto.
4620 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4621 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4623 2018-05-10 Carl Love <cel@us.ibm.com>
4625 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4626 dcbtt and dcbtstt if operands[2] is 0.
4628 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4631 * config/i386/sse.md (usadv64qi): New expander.
4633 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4635 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4636 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4637 -maltivec=be support.
4638 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4639 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4640 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4641 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4642 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4643 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4644 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4645 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4646 altivec_vsumsws): Adjust.
4647 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4648 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4649 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4651 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4652 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4653 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4654 (altivec_lve<VI_char>x): Delete expand.
4655 (*altivec_lve<VI_char>x_internal): Rename to...
4656 (altivec_lve<VI_char>x): ... this.
4657 (altivec_lvxl_<mode>): Delete expand.
4658 (*altivec_lvxl_<mode>_internal): Rename to ...
4659 (altivec_lvxl_<mode>): ... this.
4660 (altivec_stvxl_<mode>): Delete expand.
4661 (*altivec_stvxl_<mode>_internal): Rename to ...
4662 (altivec_stvxl_<mode>): ... this.
4663 (altivec_stve<VI_char>x): Delete expand.
4664 (*altivec_stve<VI_char>x_internal): Rename to ...
4665 (altivec_stve<VI_char>x): ... this.
4666 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4667 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4668 reduc_plus_scal_<mode>): Adjust.
4669 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4671 (rs6000_cpu_cpp_builtins): Adjust.
4672 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4673 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4674 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4675 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4676 -maltivec=be support.
4677 (rs6000_split_vec_extract_var): Adjust.
4678 (rs6000_split_v4si_init): Adjust.
4679 (swap_selector_for_mode): Delete.
4680 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4681 altivec_expand_stvex_be): Delete.
4682 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4683 -maltivec=be support.
4684 (rs6000_gimple_fold_builtin): Ditto.
4685 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4687 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4688 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4689 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4690 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4691 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4692 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4693 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4694 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4695 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4696 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4697 anonymous split): Adjust.
4698 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4699 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4701 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4703 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4704 when --with-gxx-include-dir is also specified.
4705 * configure: Regenerate.
4707 2018-05-09 Jim Wilson <jimw@sifive.com>
4710 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4711 * config/riscv/t-withmultilib: New.
4712 * config/riscv/withmultilib.h: New.
4713 * doc/install.texi: Document RISC-V --with-multilib-list support.
4715 2018-05-09 Richard Biener <rguenther@suse.de>
4717 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4719 (vect_bb_vectorization_profitable_p): Adjust. Compute
4720 actual scalar cost using the cost vector and the add_stmt_cost
4723 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4725 PR rtl-optimization/85645
4726 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4727 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4729 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4731 PR rtl-optimization/85645
4732 * shrink-wrap.c (spread_components): Return a boolean saying if
4733 anything was changed.
4734 (try_shrink_wrapping_separate): Iterate spread_components until
4735 nothing changes anymore.
4737 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4739 PR rtl-optimization/85645
4740 * regrename.c (build_def_use): Also kill the chains that include the
4741 destination of a REG_CFA_REGISTER note.
4743 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4745 PR rtl-optimization/85645
4746 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4747 insn that has a REG_CFA_REGISTER note.
4749 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4751 * cfgexpand.c (expand_clobber): New function.
4752 (expand_gimple_stmt_1): Use it.
4753 * tree-vect-stmts.c (vect_clobber_variable): New function,
4755 (vectorizable_simd_clone_call): ...here.
4756 (vectorizable_store): Emit a clobber either side of an
4757 IFN_STORE_LANES sequence.
4758 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4760 2018-05-09 Tom de Vries <tom@codesourcery.com>
4763 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4764 (define_insn "trap_if_false"): Add exit after trap.
4766 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4768 PR rtl-optimization/85638
4769 * bb-reorder.c: Include common/common-target.h.
4770 (create_forwarder_block): New function extracted from...
4771 (fix_up_crossing_landing_pad): ...here. Rename into...
4772 (dw2_fix_up_crossing_landing_pad): ...this.
4773 (sjlj_fix_up_crossing_landing_pad): New function.
4774 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4775 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4776 from both partitions and exit the loop after one iteration.
4778 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4781 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4783 (Basic PowerPC Built-in Functions): The new name of the
4784 subsection previously known as "PowerPC Built-in Functions".
4785 (Basic PowerPC Built-in Functions Available on all Configurations):
4787 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4789 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4791 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4793 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4796 2018-05-08 Jim Wilson <jimw@sifive.com>
4798 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4799 (LD_EMUL_SUFFIX): New.
4800 (LINK_SPEC): Use it.
4802 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4804 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4806 (Basic PowerPC Built-in Functions): The new name of the
4807 subsection previously known as "PowerPC Built-in Functions".
4808 (Basic PowerPC Built-in Functions Available on all Configurations):
4810 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4812 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4814 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4816 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4819 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4822 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4823 after cmpelim optimization.
4825 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4827 * config.gcc: Support "goldmont".
4828 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4829 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4831 * config/i386/i386.c (m_GOLDMONT): Define.
4832 (processor_target_table): Add "goldmont".
4833 (PTA_GOLDMONT): Define.
4834 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4835 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4836 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4837 (fold_builtin_cpu): Add "goldmont".
4838 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4839 (ix86_option_override_internal): Add "goldmont".
4840 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4841 (processor_type): Add PROCESSOR_GOLDMONT.
4842 * config/i386/i386.md: Add CPU "glm".
4843 * config/i386/glm.md: New file.
4844 * config/i386/x86-tune.def: Add m_GOLDMONT.
4845 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4847 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4850 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4852 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4853 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4854 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4857 * config/i386/i386.c (ix86_fold_builtin): Handle
4858 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4861 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4862 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4864 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4867 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4868 (check_arch): Likewise.
4869 (check_fpu): Return the result rather than printing it.
4870 (end arch): Fix operator precedence.
4871 (end cpu): Likewise.
4872 (END): Print the result from check_fpu.
4874 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4875 Alan Hayward <alan.hayward@arm.com>
4876 David Sherwood <david.sherwood@arm.com>
4878 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4879 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4880 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4881 (*fcmuo<mode>_and): New patterns.
4883 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4885 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4886 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4887 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4888 (cmp_op, sve_imm_con): New code attributes.
4889 (SVE_COND_INT_CMP, imm_con): Delete.
4890 (cmp_op): Remove above unspecs from int attribute.
4891 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4893 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4894 comparison-specific unspecs.
4895 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4896 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4897 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4898 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4899 (*vec_fcm<cmp_op><mode>): Rename to...
4900 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4901 (*vec_fcmuo<mode>): Rename to...
4902 (*fcmuo<mode>): ...this and adjust likewise.
4903 (*pred_fcm<cmp_op><mode>): New pattern.
4904 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4905 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4907 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4909 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4910 (aarch64_emit_sve_predicated_cond): New function.
4911 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4912 (aarch64_emit_unspec_cond_or): Replace with...
4913 (aarch64_emit_sve_or_conds): ...this new function. Use
4914 aarch64_emit_sve_ptrue_op for the individual comparisons and
4915 aarch64_emit_binop to OR them together.
4916 (aarch64_emit_inverted_unspec_cond): Replace with...
4917 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4918 aarch64_emit_sve_ptrue_op for the comparison and
4919 aarch64_emit_unop to invert the result.
4920 (aarch64_expand_sve_vec_cmp_float): Update after the above
4921 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4923 2018-05-07 Nathan Sidwell <nathan@acm.org>
4925 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4926 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4927 (Backwards Compatibility): Likewise.
4929 2018-05-07 Luis Machado <luis.machado@linaro.org>
4933 2018-05-07 Luis Machado <luis.machado@linaro.org>
4935 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4936 <prefetch_dynamic_strides>: New const bool field.
4937 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4938 prefetch_dynamic_strides.
4939 (exynosm1_prefetch_tune): Likewise.
4940 (thunderxt88_prefetch_tune): Likewise.
4941 (thunderx_prefetch_tune): Likewise.
4942 (thunderx2t99_prefetch_tune): Likewise.
4943 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4945 (aarch64_override_options_internal): Update to set
4946 PARAM_PREFETCH_DYNAMIC_STRIDES.
4947 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4948 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4949 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4950 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4951 prefetch-dynamic-strides setting.
4953 2018-05-07 Luis Machado <luis.machado@linaro.org>
4955 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4956 <minimum_stride>: New const int field.
4957 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4958 minimum_stride field.
4959 (exynosm1_prefetch_tune): Likewise.
4960 (thunderxt88_prefetch_tune): Likewise.
4961 (thunderx_prefetch_tune): Likewise.
4962 (thunderx2t99_prefetch_tune): Likewise.
4963 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4964 (aarch64_override_options_internal): Update to set
4965 PARAM_PREFETCH_MINIMUM_STRIDE.
4966 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4967 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4968 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4969 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4970 stride is constant and is below the minimum stride threshold.
4972 2018-05-07 Luis Machado <luis.machado@linaro.org>
4974 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4977 2018-05-07 Luis Machado <luis.machado@linaro.org>
4979 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4980 <prefetch_dynamic_strides>: New const bool field.
4981 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4982 prefetch_dynamic_strides.
4983 (exynosm1_prefetch_tune): Likewise.
4984 (thunderxt88_prefetch_tune): Likewise.
4985 (thunderx_prefetch_tune): Likewise.
4986 (thunderx2t99_prefetch_tune): Likewise.
4987 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4989 (aarch64_override_options_internal): Update to set
4990 PARAM_PREFETCH_DYNAMIC_STRIDES.
4991 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4992 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4993 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4994 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4995 prefetch-dynamic-strides setting.
4997 2018-05-07 Luis Machado <luis.machado@linaro.org>
4999 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5000 <minimum_stride>: New const int field.
5001 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5002 minimum_stride field.
5003 (exynosm1_prefetch_tune): Likewise.
5004 (thunderxt88_prefetch_tune): Likewise.
5005 (thunderx_prefetch_tune): Likewise.
5006 (thunderx2t99_prefetch_tune): Likewise.
5007 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5008 (aarch64_override_options_internal): Update to set
5009 PARAM_PREFETCH_MINIMUM_STRIDE.
5010 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5011 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5012 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5013 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5014 stride is constant and is below the minimum stride threshold.
5016 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5019 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5020 the type is addressable. Don't force op into register if it has
5023 2018-05-05 Roland McGrath <mcgrathr@google.com>
5026 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5027 any section for which we don't know a specific type it should have,
5028 regardless of name. Previously this was done only for the exact
5029 names ".init_array", ".fini_array", and ".preinit_array".
5030 (default_elf_asm_named_section): Add comment about
5031 relationship with default_section_type_flags and SECTION_NOTYPE.
5032 (get_section): Don't consider it a type conflict if one side has
5033 SECTION_NOTYPE and the other doesn't, as long as neither has the
5034 SECTION_BSS et al used in the default_section_type_flags logic.
5036 2018-05-05 Tom de Vries <tom@codesourcery.com>
5039 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5040 (workaround_barsyncs): New function.
5041 (nvptx_reorg): Use workaround_barsyncs.
5042 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5043 (define_expand "nvptx_membar_cta"): New define_expand.
5044 (define_insn "*nvptx_membar_cta"): New insn.
5046 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5048 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5049 To improve optimization opportunities.
5050 * builtin-types.def: The new needed builtin types for the above.
5052 2018-05-04 Richard Biener <rguenther@suse.de>
5054 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5055 * gimple-ssa-store-merging.c
5056 (imm_store_chain_info::output_merged_store): Remove redundant create,
5057 release split_store vector contents on failure.
5058 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5059 scalar stmt vector on cache hit.
5061 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5063 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5065 * config.gcc (powerpc-xilinx-eabi*): Remove.
5066 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5068 (fusion_addis_mem_combo_load): Ditto.
5069 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5071 (rs6000_cpu_cpp_builtins): Ditto.
5072 * config/rs6000/rs6000-linux.c
5073 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5074 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5075 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5077 (rs6000_setup_reg_addr_masks): Ditto.
5078 (rs6000_init_hard_regno_mode_ok): Ditto.
5079 (rs6000_option_override_internal): Ditto.
5080 (legitimate_lo_sum_address_p): Ditto.
5081 (rs6000_legitimize_address): Ditto.
5082 (rs6000_legitimize_reload_address): Ditto.
5083 (rs6000_legitimate_address_p): Ditto.
5084 (abi_v4_pass_in_fpr): Ditto.
5085 (setup_incoming_varargs): Ditto.
5086 (rs6000_gimplify_va_arg): Ditto.
5087 (rs6000_split_multireg_move): Ditto.
5088 (rs6000_savres_strategy): Ditto.
5089 (rs6000_emit_prologue_components): Ditto.
5090 (rs6000_emit_epilogue_components): Ditto.
5091 (rs6000_emit_prologue): Ditto.
5092 (rs6000_emit_epilogue): Ditto.
5093 (rs6000_elf_file_end): Ditto.
5094 (rs6000_function_value): Ditto.
5095 (rs6000_libcall_value): Ditto.
5096 * config/rs6000/rs6000.h: Ditto.
5097 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5098 (TARGET_MINMAX): ... this. New.
5099 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5100 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5101 (*movsi_internal1_single): Delete.
5102 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5103 mfpu=, mxilinx-fpu): Delete.
5104 * config/rs6000/singlefp.h: Delete.
5105 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5106 * config/rs6000/t-rs6000: Ditto.
5107 * config/rs6000/t-xilinx: Delete.
5108 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5109 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5110 (VStype_simple): Delete.
5111 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5112 * config/rs6000/xfpu.h: Delete.
5113 * config/rs6000/xfpu.md: Delete.
5114 * config/rs6000/xilinx.h: Delete.
5115 * config/rs6000/xilinx.opt: Delete.
5116 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5117 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5119 2018-05-04 Tom de Vries <tom@codesourcery.com>
5122 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5125 2018-05-04 Richard Biener <rguenther@suse.de>
5128 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5129 (expand_complex_div_wide): Likewise.
5130 (expand_complex_operations_1): Likewise.
5131 (expand_complex_libcall): Preserve EH info of the original stmt.
5132 (tree_lower_complex): Handle removed blocks.
5133 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5134 on complex multiplication and division libcall builtins.
5136 2018-05-04 Richard Biener <rguenther@suse.de>
5139 * fold-const.c (negate_expr_p): Restrict negation of operand
5140 zero of a division to when we know that can happen without
5142 (fold_negate_expr_1): Likewise.
5144 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5147 * real.h (real_nextafter): Declare.
5148 * real.c (real_nextafter): New function.
5149 * fold-const-call.c (fold_const_nextafter): New function.
5150 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5151 CASE_CFN_NEXTTOWARD.
5152 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5153 even when arg1_mode is different from arg0_mode.
5155 2018-05-03 Nathan Sidwell <nathan@acm.org>
5157 * doc/extend.texi (Deprecated Features): Remove
5159 (Backwards Compatibility): Likewise.
5160 * doc/invoke.texi (C++ Language Options): Likewise.
5161 (C++ Dialect Options): Likewise.
5163 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5166 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5167 _mm512_mask_mullox_epi64): New intrinsics.
5169 2018-05-03 Tom de Vries <tom@codesourcery.com>
5172 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5173 dump files): Add offload-tree.
5175 2018-05-03 Richard Biener <rguenther@suse.de>
5177 PR tree-optimization/85615
5178 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5179 to loops not nested in BBs loop father to avoid creating multi-entry
5182 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5184 PR tree-optimization/70291
5185 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5186 arguments. Change return type to tree. Emit libcall as a new
5187 statement rather than replacing existing one when inplace_p is true.
5188 (expand_complex_multiplication_components): New function.
5189 (expand_complex_multiplication): Expand floating-point complex
5190 multiplication using the above.
5191 (expand_complex_division): Rename inner_type parameter to type.
5192 Update expand_complex_libcall call-site.
5193 (expand_complex_operations_1): Update expand_complex_multiplication
5194 and expand_complex_division call-sites.
5196 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5199 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5200 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5201 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5202 the highest significant bit of the shift count mask is clear. In
5203 check whether and[sq]i3 is needed verify that all significant bits
5204 of the shift count other than the highest are set.
5206 2018-05-02 Tom de Vries <tom@codesourcery.com>
5209 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5210 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5211 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5212 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5213 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5214 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5215 __builtin_goacc_parlevel_size.
5217 2018-05-02 Richard Biener <rguenther@suse.de>
5219 PR tree-optimization/85597
5220 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5221 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5224 2018-05-02 Tom de Vries <tom@codesourcery.com>
5227 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5228 dump files): Add ltrans-tree.
5230 2018-05-02 Tom de Vries <tom@codesourcery.com>
5233 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5234 dump files): Add wpa-ipa.
5236 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5238 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5239 powerpc*-*-linux*paired* target.
5240 * config/rs6000/750cl.h: Delete.
5241 * config/rs6000/paired.h: Delete.
5242 * config/rs6000/paired.md: Delete.
5243 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5245 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5246 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5247 comment. Remove paired float support.
5248 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5249 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5251 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5252 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5254 * config/rs6000/rs6000.c: Remove paired float support.
5255 (paired_expand_vector_init, paired_expand_vector_move,
5256 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5257 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5258 paired_expand_builtin, paired_expand_predicate_builtin,
5259 paired_init_builtins): Delete.
5260 * config/rs6000/rs6000.h: Remove paired float support.
5261 * config/rs6000/rs6000.md: Remove paired float support.
5262 (move_from_CR_ov_bit): Delete.
5263 * config/rs6000/rs6000.opt (mpaired): Delete.
5264 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5265 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5267 2018-05-02 Richard Biener <rguenther@suse.de>
5270 * gimplify.c (gimplify_save_expr): When in SSA form allow
5271 SAVE_EXPRs to compute to SSA vars.
5273 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5276 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5277 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5278 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5279 clobber operands[2], instead use a new pseudo. Formatting fixes.
5281 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5283 PR tree-optimization/85586
5284 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5285 exit early for statements in the same group if the accesses are
5288 2018-05-02 Tom de Vries <tom@codesourcery.com>
5291 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5294 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5296 PR tree-optimization/85143
5297 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5299 2018-05-01 Tom de Vries <tom@codesourcery.com>
5302 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5303 not found" error message.
5305 2018-05-01 Tom de Vries <tom@codesourcery.com>
5308 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5309 * vec.c (test_ordered_remove_if): New function.
5310 (vec_c_tests): Call test_ordered_remove_if.
5311 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5312 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5313 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5314 VEC_ORDERED_REMOVE_IF.
5316 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5318 PR tree-optimization/82665
5319 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5320 pointer subtraction where arguments come from a memchr call.
5322 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5324 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5325 --push-state --as-needed and --pop-state instead of --as-needed and
5326 --no-as-needed if ld supports it.
5327 * configure: Regenerated.
5330 * doc/install.texi2html: Replace _002d with - and _002a with * in
5331 generated html files using sed.
5333 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5336 * gcc-rich-location.c (blank_line_before_p): New function.
5337 (use_new_line): New function.
5338 (gcc_rich_location::add_fixit_insert_formatted): New function.
5339 * gcc-rich-location.h
5340 (gcc_rich_location::add_fixit_insert_formatted): New function.
5342 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5344 * selftest.c (assert_streq): Rename "expected" and "actual" to
5345 "val1" and "val2". Extend NULL-handling to cover both inputs
5346 symmetrically, while still requiring both to be non-NULL for a pass.
5347 * selftest.h (assert_streq): Rename "expected" and "actual" to
5349 (ASSERT_EQ): Likewise.
5350 (ASSERT_EQ_AT): Likewise.
5351 (ASSERT_KNOWN_EQ): Likewise.
5352 (ASSERT_KNOWN_EQ_AT): Likewise.
5353 (ASSERT_NE): Likewise.
5354 (ASSERT_MAYBE_NE): Likewise.
5355 (ASSERT_MAYBE_NE_AT): Likewise.
5356 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5357 the assertion to pass.
5358 (ASSERT_STREQ_AT): Likewise.
5360 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5362 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5363 interaction with -pie.
5365 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5367 * selftest.h: Fix alphabetization of per-source-file selftest
5370 2018-04-30 Jason Merrill <jason@redhat.com>
5372 PR c++/61982 - dead stores to destroyed objects.
5373 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5376 2018-04-30 Jason Merrill <jason@redhat.com>
5378 * tree.c (build_clobber): New.
5379 * tree.h: Declare it.
5380 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5382 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5384 * diagnostic-show-locus.c (layout::layout): Update for
5385 location_get_source_line returning a char_span.
5386 (struct char_span): Move to input.h.
5387 (struct correction): Update for fields in char_span becoming
5389 (struct source_line): Update for location_get_source_line
5390 returning a char_span.
5391 (layout::print_line): Likewise.
5392 * edit-context.c (edited_file::print_content): Likewise.
5393 (edited_file::print_diff_hunk): Likewise.
5394 (edited_file::print_run_of_changed_lines): Likewise.
5395 (edited_file::get_num_lines): Likewise.
5396 (edited_line::edited_line): Likewise.
5397 * final.c (asm_show_source): Likewise.
5398 * input.c (location_get_source_line): Convert return type
5399 from const char * to char_span, losing the final "line_len"
5401 (dump_location_info): Update for the above.
5402 (get_substring_ranges_for_loc): Likewise. Use a char_span
5403 when handling the literal within the line.
5404 (test_reading_source_line): Update for location_get_source_line
5405 returning a char_span.
5406 * input.h (class char_span): Move here from
5407 diagnostic-show-locus.c, converting from a struct to a class.
5408 Make data members private.
5409 (char_span::operator bool): New.
5410 (char_span::length): New.
5411 (char_span::get_buffer): New.
5412 (char_span::operator[]): New.
5413 (char_span::subspan): Make const.
5414 (char_span::xstrdup): New.
5415 (location_get_source_line): Convert return type from const char *
5416 to char_span, losing the final "line_size" param.
5418 2018-04-30 Jan Hubicka <jh@suse.cz>
5420 * lto-wrapper.c (ltrans_priorities): New static var.
5421 (cmp_priority): New.
5422 (run_gcc): Read priorities and if doing parallel build order
5423 the Makefile by them.
5425 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5427 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5429 2018-04-30 Richard Biener <rguenther@suse.de>
5431 * tree-cfg.c (verify_address): Remove base argument, add
5432 flag whether to check TREE_ADDRESSABLE and do that.
5433 (verify_expr): Remove.
5434 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5435 (verify_gimple_assign_single): Likewise.
5436 (verify_gimple_switch): Likewise.
5437 (verify_expr_location_1): Dereference tp once. Add (disabled)
5438 piece from verify_expr.
5439 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5441 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5443 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5445 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5447 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5448 (small_data_pattern): Likewise.
5449 (arc_rewrite_small_data): Likewise.
5450 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5451 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5452 (get_symbol_alignment): New function.
5453 (legitimate_small_data_address_p): Likewise.
5454 (legitimate_scaled_address): Update, call
5455 legitimate_small_data_address_p.
5456 (output_sdata): New static variable.
5457 (arc_print_operand): Update how we handle small data operands.
5458 (arc_print_operand_address): Likewise.
5459 (arc_legitimate_address_p): Update, use
5460 legitimate_small_data_address_p.
5461 (arc_rewrite_small_data_p): Remove.
5462 (arc_rewrite_small_data_1): Likewise.
5463 (arc_rewrite_small_data): Likewise.
5464 (small_data_pattern): Likewise.
5465 (compact_sda_memory_operand): Update to use
5466 legitimate_small_data_address_p and get_symbol_alignment.
5467 (prepare_move_operands): Don't rewite sdata pattern.
5468 (prepare_extend_operands): Remove.
5469 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5471 (zero_extendqisi2): Likewise.
5472 (zero_extendhisi2): Likewise.
5473 (extendqihi2): Likewise.
5474 (extendqisi2): Likewise.
5475 (extendhisi2): Likewise.
5479 * config/arc/constraints.md (Usd): Change it to memory constraint.
5481 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5483 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5484 as source of std instructions.
5485 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5486 allow 6-bit constants as source for store instructions.
5487 (movdi_insn): Update instruction pattern to allow 6-bit constants
5488 as source for store instructions.
5490 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5492 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5494 2018-04-30 Nathan Sidwell <nathan@acm.org>
5495 Sandra Loosemore <sandra@codesourcery.com>
5497 * dumpfile.c (dump_open): Allow '-' for stdout.
5498 * doc/invoke.texi (Developer Options): Document dump filename
5499 determination early. Document stdin/stdout selection.
5501 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5503 Microblaze Target: PIC data text relative
5505 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5506 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5508 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5509 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5510 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5511 New addressing mode for data-text relative position indepenedent code.
5512 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5513 'ADDRESS_SYMBOLIC_TXT_REL'.
5514 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5515 (microblaze_legitimate_pic_operand): Exclude function calls from
5516 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5517 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5519 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5520 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5521 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5522 for 'address + offset'.
5523 (microblaze_expand_prologue): Add new function prologue call for
5525 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5526 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5527 table in case of TARGET_PIC_DATA_TEXT_REL.
5528 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5529 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5530 Add new macros 'UNSPEC_TEXT',
5531 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5532 + exclude function calls from 'UNSPEC_PLT' in case of data text
5534 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5535 new target hook for generating address diff vector tables in case of
5537 * doc/tm.texi : Regenerate.
5538 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5539 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5540 of addr diff vector generation.
5541 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5542 target hook definition.
5543 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5544 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5545 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5546 option for data text relative.
5548 2018-04-30 Richard Biener <rguenther@suse.de>
5550 * tree-chrec.h (evolution_function_is_constant_p): Remove
5552 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5554 2018-04-30 Richard Biener <rguenther@suse.de>
5557 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5559 2018-04-30 Richard Biener <rguenther@suse.de>
5561 PR tree-optimization/28364
5562 PR tree-optimization/85275
5563 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5564 copying first exit test.
5566 2018-04-28 Mark Wielaard <mark@klomp.org>
5568 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5570 (dwarf_AT): Handle DW_AT_addr_base.
5571 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5573 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5576 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5577 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5578 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5579 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5581 2018-04-28 Richard Biener <rguenther@suse.de>
5583 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5584 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5585 to reflect use. Only add interesting stmts.
5587 2018-04-27 Martin Jambor <mjambor@suse.cz>
5590 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5591 the jump function allows for passing through aggregate values.
5593 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5595 * input.h (in_system_header_at): Convert from macro to inline
5597 (from_macro_expansion_at): Likewise.
5598 (from_macro_definition_at): Likewise.
5600 2018-04-27 Jeff Law <law@redhat.com>
5602 * config.gcc: Mark tile* targets as deprecated/obsolete.
5604 2018-04-27 Richard Biener <rguenther@suse.de>
5606 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5609 2018-04-27 Richard Biener <rguenther@suse.de>
5611 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5613 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5615 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5616 with Yd constraint. Set "preferred_for_speed" attribute from
5617 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5619 (*movdi_internal): Ditto.
5620 (movti_interunit splitters): Remove
5621 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5622 (movdi_interunit splitters): Ditto.
5623 * config/i386/constraints.md (Ye): Remove.
5624 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5626 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5629 * config/aarch64/constraints.md (Usg): Limit to 31.
5632 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5634 PR tree-optimization/85529
5635 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5636 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5637 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5638 zero extension or masking of the MSB bit.
5639 (optimize_range_tests): Add FIRST_BB argument, pass it through
5640 to optimize_range_tests_var_bound.
5641 (maybe_optimize_range_tests, reassociate_bb): Adjust
5642 optimize_range_tests callers.
5644 2018-04-26 Richard Biener <rguenther@suse.de>
5645 Jakub Jelinek <jakub@redhat.com>
5647 * cgraph.h (symbol_table): Just declare debug method here.
5648 * symtab.c (symbol_table::debug): Define.
5650 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5652 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5654 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5656 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5657 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5658 (*movdi_internal): Substitute Yi and Yj constraint with x
5659 and Ym and Yn constraint with y constraint. Update "isa"
5660 attribute and set "preferred_for_speed" attribute from
5661 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5662 (*movsi_internal): Ditto.
5663 (*movdf_internal): Ditto.
5664 (*movsf_internal): Ditto.
5665 (*zero_extendsidi2): Ditto.
5666 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5667 (sse2_loadld): Ditto.
5668 (*vec_extract<ssevecmodelower>_0): Ditto.
5669 (*vec_extractv4si_0_zext_sse4): Ditto.
5670 (vec_concatv2di): Ditto.
5671 (*vec_dup<mode>): Ditto.
5672 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5673 * config/i386/constraints.md (Yi): Remove.
5678 2018-04-26 Nathan Sidwell <nathan@acm.org>
5680 * dumpfile.c (dump_open): New.
5681 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5682 (dump_finish): Detect stdio/stderr by value not name.
5684 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5686 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5688 2018-04-26 Tom de Vries <tom@codesourcery.com>
5691 * config/nvptx/nvptx.c (verify_neutering_jumps)
5692 (verify_neutering_labels): New function
5693 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5695 2018-04-26 Tom de Vries <tom@codesourcery.com>
5698 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5699 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5702 2018-04-26 Richard Biener <rguenther@suse.de>
5703 Tom de Vries <tom@codesourcery.com>
5706 * lto-streamer-out.c (output_function): Fixup loops if required to match
5707 discovery done in the reader.
5709 2018-04-26 Richard Biener <rguenther@suse.de>
5711 PR tree-optimization/85116
5712 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5713 have a loop exit from the single latch predecessor. Remove
5714 case of header with just condition.
5715 (ch_base::copy_headers): Exclude infinite loops from any
5717 (pass_ch::execute): Record exits.
5719 2018-04-26 Richard Biener <rguenther@suse.de>
5721 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5722 prologue cost vector and pass it to vect_get_load_cost.
5723 (vect_get_peeling_costs_all_drs): Likewise.
5724 (vect_peeling_hash_get_lowest_cost): Likewise.
5725 (vect_enhance_data_refs_alignment): Likewise.
5727 2018-04-26 Richard Biener <rguenther@suse.de>
5730 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5731 checking of integer<->pointer conversions.
5732 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5733 sign-/zero-extending pointer types.
5734 (expand_omp_for_static_chunk): Likewise.
5736 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5737 Jean Lee <xiaoyur347@gmail.com>
5739 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5740 (TARGET_ASAN_SHADOW_OFFSET): Define.
5741 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5742 true for -fsanitize=address.
5744 2018-04-25 Mark Wielaard <mark@klomp.org>
5746 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5749 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5751 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5752 than "alu", remove explicit "memory" and "imm_disp" attributes.
5753 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5756 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5757 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5758 gen_lowpart_no_emit.
5760 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5763 * config/i386/i386.c (ix86_expand_builtin): Change memory
5764 operand to XI, extend p0 to Pmode.
5765 * config/i386/i386.md: Change unspec volatile and operand
5766 1 mode to XI, change operand 0 mode to P.
5768 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5770 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5771 GET_MODE_MASK before any checking.
5772 (nds32_can_use_bset_p): Likewise.
5773 (nds32_can_use_btgl_p): Likewise.
5775 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5777 * config/nds32/nds32-doubleword.md: New define_split pattern for
5778 illegal register number.
5780 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5782 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5784 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5786 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5788 2018-04-25 Richard Biener <rguenther@suse.de>
5790 * lto-streamer.h (LTO_major_version): Bump to 8.
5792 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5794 * BASE-VER: Set to 9.0.0.
5796 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5798 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5799 in __abskf2 and __powikf2.
5801 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5804 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5805 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5806 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5807 Use the above on operand 2. Reindent.
5808 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5810 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5813 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5815 * config/i386/i386.opt (mcet): Removed.
5816 * doc/install.texi: Remove -mcet documentation.
5817 * doc/invoke.texi: Likewise.
5819 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5822 * doc/install.texi: Remove -mcet from bootstrap-cet.
5824 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5827 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5828 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5832 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5833 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5834 containing a CONST_VECTOR.
5836 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5838 * doc/install.texi: Update newlib dependency for nvptx.
5840 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5843 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5844 instead of INTVAL when shifting x left.
5846 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5848 PR tree-optimization/85478
5849 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5850 vect_grouped_store_supported for single element vectors.
5852 2018-04-24 Richard Biener <rguenther@suse.de>
5855 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5856 load cost increase to the case of non-constant step.
5858 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5861 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5862 destination if any_malformed_asm.
5864 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5867 * expr.c (store_field): In the bitfield case, if the value comes from
5868 a function call and is returned in registers by means of a PARALLEL,
5869 do not change the mode of the temporary unless BLKmode and VOIDmode.
5871 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5873 PR rtl-optimization/85423
5874 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5875 dependencies to debug insns when the previous insn is non-debug.
5877 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5879 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5880 enums into a single definition.
5881 (fls): Fix predicates and printing.
5884 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5886 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5887 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5888 and short u6 immediate.
5889 (check_if_valid_sleep_operand): Remove.
5890 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5892 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5894 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5895 flag_always_save_lp condition.
5896 * config/nds32/nds32.opt (malways-save-lp): New option.
5898 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5900 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5901 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5902 * config/nds32/nds32.h
5903 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5904 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5906 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5908 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5909 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5911 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5912 Chung-Ju Wu <jasonwucj@gmail.com>
5914 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5916 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5917 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5919 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5921 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5923 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5925 * config/nds32/nds32-protos.h (nds32_data_alignment,
5926 nds32_local_alignment): Declare.
5927 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5928 nds32_local_alignment): New functions.
5929 (TARGET_CONSTANT_ALIGNMENT): Define.
5930 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5932 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5934 * config/nds32/nds32.c
5935 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5936 (TARGET_MODES_TIEABLE_P): Likewise.
5938 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5940 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5943 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5944 Chung-Ju Wu <jasonwucj@gmail.com>
5946 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5947 for unaligned access.
5948 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5950 * config/nds32/nds32-intrinsic.md: Likewise.
5951 * config/nds32/nds32_intrinsic.h: Likewise.
5952 * config/nds32/nds32.h (nds32_builtins): Likewise.
5953 * config/nds32/nds32.opt (munaligned-access): New option.
5954 * config/nds32/nds32.c (nds32_asm_file_start): Display
5955 flag_unaligned_access status.
5957 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5959 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5960 -mno-relax is present.
5961 * config/riscv/linux.h (LINK_SPEC): Ditto.
5963 2018-04-20 Martin Sebor <msebor@redhat.com>
5966 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5968 (gimple_fold_builtin_stxcpy_chk): Same.
5969 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5971 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5974 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5975 __powikf2 when long double is IEEE 128-bit.
5977 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5979 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5980 step to make sure stack always aligned.
5982 2018-04-20 Carl Love <cel@us.ibm.com>
5985 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5986 size check for arg0.
5988 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5989 Tom de Vries <tom@codesourcery.com>
5992 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5993 Emit insns for calls too.
5994 (nvptx_find_par): Always look for worker-level predecessor insn.
5995 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5997 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5998 (nvptx_process_pars): Propagate frames for calls.
6000 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6003 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6005 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6006 (ix86_handle_option): Don't handle OPT_mibt.
6007 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6009 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6011 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6012 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6013 (ix86_target_macros): Define __CET__ with flag_cf_protection
6014 for -fcf-protection.
6015 * config/i386/i386.c (isa2_opts): Remove -mibt.
6016 * config/i386/i386.h (TARGET_IBT): Removed.
6017 (TARGET_IBT_P): Likewise.
6018 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6019 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6020 * config/i386/i386.opt (mcet): Update help message.
6023 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6024 -mcet as an alias for -mshstk.
6026 2018-04-20 Richard Biener <rguenther@suse.de>
6029 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6030 complexity by forcing a single use of the multiply operand.
6032 2018-04-20 Martin Jambor <mjambor@suse.cz>
6035 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6036 recursion dependency to only apply to non-clones.
6038 2018-04-20 Martin Jambor <mjambor@suse.cz>
6041 * ipa-cp.c (create_specialized_node): Check that clones of
6042 self-recursive edges exist during IPA-CP.
6044 2018-04-19 Toon Moene <toon@moene.org>
6046 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6049 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6051 PR tree-optimization/85467
6052 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6053 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6054 VECTOR_CST element to type.
6056 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6059 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6060 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6061 (builtin_longjmp): Likewise.
6062 (save_stack_nonlocal): New pattern.
6063 (restore_stack_nonlocal): Likewise.
6065 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6068 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6069 Replace ASM_OUTPUT_LABEL with fprintf.
6071 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6074 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6075 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6076 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6077 define __IBT__ and __SHSTK__ for -fcf-protection.
6078 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6080 (ix86_trampoline_init): Likewise.
6081 (x86_output_mi_thunk): Likewise.
6082 (ix86_notrack_prefixed_insn_p): Likewise.
6083 (ix86_option_override_internal): Don't disallow -fcf-protection.
6084 * config/i386/i386.md (rdssp<mode>): Also enable for
6086 (incssp<mode>): Likewise.
6087 (nop_endbr): Likewise.
6088 * config/i386/i386.opt (mcet): Change help message to built-in
6092 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6093 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6094 enable CET built-in functions.
6096 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6098 * common/config/i386/i386-common.c
6099 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6100 OPTION_MASK_ISA_MOVDIRI_UNSET,
6101 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6102 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6103 * config.gcc (movdirintrin.h): New header.
6104 * config/i386/cpuid.h (bit_MOVDIRI,
6105 bit_MOVDIR64B): New bits.
6106 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6108 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6109 (VOID, PVOID, PCVOID)): New function types.
6110 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6111 __builtin_ia32_directstoreu_u64,
6112 __builtin_ia32_movdir64b): New builtins.
6113 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6114 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6116 (ix86_valid_target_attribute_inner_p): Ditto.
6117 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6118 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6119 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6120 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6121 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6122 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6123 (movdiri<mode>, movdir64b_<mode>): New.
6124 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6125 * config/i386/immintrin.h: Include movdirintrin.h.
6126 * config/i386/movdirintrin.h: New file.
6127 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6129 2018-04-19 Richard Biener <rguenther@suse.de>
6132 * cfg.c (clear_bb_flags): When loop state says we have
6133 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6135 2018-04-19 Richard Biener <rguenther@suse.de>
6137 PR tree-optimization/84737
6138 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6139 copying restrict info.
6140 (vect_setup_realignment): Use it.
6141 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6142 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6143 the first DR to all generated stores.
6144 (vectorizable_load): Likewise for loads.
6146 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6148 PR tree-optimization/85446
6149 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6150 the integral and pointer types to have the same precision.
6152 * doc/install.texi: Document --disable-cet being the default and
6155 2018-04-18 Martin Liska <mliska@suse.cz>
6157 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6160 2018-04-18 Martin Liska <mliska@suse.cz>
6163 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6166 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6167 arguments if they are comparable.
6169 2018-04-18 Martin Liska <mliska@suse.cz>
6172 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6175 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6178 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6181 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6182 ENDBR after calling __morestack.
6184 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6187 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6188 by using gcc_base_ver to generate a gcc_driver_version, and use
6189 it when generating GCC_DRIVER_NAME.
6190 * configure: Regenerate.
6192 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6195 * config.gcc: Obsolete powerpc*-*-*spe*.
6197 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6200 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6201 (stabstr_D): Change type of unum from unsigned int to
6202 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6205 2018-04-17 Jim Wilson <jimw@sifive.com>
6208 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6209 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6210 Set arg_pointer_offset after using pretend_args_size.
6212 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6214 PR rtl-optimization/85431
6215 * dse.c (record_store): Ignore zero width stores.
6218 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6219 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6220 __builtin_stack_restore rather than after it.
6221 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6222 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6223 argument instead of virtual_dynamic_stack_rtx.
6225 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6227 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6229 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6230 Add note to error message to explain internal mapping of overloaded
6231 built-in function name to non-overloaded built-in function name.
6232 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6235 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6238 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6239 where the inputs overlap with the output.
6241 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6243 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6244 (=v, v) alternative and explicit "memory" attribute.
6245 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6246 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6248 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6249 "sselog1" type instead of "sselog".
6250 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6251 "sselog". Remove explicit "memory" attribute.
6252 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6253 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6255 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6256 "sselog1" type instead of "sselog". Remove explicit "memory"
6258 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6259 use "sselog1" type instead of "sselog". Remove explicit "memory"
6261 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6262 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6264 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6265 "sselog1" type instead of "sselog". Remove explicit "memory"
6267 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6268 use "sselog1" type instead of "sselog". Remove explicit "memory"
6272 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6275 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6278 2018-04-17 Martin Jambor <mjambor@suse.cz>
6281 * ipa-cp.c (create_specialized_node): Call
6282 expand_all_artificial_thunks if necessary.
6284 2018-04-17 Martin Liska <mliska@suse.cz>
6287 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6288 in message, remote space in between '_G' and '('.
6290 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6293 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6294 avx512f_vmcmp<mode>3<round_saeonly_name>,
6295 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6296 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6297 avx512f_rndscale<mode><round_saeonly_name>,
6298 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6299 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6300 Use %<iptr>2 instead of %2 for -masm=intel.
6301 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6302 avx512f_vcvttss2usi<round_saeonly_name>,
6303 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6305 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6306 avx512f_vcvttsd2usi<round_saeonly_name>,
6307 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6308 Use %q1 instead of %1 for -masm=intel.
6309 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6310 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6311 of %3 for -masm=intel.
6312 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6314 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6316 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6318 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6319 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6321 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6323 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6324 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6325 %g1 and one with %0 and %1.
6326 (avx512er_vmrcp28<mode><round_saeonly_name>,
6327 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6329 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6330 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6331 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6332 of %0 and %{%4%} for -masm=intel.
6333 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6334 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6335 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6336 order of %0 and %{%5%}%{z%} for -masm=intel.
6338 2018-04-17 Jan Hubicka <jh@suse.cz>
6341 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6343 2018-04-17 Martin Liska <mliska@suse.cz>
6346 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6347 for target_clone error message. Make default implementation
6348 clone to be a local declaration.
6349 (separate_attrs): Add new argument and check for an empty
6351 (expand_target_clones): Handle it.
6352 (ipa_target_clone): Make redirection just for target_clones
6355 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6356 Tom de Vries <tom@codesourcery.com>
6359 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6360 tiled basic blocks without omp continue statements.
6362 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6365 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6366 vec_extract expression as having side effects to make sure it gets
6369 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6372 * config/i386/i386.c (get_builtin_code_for_version): Check
6375 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6378 * gcc/config.gcc: Support "skylake".
6379 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6381 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6382 (processor_target_table): Add "skylake".
6383 (ix86_option_override_internal): Add "skylake".
6384 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6385 PROCESSOR_CANNONLAKE.
6386 (get_builtin_code_for_version): Fix priority for
6387 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6388 PROCESSOR_SKYLAKE-AVX512.
6389 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6390 (processor_type): Add PROCESSOR_SKYLAKE.
6392 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6393 Jason Merrill <jason@redhat.com>
6396 * convert.c (convert_to_integer_1): Use direct recursion for
6397 enumeral types and types with a precision less than the number
6398 of bits in their mode.
6400 2018-04-16 Julia Koval <julia.koval@intel.com>
6403 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6404 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6406 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6409 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6410 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6411 and -mno-direct-move.
6413 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6416 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6417 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6418 Ensure negative shifts result in {0}.
6420 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6422 PR rtl-optimization/79916
6423 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6424 regs (if any) to define how to gnerate SD moves when LRA is in
6427 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6429 PR rtl-optimization/85393
6430 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6431 * except.c (expand_dw2_landing_pad_for_region): Make static.
6432 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6433 a label and unconditional jump to old_bb, rather than
6434 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6437 PR rtl-optimization/85376
6438 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6439 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6440 instead of a specific value.
6442 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6443 Bin Cheng <bin.cheng@arm.com>
6445 PR tree-optimization/82965
6446 PR tree-optimization/83991
6447 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6448 by_profile_only parameter.
6449 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6450 information if the loop was predicted to iterate too many times.
6451 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6453 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6456 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6459 2018-04-13 Martin Liska <mliska@suse.cz>
6460 Jakub Jelinek <jakub@redhat.com>
6463 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6464 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6465 * builtins.c (expand_builtin_memory_copy_args): Use
6466 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6467 handle dest_addr == pc_rtx.
6469 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6472 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6473 asked to not generate direct moves.
6474 (fix_trunc<mode>si2_stfiwx): Similar.
6475 (fix_trunc<mode>si2_internal): Similar.
6477 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6480 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6481 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6482 lookup if dest in some wider mode is known to be const0_rtx and
6483 if so, record permanent equivalence for it to be ZERO_EXTEND of
6484 the narrower mode destination.
6486 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6488 * lto-streamer-out.c (output_function): Revert 259346.
6489 * omp-expand.c (expand_oacc_for): Likewise.
6491 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6493 PR rtl-optimization/85354
6494 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6495 * sel-sched.c (sel_global_init): ... here.
6497 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6500 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6501 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6502 mode for PE-COFF targets.
6503 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6504 (i386_pe_asm_lto_end): Likewise.
6505 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6506 (TARGET_ASM_LTO_END): Likewise.
6507 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6508 (i386_pe_asm_lto_start): New function.
6509 (i386_pe_asm_lto_end): Likewise.
6511 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6512 Richard Biener <rguenther@suse.de>
6515 * lto-streamer-out.c (output_function): Fix CFG loop state before
6517 * omp-expand.c (expand_oacc_for): Handle calls to internal
6518 functions like regular functions.
6520 2018-04-12 Richard Biener <rguenther@suse.de>
6523 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6524 for the early LTO debug to properly generate references to it
6525 during DIE emission. Do not re-use that for the skeleton for
6527 (dwarf2out_early_finish): Likewise.
6529 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6532 * config/i386/sse.md
6533 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6534 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6535 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6536 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6537 and output is a reg, avoid creating invalid lowpart subreg, but
6538 instead split into a 512-bit move. Don't split if not AVX512VL,
6539 input is xmm16+ reg and output is a mem.
6540 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6541 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6542 xmm16+ reg and output is a mem.
6544 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6546 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6547 also for flag_dwarf2_cfi_asm.
6549 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6551 PR rtl-optimization/85342
6552 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6553 a bool scalar var inside of the loop instead. Don't try to update
6554 recog_data.operand after failed apply_change_group.
6556 2018-04-12 Tom de Vries <tom@codesourcery.com>
6559 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6560 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6561 array with flexible array member as array without given dimension.
6562 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6563 argument for undefined param to true.
6565 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6568 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6569 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6570 from PowerPC section.
6571 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6572 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6573 help text that is too long.
6574 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6575 help text that is too long.
6576 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6577 help text that is too long.
6579 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6581 * config/alpha/alpha.md (stack_probe_internal): Rename
6582 from "probe_stack". Update all callers.
6584 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6586 PR rtl-optimization/84566
6587 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6588 sched_macro_fuse_insns.
6590 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6593 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6594 (compute_block_dependences): ... from here.
6596 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6598 PR tree-optimization/85331
6599 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6600 from int to HOST_WIDE_INT.
6602 2018-04-11 Martin Jambor <mjambor@suse.cz>
6605 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6606 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6607 not the same as the source val.
6608 (cgraph_edge_brings_value_p): New parameter.
6609 (gather_edges_for_value): Pass destination value to
6610 cgraph_edge_brings_value_p.
6611 (perhaps_add_new_callers): Likewise.
6612 (get_info_about_necessary_edges): Likewise and exclude values brought
6613 only by self-recursive edges.
6614 (create_specialized_node): Redirect only clones of self-calling edges.
6615 (+self_recursive_pass_through_p): New function.
6616 (find_more_scalar_values_for_callers_subset): Use it.
6617 (find_aggregate_values_for_callers_subset): Likewise.
6618 (known_aggs_to_agg_replacement_list): Removed.
6619 (decide_whether_version_node): Re-calculate known constants for all
6620 remaining context clones.
6622 2018-04-11 Richard Biener <rguenther@suse.de>
6625 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6626 from early DWARF output.
6627 (dwarf2out_early_finish): Output line info unconditionally into
6628 early DWARF and add reference to it.
6630 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6633 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6634 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6635 other than V2DFmode using iptr mode attribute.
6636 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6638 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6640 PR rtl-optimization/84659
6641 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6643 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6646 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6648 (output_loc_list): Pass address of a dummy size variable even in the
6649 locview handling loop.
6650 (index_location_lists): Add comment on why skip_loc_list_entry can't
6653 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6656 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6659 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6662 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6663 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6664 and -mstring-compare-inline-limit.
6666 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6669 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6670 for stack clash protection in a register whenever we need it to be in
6673 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6675 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6676 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6678 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6681 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6683 (mlong-double-): Ditto.
6684 * config/rs6000/sysv4.opt (msdata=): Ditto.
6685 (mtls-size=): Ditto.
6687 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6689 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6690 erroneous entries for
6691 "vector int vec_ldl (int, long int *)", and
6692 "vector unsigned int vec_ldl (int, unsigned long int *)".
6693 Add comments and entries for
6694 "vector bool char vec_ldl (int, bool char *)",
6695 "vector bool short vec_ldl (int, bool short *)",
6696 "vector bool int vec_ldl (int, bool int *)",
6697 "vector bool long long vec_ldl (int, bool long long *)",
6698 "vector pixel vec_ldl (int, pixel *)",
6699 "vector long long vec_ldl (int, long long *)",
6700 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6701 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6702 type tree bool_long_long_type_node and correct definition of
6703 bool_V2DI_type_node to make reference to this new type tree.
6704 (rs6000_mangle_type): Replace erroneous reference to
6705 bool_long_type_node with bool_long_long_type_node.
6706 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6707 comments to emphasize sign distinctions for char and int types and
6708 replace RS6000_BTI_bool_long constant with
6709 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6710 use of RS6000_BTI_pixel.
6711 (bool_long_type_node): Remove this macro definition.
6712 (bool_long_long_type_node): New macro definition
6714 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6716 PR rtl-optimization/85300
6717 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6718 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6719 simplify_unary_operation fails.
6721 2018-04-10 Martin Liska <mliska@suse.cz>
6723 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6724 cgraph_edge and ipa_ref.
6726 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6730 * config/i386/sse.md
6731 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6732 computation of the VEC_MERGE selector from mask.
6733 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6734 Fix decoding of the VEC_MERGE selector into mask.
6736 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6738 PR tree-optimization/85286
6739 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6741 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6743 * final.c (final_1): Set insn_last_address as well as
6744 insn_current_address.
6746 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6749 * explow.c (emit_stack_probe): Call validize_mem on memory location
6750 before passing it to gen_probe_stack. Create address operand and
6751 legitimize it for the probe_stack_address case.
6753 2018-04-09 Jan Hubicka <jh@suse.cz>
6756 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6757 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6758 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6760 (free_lang_data): Rebuild type inheritance graph.
6762 2018-04-09 Martin Sebor <msebor@redhat.com>
6764 * invoke.texi (-finline-small-functions): Mention other optimization
6766 (-findirect-inlining, -fpartial-inlining): Same.
6767 (-finline-functions-called-once): Same.
6768 (-freorder-blocks-and-partition): Same.
6770 2018-04-09 Jan Hubicka <jh@suse.cz>
6773 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6774 jumps; choose last target that matches the criteria (i.e.
6775 no partition changes for non-crossing jumps).
6776 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6777 support for redirecting crossing jumps to non-crossing.
6779 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6781 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6782 also for naked functions.
6784 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6786 * config/arc/arc.md (add_shift): New pattern.
6787 (add_shift2): Likewise.
6788 (sub_shift): Likewise.
6789 (sub_shift_cmp0_noout): Likewise.
6790 (compare_si_ashiftsi): Likewise.
6791 (xbfu_cmp0_noout): New combine pattern.
6792 (xbfu_cmp0"): Likewise.
6793 (movsi_set_cc_insn): Place the predicable variant first.
6794 (commutative_binary_cmp0_noout): Remove clobber.
6795 (commutative_binary_cmp0): New pattern.
6796 (noncommutative_binary_cmp0): Likewise.
6797 (noncommutative_binary_cmp0_noout): Likewise.
6798 (noncommutative_binary_comparison_result_used): Removed.
6799 (rsub_cmp0): New pattern.
6800 (rsub_cmp0_noout): Likewise.
6801 (extzvsi): Changed, keep only meaningful variants.
6802 (SQH, SEZ): New iterators.
6803 (SQH_postfix): New mode attribute.
6804 (SEZ_prefix): New code attribute.
6805 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6806 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6807 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6809 (noncommutative_operator): Check the availability of barrel
6812 2018-04-09 Richard Biener <rguenther@suse.de>
6814 PR tree-optimization/85284
6815 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6816 Only use the niter constraining form of simple_iv when the exit
6819 2018-04-09 Tom de Vries <tom@codesourcery.com>
6822 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6823 (define_expand "*memory_barrier"): New define_expand.
6824 (define_insn "memory_barrier"): New insn.
6826 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6828 PR rtl-optimization/80463
6829 PR rtl-optimization/83972
6830 PR rtl-optimization/83480
6832 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6833 correct producer for the insn.
6834 (tidy_control_flow): Fixup seqnos in case of debug insns.
6836 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6838 PR rtl-optimization/83913
6840 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6841 different sched-times when merging exprs.
6843 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6845 PR rtl-optimization/83962
6847 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6848 tidy_fallthru_edge and tidy_control_flow.
6850 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6852 PR rtl-optimization/83530
6854 * sel-sched.c (force_next_insn): New global variable.
6855 (remove_insn_for_debug): When force_next_insn is true, also leave only
6856 next insn in the ready list.
6857 (sel_sched_region): When the region wasn't scheduled, make another pass
6858 over it with force_next_insn set to 1.
6860 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6862 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6864 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6865 for interrupt control.
6866 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6867 functions for interrupt control.
6868 * config/nds32/nds32-intrinsic.md: Likewise.
6869 * config/nds32/nds32_intrinsic.h: Likewise.
6870 * config/nds32/nds32.h (nds32_builtins): Likewise.
6872 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6874 * config/nds32/nds32.c (nds32_init_machine_status,
6875 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6876 strict_aligned_p field.
6877 (nds32_expand_to_rtl_hook): New function.
6878 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6879 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6881 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6882 Chung-Ju Wu <jasonwucj@gmail.com>
6884 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6885 * config/nds32/nds32-n7.md: New file.
6886 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6887 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6889 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6890 * config/nds32/nds32.md (pipeline_model): Add n7.
6891 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6892 * config/nds32/pipelines.md: Include n7 settings.
6894 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6895 Chung-Ju Wu <jasonwucj@gmail.com>
6897 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6898 * config/nds32/nds32-e8.md: New file.
6899 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6900 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6902 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6903 * config/nds32/nds32.md (pipeline_model): Add e8.
6904 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6905 * config/nds32/pipelines.md: Include e8 settings.
6907 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6908 Chung-Ju Wu <jasonwucj@gmail.com>
6910 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6911 * config/nds32/nds32-n8.md: New file.
6912 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6913 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6915 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6916 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6917 * config/nds32/nds32.md (pipeline_model): Add n8.
6918 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6919 * config/nds32/pipelines.md: Include n8 settings.
6921 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6922 Chung-Ju Wu <jasonwucj@gmail.com>
6924 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6925 * config/nds32/nds32-n9-2r1w.md: New file.
6926 * config/nds32/nds32-n9-3r2w.md: New file.
6927 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6928 nds32_register_ports): New or modify for cpu n9.
6929 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6931 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6932 * config/nds32/nds32-utils.c: New file.
6933 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6934 TARGET_MUL_SLOW): Define.
6935 * config/nds32/nds32.md (pipeline_model): New attribute.
6936 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6937 New options that support cpu n9.
6938 * config/nds32/pipelines.md: Include n9 settings.
6939 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6941 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6943 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6944 information if necessary.
6945 (output_cond_branch_compare_zero): Likewise.
6946 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6947 (nds32_target_alignment): Refine for alignment.
6948 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6949 (FUNCTION_BOUNDARY): Modify.
6950 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6952 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6954 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6956 * config/nds32/constants.md (unspec_volatile_element): Add values for
6957 TLB operation and data prefetch.
6958 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6959 functions for TLB operation and data prefetch.
6960 * config/nds32/nds32-intrinsic.md: Likewise.
6961 * config/nds32/nds32_intrinsic.h: Likewise.
6962 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6963 (nds32_print_operand): Likewise.
6964 * config/nds32/nds32.h (nds32_builtins): Likewise.
6966 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6967 Andrew Pinski <pinsika@gcc.gnu.org>
6970 * match.pd: Use constant_boolean_node of correct type instead of
6971 boolean_true_node or boolean_false_node for simplifying
6972 pointer comparisons to zero.
6974 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6976 PR tree-optimization/80021
6977 * tree.c (verify_type_variant): Make error call in verify_variant_match
6978 translatable and remove final full stop.
6980 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6982 * config/nds32/constants.md (unspec_volatile_element): Add
6983 UNSPEC_VOLATILE_EH_RETURN.
6984 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6985 nds32_output_stack_pop): Support dwarf exception handling process.
6986 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6987 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6988 exception handling process.
6989 (nds32_compute_stack_frame): Likewise.
6990 (nds32_return_addr_rtx): Likewise.
6991 (nds32_initial_elimination_offset): Likewise.
6992 (nds32_expand_prologue): Likewise.
6993 (nds32_expand_epilogue): Likewise.
6994 (nds32_dynamic_chain_address): New function.
6995 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6997 (DYNAMIC_CHAIN_ADDRESS): Define.
6998 (EH_RETURN_DATA_REGNO): Define.
6999 (EH_RETURN_STACKADJ_RTX): Define.
7000 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7001 patterns for dwarf exception handling.
7003 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7005 * config/nds32/nds32.h: Clean up obsolete macros.
7007 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7009 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7010 Add enum values for particular instructions.
7011 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7012 particular intrinsic functions.
7013 * config/nds32/nds32-intrinsic.md: Likewise.
7014 * config/nds32/nds32_intrinsic.h: Likewise.
7015 * config/nds32/nds32.h (nds32_builtins): Likewise.
7016 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7017 (btst, ave): New patterns for particular instructions.
7019 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7021 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7022 Add enum values for atomic load/store and memory sync.
7023 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7025 * config/nds32/nds32-intrinsic.md: Likewise.
7026 * config/nds32/nds32_intrinsic.h: Likewise.
7027 * config/nds32/nds32.h (nds32_builtins): Likewise.
7029 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7031 PR tree-optimization/85257
7032 * fold-const.c (native_encode_vector): If not all elts could fit
7033 and off is -1, return 0 rather than offset.
7034 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7035 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7036 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7037 adjust buffer in native_interpret_expr call.
7039 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7041 * config/nds32/constants.md (unspec_volatile_element): Add cache
7042 control enum values.
7043 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7044 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7045 * config/nds32/nds32.c (nds32_cctl_names): New.
7046 (nds32_print_operand): Handle cache control register names.
7047 * config/nds32/nds32.h (nds32_builtins): New enum values.
7048 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7050 * config/nds32/nds32.md (type): Add mmu.
7051 * config/nds32/pipelines.md (simple_insn): Add mmu.
7053 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7055 * config/nds32/nds32.md (type): Remove call.
7056 * config/nds32/pipelines.md (simple_insn): Likewise.
7058 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7060 * config/nds32/constants.md (unspec_volatile_element): Add
7061 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7062 UNSPEC_VOLATILE_FMFCFG.
7063 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7064 description for fmfcfg and fmfcsr.
7065 (bdesc_1arg): Add fmtcsr.
7066 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7067 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7068 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7069 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7070 unspec_fmfcfg): New patterns.
7071 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7072 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7073 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7074 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7075 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7076 __nds32__fmfcfg): Define.
7078 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7080 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7081 intrinsic register names.
7082 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7083 intrinsic register enum values and macros.
7085 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7087 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7088 for load/store addressing form.
7089 (nds32_print_operand_address): Likewise.
7091 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7094 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7095 based on LABEL_REF. Remove useless assertion.
7096 (pic_address_needs_scratch): Fix formatting.
7097 (sparc_legitimize_pic_address): Minor tweaks.
7098 (sparc_delegitimize_address): Adjust assertion accordingly.
7099 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7100 into symbolic_operand.
7101 (movsi_high_pic_label_ref): Likewise.
7102 (movsi_lo_sum_pic_label_ref): Likewise.
7103 (movdi_pic_label_ref): Likewise.
7104 (movdi_high_pic_label_ref): Likewise.
7105 (movdi_lo_sum_pic_label_ref): Likewise.
7107 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7109 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7110 custom LIB_SPEC setup.
7112 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7113 Kito Cheng <kito.cheng@gmail.com>
7115 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7116 * config/riscv/freebsd.h: New.
7118 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7120 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7121 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7124 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7125 Kito Cheng <kito.cheng@gmail.com>
7127 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7128 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7129 * config/nds32/nds32-protos.h (nds32_output_call,
7130 nds32_output_return): Declare.
7131 * config/nds32/nds32.md: Refine all the call and return patterns.
7133 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7136 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7137 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7139 PR rtl-optimization/84872
7140 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7141 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7144 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7146 * expr.c (copy_blkmode_to_reg): Revert 254862.
7147 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7149 2018-04-06 Richard Biener <rguenther@suse.de>
7152 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7153 after seeing a component reference with an adjacent field. Treat
7154 refs to arrays at struct end of external decls similar to
7155 refs to unconstrained commons.
7157 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7160 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7161 look through SAVE_EXPRs with non-side-effects argument. Adjust
7163 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7166 2018-04-06 Richard Biener <rguenther@suse.de>
7169 * alias.c (find_base_term): New wrapper around find_base_term
7170 unwinding CSELIB_VAL_PTR changes.
7171 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7174 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7176 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7178 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7179 constant definitions.
7180 ("nop"): lr 0,0 -> nopr r0
7181 ("nop_lr0", "nop_lr1"): New insn definitions.
7183 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7185 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7186 NDS32_V3PUSH_AVAILABLE_P macro.
7188 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7189 Chung-Ju Wu <jasonwucj@gmail.com>
7191 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7192 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7193 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7194 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7195 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7196 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7197 * config/nds32/constraints.md: New constraints and checking for hard
7198 float configuration.
7199 * config/nds32/iterators.md: New mode iterator and attribute for hard
7200 float configuration.
7201 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7203 * config/nds32/nds32-fpu.md: New file.
7204 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7205 deal with hard float code generation.
7206 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7208 (abi_type, float_reg_number): New enum type.
7209 * config/nds32/nds32-predicates.c: New predicates for hard float.
7210 * config/nds32/nds32-protos.h: Declare functions for hard float.
7211 * config/nds32/nds32.c: Implementation for hard float configuration.
7212 * config/nds32/nds32.h: Definitions for hard float configuration.
7213 * config/nds32/nds32.md: Include hard float machine description and
7214 modify patterns for hard float configuration.
7215 * config/nds32/nds32.opt: New options for hard float configuration.
7216 * config/nds32/predicates.md: New predicates for hard float
7219 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7221 * common/config/nds32/nds32-common.c
7222 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7224 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7227 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7228 CONSTRUCTOR_ELT (ctor, ...)->value.
7230 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7233 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7235 2018-04-05 Tom de Vries <tom@codesourcery.com>
7238 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7241 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7242 Kito Cheng <kito.cheng@gmail.com>
7244 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7245 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7246 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7247 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7249 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7250 Kito Cheng <kito.cheng@gmail.com>
7252 * config/nds32/constraints.md (Ufe): New memory constraint.
7253 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7254 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7255 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7257 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7258 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7260 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7262 * config/nds32/nds32.md: Use optimize_size in the condition for
7263 alu-shift instructions.
7265 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7267 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7269 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7271 * config/nds32/nds32.md (negsi2): Refine pattern.
7273 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7274 Chung-Ju Wu <jasonwucj@gmail.com>
7276 * config/nds32/iterators.md (shift_rotate): New code iterator.
7277 (shift): New code attribute.
7278 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7279 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7280 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7281 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7282 bit-wise operations.
7283 (andsi3, *andsi3): Ditto.
7284 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7285 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7286 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7287 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7288 nds32_ior_operand, nds32_xor_operand): New predicates.
7290 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7292 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7293 (addsi3, subsi3): ... this.
7295 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7297 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7299 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7301 * config/nds32/nds32.md: Adjust indention.
7303 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7305 * config/nds32/nds32.md (feature): New attribute.
7307 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7309 * config/nds32/nds32.md (subtype): New attribute.
7311 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7314 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7315 expansion to perform a bitwise AND of the argument followed by a
7316 boolean negation of the result.
7318 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7320 PR rtl-optimization/84878
7321 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7322 the basic block. Assert the use reference is not artificial and that
7323 it has an associated insn.
7325 2018-04-04 Michael Matz <matz@suse.de>
7327 * builtins.c (compute_objsize): Pass correct operand
7328 to array_at_struct_end_p.
7330 2018-04-04 Richard Biener <rguenther@suse.de>
7333 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7334 from contexts for DINFO_LEVEL_TERSE and below.
7336 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7338 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7339 resiter_operand condition.
7340 * config/nds32/nds32.md (*move<mode>): Ditto.
7342 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7343 Monk Chiang <sh.chiang04@gmail.com>
7345 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7347 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7351 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7352 Kito Cheng <kito.cheng@gmail.com>
7354 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7355 nds32_cond_code_str, output_cond_branch,
7356 output_cond_branch_compare_zero, nds32_expand_cbranch,
7357 nds32_expand_cstore, nds32_expand_movcc,
7358 nds32_output_cbranchsi4_equality_zero,
7359 nds32_output_cbranchsi4_equality_reg,
7360 nds32_output_cbranchsi4_equality_reg_or_const_int,
7361 nds32_output_cbranchsi4_greater_less_zero: New functions.
7362 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7363 nds32_expand_cstore, nds32_expand_movcc,
7364 nds32_output_cbranchsi4_equality_zero,
7365 nds32_output_cbranchsi4_equality_reg,
7366 nds32_output_cbranchsi4_equality_reg_or_const_int,
7367 nds32_output_cbranchsi4_greater_less_zero): Declare.
7368 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7369 nds32_rimm11s_operand): New predicates.
7370 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7371 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7374 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7376 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7377 * config/nds32/nds32.md: Ditto.
7378 * config/nds32/pipelines.md: Ditto.
7380 2018-04-04 Richard Biener <rguenther@suse.de>
7382 PR tree-optimization/85168
7383 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7384 propagating abnormals.
7386 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7388 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7390 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7391 Kito Cheng <kito.cheng@gmail.com>
7393 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7394 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7395 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7396 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7397 * config/nds32/nds32.md (sibcall_internal): New.
7398 (sibcall_register): Remove.
7399 (sibcall_immediate): Remove.
7400 (sibcall_value_internal): New.
7401 (sibcall_value_register): Remove.
7402 (sibcall_value_immediate): Remove.
7403 * config/nds32/predicates.md (nds32_general_register_operand): New.
7404 (nds32_call_address_operand): New.
7406 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7408 PR rtl-optimization/85167
7409 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7410 bb_defs if *split_p, instead preinitialize it to NULL.
7412 PR tree-optimization/85156
7413 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7414 evaluating the argument multiple times.
7416 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7418 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7420 (_mm_cvtpd_ps): Likewise.
7421 (_mm_cvttpd_epi32): Likewise.
7422 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7423 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7424 vector, pixel, and bool following altivec.h include.
7426 2018-04-03 Martin Sebor <msebor@redhat.com>
7428 * doc/extend.texi (Common Function Attributes): Clarify.
7429 (const attribute): Likewise.
7430 (pure attribute): Likewise.
7432 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7435 * config/i386/i386.c (ix86_expand_vector_set): Use
7436 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7438 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7440 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7441 instructions when changing rounding bits to preserve precision bits
7442 in the x87 control word.
7444 2018-04-03 Martin Liska <mliska@suse.cz>
7446 PR tree-optimization/82491
7447 * rtl.h (strip_offset_and_add): Replace += suboffset with
7448 poly_uint64 () + suboffset.
7450 2018-03-29 Martin Liska <mliska@suse.cz>
7451 Martin Jambor <mjambor@suse.cz>
7454 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7455 param_type is not an integral or pointer type.
7457 2018-04-03 Richard Biener <rguenther@suse.de>
7459 * sese.h (recompute_all_dominators): Remove.
7461 2018-04-02 Martin Sebor <msebor@redhat.com>
7463 * doc/invoke.texi (-Wrestrict): Fix typos.
7465 2018-04-02 Jim Wilson <jimw@sifive.com>
7467 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7468 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7469 (<optab>di3, <optab>si3_extend): Likewise.
7470 (<optab>si3_mask, <optab>si3_mask_1): New.
7471 (<optab>di3_mask, <optab>di3_mask_1): New.
7472 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7473 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7474 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7476 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7478 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7481 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7483 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7484 (nds32_canonicalize_comparison): New function.
7486 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7487 Kito Cheng <kito.cheng@gmail.com>
7488 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7490 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7491 * config/nds32/constants.md (unspec_volatile_element): Add
7492 UNSPEC_VOLATILE_RELAX_GROUP.
7493 * config/nds32/nds32-relax-opt.c: New file.
7494 * config/nds32/nds32-predicates.c
7495 (nds32_symbol_load_store_p): New function.
7496 * config/nds32/nds32-protos.h
7497 (nds32_symbol_load_store_p): Declare function.
7498 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7499 * config/nds32/nds32.c
7500 (nds32_register_pass): New function to register pass.
7501 (nds32_register_passes): New function to register passes.
7502 * config/nds32/nds32.md (relax_group): New pattern.
7503 * config/nds32/nds32.opt (mrelax-hint): New option.
7504 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7506 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7508 * config/nds32/t-nds32: Modify files dependency.
7510 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7512 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7513 (PROFILE_HOOK): Define its implementation.
7515 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7517 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7518 type and 32-bit size.
7520 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7523 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7524 (V_128_256): New mode iterator.
7525 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7526 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7527 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7529 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7530 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7532 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7535 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7536 NaN inputs correctly.
7538 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7541 * config/rs6000/vsx.md (??r): New mode attribute.
7542 (*vsx_mov<mode>_64bit): Use it.
7543 (*vsx_mov<mode>_32bit): Likewise.
7545 2018-03-30 Martin Sebor <msebor@redhat.com>
7547 PR tree-optimization/84818
7548 * builtins.c (check_access): Use warning_n.
7550 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7553 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7555 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7558 2018-03-30 Julia Koval <julia.koval@intel.com>
7561 * x86-tune.def (movx, partial_reg_dependency): Enable for
7564 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7567 * lra-constraints.c (process_alt_operands): Move setting
7568 this_alternative_matches below.
7570 2018-03-29 Martin Liska <mliska@suse.cz>
7573 * doc/invoke.texi: Document how LTO works with debug info.
7574 Describe auto-load support of binutils. Mention 'x86-64'
7575 as valid option value of -march option.
7577 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7579 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7582 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7583 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7584 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7587 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7590 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7591 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7592 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7593 for RS6000_BTM_POWERPC64.
7594 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7595 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7596 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7601 2018-03-28 Carl Love <cel@us.ibm.com>
7604 2017-09-27 Carl Love <cel@us.ibm.com>
7606 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7607 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7608 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7611 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7613 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7614 instead of __vector bool.
7615 (_mm_max_pu8): Likewise.
7616 (_mm_min_pi16): Likewise.
7618 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7621 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7622 (DIVWEUO): Likewise.
7624 (DIVDEUO): Likewise.
7625 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7626 DIVWEUO and DIVDEUO.
7627 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7628 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7629 (div_extend): Likewise.
7630 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7632 (__builtin_divweuo): Likewise.
7633 (__builtin_divdeo): Likewise.
7634 (__builtin_divdeuo): Likewise.
7636 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7639 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7640 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7642 PR tree-optimization/82004
7643 * gimple-match-head.c (optimize_pow_to_exp): New function.
7644 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7645 Don't fold to exp if optimize_pow_to_exp is false.
7647 2018-03-28 Martin Liska <mliska@suse.cz>
7650 * calls.c (initialize_argument_information): Fix trailing space.
7651 * common.opt: Fix typo and provide better explanation for
7652 -fsanitize-coverage option.
7653 * config/i386/i386.opt: Fix typo.
7655 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7656 Martin Liska <mliska@suse.cz>
7659 * gimplify.c (asan_poison_variable): Don't do the check for
7660 gimplify_omp_ctxp here.
7661 (gimplify_decl_expr): Do it here.
7662 (gimplify_target_expr): Likewise.
7664 2018-03-28 Martin Liska <mliska@suse.cz>
7667 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7668 chkp_type_bounds_count if MPX is not enabled.
7670 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7672 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7674 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7677 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7678 function to create the function decl for complex long double
7679 multiply and divide for -mabi=ieeelongdouble.
7680 (init_float128_ieee): Call it.
7682 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7685 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7686 -fcf-protection=branch -mibt.
7687 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7689 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7692 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7693 arm_word_relocations.
7695 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7698 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7699 extern array declarations.
7701 2018-03-27 Richard Biener <rguenther@suse.de>
7704 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7705 explicit single_use checks.
7707 2018-03-27 Richard Biener <rguenther@suse.de>
7709 PR tree-optimization/85082
7710 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7713 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7715 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7716 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7717 Turn on fasynchronous-unwind-tables and funwind-tables.
7719 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7722 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7723 (*bmi_blsr_<mode>_ccz): Ditto.
7725 2018-03-26 Tom de Vries <tom@codesourcery.com>
7727 PR tree-optimization/85063
7728 * omp-general.c (offloading_function_p): New function. Factor out
7730 * omp-offload.c (pass_omp_target_link::gate): ... here.
7731 * omp-general.h (offloading_function_p): Declare.
7732 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7733 with attribute omp declare target for offloading functions.
7735 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7737 PR tree-optimization/84005
7738 * tree-data-ref.h (get_base_for_alignment): Declare.
7739 * tree-data-ref.c (get_base_for_alignment_1): New function.
7740 (get_base_for_alignment): Likewise.
7741 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7742 get_base_for_alignment to find a suitable base object, instead
7743 of always using drb->base_address.
7745 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7748 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7749 known size by default.
7751 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7754 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7755 and non BLKmode operands.
7757 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7760 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7761 Clean up attributes.
7763 2018-03-23 Richard Biener <rguenther@suse.de>
7766 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7767 we are going to emit early debug for LTO.
7769 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7772 * function.c (match_asm_constraints_1): Don't optimize if input
7773 doesn't satisfy general_operand predicate for output's mode.
7776 * alias.c (write_dependence_p): Don't require for x_canonicalized
7777 non-VOIDmode if x has VOIDmode.
7780 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7781 just don't try to optimize it rather than assert it never happens.
7783 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7785 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7786 macro expansions for definition of ST_INTERNAL_<mode> and
7787 LD_INTERNAL_<mode> builtins.
7788 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7790 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7792 (altivec_expand_st_builtin): Likewise.
7793 (altivec_expand_builtin): Remove calls to deleted functions.
7794 (rs6000_address_for_altivec): Delete this function.
7795 * config/rs6000/vector.md: Remove expands for
7796 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7798 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7801 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7802 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7803 re-computing once computed.
7804 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7805 (arm_init_machine_status): Initialize
7806 machine->static_chain_stack_bytes.
7808 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7811 * doc/extend.texi: Add four new prototypes for vec_ld.
7812 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7813 definitions for more logical presentation.
7814 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7815 entries for V1TI variants of __builtin_altivec_ld builtin.
7816 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7817 handling of V1TI variant of LVX icode pattern.
7818 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7819 (rs6000_gimple_fold_builtin): Likewise.
7820 (altivec_init_builtins): Add code to define
7821 __builtin_altivec_lvx_v1ti function.
7823 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7826 * function.c (match_asm_constraints_1): Don't do the optimization
7827 if input isn't a REG, SUBREG, MEM or constant.
7829 2018-03-22 Tom de Vries <tom@codesourcery.com>
7831 PR tree-optimization/84956
7832 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7833 bb_has_abnormal_pred.
7835 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7838 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7839 DECL_INITIAL (decl) to decl at the end.
7840 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7843 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7845 * doc/extend.texi (__builtin_tgmath): Document when complex
7846 integer types are treated as _Complex _Float64.
7848 2018-03-21 Tom de Vries <tom@codesourcery.com>
7850 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7852 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7854 PR tree-optimization/84960
7855 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7856 if it is ENTRY block, move them into single succ of ENTRY in that case.
7858 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7860 PR tree-optimization/84811
7861 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7862 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7863 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7864 (known_subrange_p): Update accordingly. Cast each value involved
7865 in the size comparison, rather than casting the result of the
7868 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7870 PR tree-optimization/84982
7871 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7872 by flipping the least significant bit rather than all bits from
7873 bitpos to bitpos + bitsize - 1.
7875 2018-03-21 Nathan Sidwell <nathan@acm.org>
7877 * doc/extend.texi (Deprecated Features): Remove mention of
7878 long-deleted deprecations.
7880 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7883 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7884 * configure: Regenerate.
7886 2018-03-21 Tom de Vries <tom@codesourcery.com>
7888 PR tree-optimization/83126
7889 * tree-parloops.c (num_phis): New function.
7890 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7892 2018-03-21 Nathan Sidwell <nathan@acm.org>
7894 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7895 mention anon-struct/union members and trailing attributes.
7897 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7899 PR tree-optimization/84969
7900 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7901 builtin memset partitions if they set different rhs values.
7903 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7905 PR rtl-optimization/84989
7906 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7907 VEC_DUPLICATE with scalar result mode.
7909 2018-03-21 Martin Liska <mliska@suse.cz>
7912 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7913 not intended return statement.
7915 2018-03-21 Martin Liska <mliska@suse.cz>
7918 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7919 (chkp_find_bound_slots_1): Limit number of iterations.
7921 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7924 * Minor grammar fixes for x86 options.
7926 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7929 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7930 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7932 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7935 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7936 (altivec_lvx_<mode>_1op): Likewise.
7937 (altivec_stvx_<mode>_2op): Likewise.
7938 (altivec_stvx_<mode>_1op): Likewise.
7939 (altivec_lvx_<VM2:mode>): New define_expand.
7940 (altivec_stvx_<VM2:mode>): Likewise.
7941 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7942 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7943 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7944 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7945 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7946 (rs6000_gen_lvx): Likewise.
7947 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7948 (altivec_expand_stv_builtin): Likewise.
7949 (altivec_expand_builtin): Likewise.
7950 * config/rs6000/vector.md: Likewise.
7952 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7955 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7958 2018-03-20 Richard Biener <rguenther@suse.de>
7961 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7962 sign-conversions as zero, fall back to standard scalar_stmt
7965 2018-03-20 Martin Liska <mliska@suse.cz>
7968 * predict.c (rebuild_frequencies): Handle case when we have
7969 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7971 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7974 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7975 flag_section_anchors.
7976 * varasm.c (use_blocks_for_decl_p): Remove hack for
7977 dw2_force_const_mem.
7980 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7982 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7983 be created, use lowpart_subreg of operands[0] rather than operands[0]
7985 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7986 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7987 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7988 and n constraint instead of aarch64_shift_imm_di and Usd.
7989 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7990 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7992 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7995 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7996 to favor GPR over NEON registers.
7997 (<shift>di3_neon): Likewise.
7999 2018-03-20 Tom de Vries <tom@codesourcery.com>
8002 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8003 (nvptx_process_pars): Emit bar.sync asap and alap.
8005 2018-03-20 Tom de Vries <tom@codesourcery.com>
8008 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8009 seen_label if seen_label is already set.
8011 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8014 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8015 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8016 Use 1U instead of 1. Formatting fixes.
8019 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8020 instead of TREE_TYPE (s1) for the return value.
8022 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8024 PR tree-optimization/84946
8025 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8026 bitsize + bitsize in poly_uint64 rather than poly_int64.
8029 * dwarf2asm.c: Include fold-const.c.
8030 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8031 of decl rather than decl itself.
8033 PR rtl-optimization/84643
8034 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8036 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8039 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8040 calling assemble_variable.
8042 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8045 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8046 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8048 2018-03-19 Jim Wilson <jimw@sifive.com>
8051 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8052 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8053 (riscv_first_stack_step): Likewise.
8054 (riscv_option_override): Use STACK_BOUNDARY instead of
8056 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8058 (BIGGEST_ALIGNMENT): Set to 128.
8059 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8060 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8063 2018-03-19 Richard Biener <rguenther@suse.de>
8065 PR tree-optimization/84933
8066 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8067 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8069 2018-03-19 Richard Biener <rguenther@suse.de>
8071 PR tree-optimization/84859
8072 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8073 (cond_if_else_store_replacement): Perform sinking operation on
8074 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8075 Generalize what a BB with a single eligible store is.
8077 2018-03-19 Richard Biener <rguenther@suse.de>
8079 PR tree-optimization/84929
8080 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8081 chrec_is_positive against non-chrec arg.
8083 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8086 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8088 2018-03-18 Martin Liska <mliska@suse.cz>
8090 PR rtl-optimization/84635
8091 * regrename.c (build_def_use): Use matches_mode only when
8094 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8096 PR tree-optimization/84913
8097 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8098 vectorize chains of COND_EXPRs.
8100 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8102 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8104 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8106 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8108 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8110 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8112 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8113 Kito Cheng <kito.cheng@gmail.com>
8115 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8116 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8117 (nds32_adjust_reg_alloc_order): New function.
8118 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8120 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8122 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8123 nds32_print_operand, nds32_print_operand_address): Use
8124 HOST_WIDE_INT_PRINT_DEC instead.
8126 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8128 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8130 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8133 * config/i386/i386.c (initial_ix86_tune_features,
8134 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8136 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8137 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8138 rather than 1u << ix86_tune. Formatting fix.
8139 (ix86_option_override_internal): Change ix86_arch_mask from
8140 unsigned int to unsigned HOST_WIDE_INT, initialize to
8141 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8142 (ix86_function_specific_restore): Likewise.
8144 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8147 * postreload.c (reload_combine_recognize_pattern): Perform
8148 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8149 truncate_int_for_mode the result for the destination's mode.
8152 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8153 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8155 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8158 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8161 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8162 regno_allocno_class_array and sorted_pseudos.
8163 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8164 insns where regno is used.
8166 2018-03-16 Martin Liska <mliska@suse.cz>
8169 * multiple_target.c (create_dispatcher_calls): Redirect
8170 reference in the symbol table.
8172 2018-03-16 Martin Liska <mliska@suse.cz>
8175 * multiple_target.c (create_dispatcher_calls): Redirect also
8178 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8182 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8183 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8184 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8186 2018-03-16 Julia Koval <julia.koval@intel.com>
8188 * doc/invoke.texi (Skylake Server): Add CLWB.
8189 Cannonlake): Remove CLWB.
8191 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8193 PR tree-optimization/84841
8194 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8196 (FLOAT_ONE_CONST_TYPE): Define.
8197 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8198 (sort_by_operand_rank): Put entries with higher constant_type last
8199 rather than first to match comments.
8201 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8203 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8206 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8209 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8210 check and use address of target always.
8212 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8215 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8216 (indirect_thunk_bnd_needed): Likewise.
8217 (indirect_thunks_used): Likewise.
8218 (indirect_thunks_bnd_used): Likewise.
8219 (indirect_return_needed): New.
8220 (indirect_return_bnd_needed): Likewise.
8221 (output_indirect_thunk_function): Add a bool argument for
8223 (output_indirect_thunk_function): Don't generate alias for
8224 function return thunk.
8225 (ix86_code_end): Call output_indirect_thunk_function to generate
8226 function return thunks.
8227 (ix86_output_function_return): Set indirect_return_bnd_needed
8228 and indirect_return_needed instead of indirect_thunk_bnd_needed
8229 and indirect_thunk_needed.
8231 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8233 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8234 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8235 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8237 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8238 Paul Hua <paul.hua.gm@gmail.com>
8241 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8243 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8245 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8246 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8249 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8252 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8253 instead of GET_MODE_SIZE when comparing Units.
8255 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8258 * varasm.c (hash_section): Return an unchangeble hash value
8259 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8260 Return !aarch64_can_use_per_function_literal_pools_p ().
8262 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8265 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8266 rather than address of cmode as last argument to prepare_cmp_insn.
8268 2018-03-15 Julia Koval <julia.koval@intel.com>
8270 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8271 F_AVX512VNNI, F_AVX512BITALG): New.
8273 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8276 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8277 insn for floating-point loads and stores.
8279 2018-03-14 Carl Love <cel@us.ibm.com>
8281 * config/rs6000/rs6000-c.c: Add macro definitions for
8282 ALTIVEC_BUILTIN_VEC_PERMXOR.
8283 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8284 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8285 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8286 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8288 * config/doc/extend.texi: Add prototypes for vec_permxor.
8290 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8293 * diagnostic-show-locus.c (class layout_point): Convert m_line
8294 from int to linenum_type.
8295 (line_span::comparator): Use linenum "compare" function when
8296 comparing line numbers.
8297 (test_line_span): New function.
8298 (layout_range::contains_point): Convert param "row" from int to
8300 (layout_range::intersects_line_p): Likewise.
8301 (layout::will_show_line_p): Likewise.
8302 (layout::print_source_line): Likewise.
8303 (layout::should_print_annotation_line_p): Likewise.
8304 (layout::print_annotation_line): Likewise.
8305 (layout::print_leading_fixits): Likewise.
8306 (layout::annotation_line_showed_range_p): Likewise.
8307 (struct line_corrections): Likewise for field m_row.
8308 (line_corrections::line_corrections): Likewise for param "row".
8309 (layout::print_trailing_fixits): Likewise.
8310 (layout::get_state_at_point): Likewise.
8311 (layout::get_x_bound_for_row): Likewise.
8312 (layout::print_line): Likewise.
8313 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8314 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8315 * input.c (selftest::test_linenum_comparisons): New function.
8316 (selftest::input_c_tests): Call it.
8317 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8318 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8319 * selftest.h (ASSERT_GT): New macro.
8320 (ASSERT_GT_AT): New macro.
8321 (ASSERT_LT): New macro.
8322 (ASSERT_LT_AT): New macro.
8324 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8326 PR rtl-optimization/84780
8327 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8329 2018-03-14 Martin Liska <mliska@suse.cz>
8331 * tree.c (record_node_allocation_statistics): Use
8332 get_stats_node_kind.
8333 (get_stats_node_kind): New function extracted from
8334 record_node_allocation_statistics.
8335 (free_node): Use get_stats_node_kind.
8337 2018-03-14 Richard Biener <rguenther@suse.de>
8339 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8340 that the value-set of ANTIC_IN doesn't grow.
8343 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8345 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8346 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8348 2018-03-14 Julia Koval <julia.koval@intel.com>
8350 * config.gcc (icelake-client, icelake-server): New.
8352 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8353 (initial_ix86_arch_features): Ditto.
8354 (PTA_SKYLAKE): Add SGX.
8355 (PTA_ICELAKE): Remove.
8356 (PTA_ICELAKE_CLIENT): New.
8357 (PTA_ICELAKE_SERVER): New.
8358 (ix86_option_override_internal): Split up icelake on icelake client and
8360 (get_builtin_code_for_version): Ditto.
8361 (fold_builtin_cpu): Ditto.
8362 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8363 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8364 * config/i386/i386.h (processor_type): Ditto.
8365 * doc/invoke.texi: Ditto.
8367 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8370 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8371 INTEGER_CST offset, add it together with bitpos / 8 and
8372 sign extend based on POINTER_SIZE.
8376 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8379 * config/i386/constraints.md (Yc): New register constraint.
8380 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8381 Use Yc constraint for alternative 2 of operand 0. Remove
8382 preferred_for_speed attribute.
8384 2018-03-14 Richard Biener <rguenther@suse.de>
8386 PR tree-optimization/84830
8387 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8388 with the old one to avoid oscillations.
8390 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8393 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8395 (assign_by_spills): Return a flag of reload assignment failure.
8396 Do not process the reload assignment failures. Do not spill other
8397 reload pseudos if they has the same reg class. Update n if
8399 (lra_assign): Add a return arg. Set up from the result of
8400 assign_by_spills call.
8401 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8402 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8403 usage_insns if it is not NULL.
8404 (spill_hard_reg_in_range): New function.
8405 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8406 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8407 function prototypes.
8408 (lra_assign): Change prototype.
8409 * lra.c (lra): Add code to deal with fails by splitting hard reg
8412 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8414 * config/riscv/riscv.opt (mrelax): New option.
8415 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8416 "norelax" when riscv_mrelax is disabled.
8417 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8419 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8422 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8423 reassociation for int modes.
8425 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8427 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8428 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8430 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8431 * config/aarch64/aarch64-sve.md
8432 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8433 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8434 (*extend<mode><Vwide>2): Rename to...
8435 (aarch64_sve_extend<mode><Vwide>2): ...this.
8436 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8437 renaming the old pattern to...
8438 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8440 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8441 define_expand, renaming the old pattern to...
8442 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8443 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8444 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8445 account when deciding which SVE instruction the optab should use.
8446 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8448 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8450 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8451 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8452 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8453 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8454 (tlsdesc_small_<mode>): Turn a define_expand and use
8455 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8456 (tlsdesc_small_advsimd_<mode>): ...this.
8457 (tlsdesc_small_sve_<mode>): New pattern.
8459 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8461 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8462 (UNSPEC_UMUL_HIGHPART): New constants.
8463 (MUL_HIGHPART): New int iteraor.
8464 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8465 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8467 (*<su>mul<mode>3_highpart): New define_insn.
8469 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8472 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8475 2018-03-13 Martin Liska <mliska@suse.cz>
8478 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8480 (sem_item_optimizer::~sem_item_optimizer): Release it.
8481 (sem_item_optimizer::merge_classes): Register variable aliases.
8482 (sem_item_optimizer::fixup_pt_set): New function.
8483 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8484 * ipa-icf.h: Declare new variables and functions.
8486 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8489 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8490 integer_pow2p@2 and test integer_pow2p in condition.
8491 (A < 0 ? C : 0): Similarly for @1.
8494 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8495 characters starting at p contain '\0' character, don't look beyond
8499 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8500 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8503 * reg-stack.c (change_stack): Change update_end var from int to
8504 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8505 also call set_block_for_insn on the newly added insns and rescan.
8508 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8509 on the last operand.
8512 * tree.c (stabilize_reference_1): Return save_expr (e) for
8513 STATEMENT_LIST even if it doesn't have side-effects.
8515 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8517 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8519 2018-03-12 Renlin Li <renlin.li@arm.com>
8521 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8522 aarch64_output_scalar_simd_mov_immediate.
8524 2018-03-12 Martin Sebor <msebor@redhat.com>
8526 PR tree-optimization/83456
8527 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8528 for perfectly overlapping calls to memcpy.
8529 (gimple_fold_builtin_memory_chk): Same.
8530 (gimple_fold_builtin_strcpy): Handle no-warning.
8531 (gimple_fold_builtin_stxcpy_chk): Same.
8532 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8534 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8536 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8537 parameter. Use it for SFmode.
8538 (rs6000_function_arg_advance_1): Adjust.
8539 (rs6000_function_arg): Adjust.
8540 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8542 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8544 PR rtl-optimization/84169
8545 PR rtl-optimization/84780
8546 * combine.c (can_combine_p): Check for a 2-insn combination whether
8547 the destination register is used between the two insns, too.
8549 2018-03-12 Richard Biener <rguenther@suse.de>
8551 PR tree-optimization/84803
8552 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8553 for refs DR analysis didn't process.
8555 2018-03-12 Richard Biener <rguenther@suse.de>
8557 PR tree-optimization/84777
8558 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8559 force-vectorize loops ignore whether we are optimizing for size.
8561 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8563 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8564 (TARGET_MD_ASM_ADJUST): Define.
8566 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8567 Kito Cheng <kito.cheng@gmail.com>
8568 Chung-Ju Wu <jasonwucj@gmail.com>
8570 * config/nds32/nds32.c (nds32_compute_stack_frame,
8571 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8572 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8573 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8574 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8575 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8576 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8577 * config/nds32/nds32.md (prologue, epilogue): Use macro
8578 NDS32_V3PUSH_AVAILABLE_P to do checking.
8580 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8583 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8584 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8585 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8586 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8587 addition of most attributes on !orig_type_die or the attribute not
8588 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8590 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8591 Chung-Ju Wu <jasonwucj@gmail.com>
8593 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8595 * config/nds32/nds32.opt (mvh): New option.
8597 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8598 Chung-Ju Wu <jasonwucj@gmail.com>
8600 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8602 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8603 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8606 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8607 Chung-Ju Wu <jasonwucj@gmail.com>
8609 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8611 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8612 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8614 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8615 Kito Cheng <kito.cheng@gmail.com>
8616 Chung-Ju Wu <jasonwucj@gmail.com>
8618 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8619 UNSPEC_FFMISM and UNSPEC_FLMISM.
8620 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8621 for ffb, ffmism and flmism.
8622 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8623 (unspec_ffmism): Ditto.
8624 (unspec_flmism): Ditto.
8625 (nds32_expand_builtin_impl): Check if string extension is available.
8626 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8627 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8629 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8632 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8635 * lra-assigns.c (assign_by_spills): Return a flag of reload
8636 assignment failure. Do not process the reload assignment
8637 failures. Do not spill other reload pseudos if they has the same
8639 (lra_assign): Add a return arg. Set up from the result of
8640 assign_by_spills call.
8641 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8642 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8643 usage_insns if it is not NULL.
8644 (spill_hard_reg_in_range): New function.
8645 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8646 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8647 function prototypes.
8648 (lra_assign): Change prototype.
8649 * lra.c (lra): Add code to deal with fails by splitting hard reg
8652 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8655 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8657 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8660 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8661 (dw_val_node): Add val_symbolic_view.
8662 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8663 (symview_upper_bound): New.
8664 (new_line_info_table): Initialize symviews_since_reset.
8665 (dwarf2out_source_line): Count symviews_since_reset and set
8666 symview_upper_bound.
8667 (dw_val_equal_p): Handle symview.
8668 (add_AT_symview): New.
8669 (print_dw_val): Handle symview.
8670 (attr_checksum, attr_checksum_ordered): Likewise.
8671 (same_dw_val_p, size_of_die): Likewise.
8672 (value_format, output_die): Likewise.
8673 (add_high_low_attributes): Use add_AT_symview for entry_view.
8674 (dwarf2out_finish): Reset symview_upper_bound, clear
8677 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8680 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8681 Add strict argument and use it.
8682 (rs6000_split_multireg_move): Update for new strict argument.
8683 (mem_operand_gpr): Disallow all non-offsettable addresses.
8684 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8686 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8689 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8690 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8691 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8694 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8695 decl, use remap_type if we want to use the type.
8697 2018-03-09 Martin Sebor <msebor@redhat.com>
8699 PR tree-optimization/84526
8700 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8702 (builtin_access::generic_overlap): Be prepared to handle non-array
8705 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8707 PR rtl-optimization/84682
8708 * lra-constraints.c (process_address_1): Check is_address flag
8709 for address constraints.
8710 (process_alt_operands): Likewise.
8711 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8712 preprocess_constraints.
8713 * recog.h (preprocess_constraints): Add oploc parameter.
8715 * recog.c (preprocess_constraints): Test address_operand for
8716 CT_ADDRESS constraints.
8718 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8721 * lra-assigns.c (assign_by_spills): Return a flag of reload
8722 assignment failure. Do not process the reload assignment
8723 failures. Do not spill other reload pseudos if they has the same
8725 (lra_assign): Add a return arg. Set up from the result of
8726 assign_by_spills call.
8727 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8728 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8729 usage_insns if it is not NULL.
8730 (spill_hard_reg_in_range): New function.
8731 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8732 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8733 function prototypes.
8734 (lra_assign): Change prototype.
8735 * lra.c (lra): Add code to deal with fails by splitting hard reg
8738 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8741 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8742 Accept complain bool parameter. Only emit errors if it is true.
8743 (arm_parse_cpu_option_name): Likewise.
8744 (arm_target_thumb_only): Adjust callers of the above.
8745 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8746 prototype to take a default true bool parameter.
8747 (arm_parse_arch_option_name): Likewise.
8749 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8750 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8754 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8755 * configure: Regenerate.
8756 * configure.ac ("linker --version-script option"): New.
8757 ("linker soname option"): New.
8759 2018-03-09 Richard Biener <rguenther@suse.de>
8761 PR tree-optimization/84775
8762 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8763 immediate uses of predicate stmts and mark them modified.
8766 PR tree-optimization/84178
8767 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8769 (version_loop_for_if_conversion): Delay update_ssa call.
8770 (tree_if_conversion): Delay update_ssa until after predicate
8773 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8776 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8777 when the function accesses prior frames.
8779 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8782 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8783 gen_llsym, otherwise call maybe_gen_llsym.
8786 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8787 has ',' character inside of it.
8789 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8792 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8793 as clobbering CC_REGNUM.
8795 2018-03-08 Richard Biener <rguenther@suse.de>
8798 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8799 (follow_copies_to_constant): Do not follow SSA names registered
8802 2018-03-08 Richard Biener <rguenther@suse.de>
8804 PR tree-optimization/84178
8805 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8807 (version_loop_for_if_conversion): Delay update_ssa call.
8808 (tree_if_conversion): Delay update_ssa until after predicate
8811 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8813 PR tree-optimization/84178
8814 * tree-if-conv.c (release_bb_predicate): Remove the
8815 the assertion that the stmts have NULL use_ops.
8816 Discard the statements, asserting that they haven't
8817 yet been added to a BB.
8819 2018-03-08 Richard Biener <rguenther@suse.de>
8821 PR tree-optimization/84746
8822 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8823 (phi_translate): Pass in destination ANTIC_OUT set.
8824 (phi_translate_1): Likewise. For a simplified result lookup
8825 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8826 (phi_translate_set): Adjust.
8827 (do_pre_regular_insertion): Likewise.
8828 (do_pre_partial_partial_insertion): Likewise.
8830 2018-03-08 Martin Liska <mliska@suse.cz>
8832 PR gcov-profile/84735
8833 * doc/gcov.texi: Document usage of profile files.
8834 * gcov-io.h: Document changes in the format.
8836 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8840 * dwarf2out.c (struct dw_line_info_table): Update comments for
8842 (FORCE_RESET_NEXT_VIEW): New.
8843 (FORCE_RESETTING_VIEW_P): New.
8844 (RESETTING_VIEW_P): Check for -1 too.
8845 (ZERO_VIEW_P): Likewise.
8846 (new_line_info_table): Force-reset next view.
8847 (dwarf2out_begin_function): Likewise.
8848 (dwarf2out_source_line): Simplify zero_view_p initialization.
8849 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8850 view directly. Omit view when omitting .loc at line 0.
8852 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8854 PR tree-optimization/84740
8855 * tree-switch-conversion.c (process_switch): Call build_constructors
8856 only if info.phi_count is non-zero.
8858 PR tree-optimization/84739
8859 * tree-tailcall.c (find_tail_calls): Check call arguments against
8860 DECL_ARGUMENTS (current_function_decl) rather than
8861 DECL_ARGUMENTS (func) when checking for tail recursion.
8863 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8865 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8866 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8867 Volker Reichelt's entry and add entries for people that perform
8868 GCC fuzzy testing and report numerous bugs.
8870 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8873 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8874 readonly data in sdata, if that is disabled.
8875 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8876 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8877 -mreadonly-in-sdata option.
8879 2018-03-07 Martin Sebor <msebor@redhat.com>
8881 PR tree-optimization/84468
8882 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8883 basic block when looking for nul assignment.
8885 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8888 * except.h (output_function_exception_table): Adjust prototype.
8889 * except.c (output_function_exception_table): Remove FNNAME parameter
8890 and add SECTION parameter. Ouput one part of the table at a time.
8891 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8892 the first part of the exception table and emit unwind directives.
8893 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8894 (i386_pe_seh_cold_init): Likewise.
8895 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8896 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8897 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8898 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8899 * config/i386/winnt.c: Include except.h.
8900 (struct seh_frame_state): Add reg_offset, after_prologue and
8901 in_cold_section fields.
8902 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8903 (i386_pe_seh_cold_init): New function.
8904 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8905 to seh->in_cold_section.
8906 (seh_emit_push): Record the offset of the push.
8907 (seh_emit_save): Record the offet of the save.
8908 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8909 Test seh->after_prologue to disregard the epilogue.
8910 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8911 (i386_pe_end_cold_function): New function.
8913 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8916 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8917 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8920 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8922 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8923 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8926 * multiple_target.c: Include tree-inline.h and intl.h.
8927 (expand_target_clones): Diagnose and fail if node->definition and
8928 !tree_versionable_function_p (node->decl).
8930 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8932 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8934 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8935 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8936 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8938 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8941 * combine.c (try_combine): Use reg_or_subregno instead of handling
8942 just paradoxical SUBREGs and REGs.
8944 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8946 * config/arc/arc.c (arc_finalize_pic): Remove function.
8947 (arc_must_save_register): We use single base PIC register, remove
8948 checks to save/restore the PIC register.
8949 (arc_expand_prologue): Likewise.
8950 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8952 (arc_verify_short): Likewise.
8953 (arc_attr_type): Likewise.
8954 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8955 (walk_stores): Likewise.
8956 (arc_address_cost): Make it static.
8957 (arc_verify_short): Likewise.
8958 (branch_dest): Likewise.
8959 (arc_attr_type): Likewise.
8960 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8961 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8962 (arc_final_prescan_insn): Remove inserting the nops due to
8963 hardware hazards. It is done in reorg step.
8964 (insn_length_variant_t): Remove.
8965 (insn_length_parameters_t): Likewise.
8966 (arc_insn_length_parameters): Likewise.
8967 (arc_get_insn_variants): Likewise.
8968 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8970 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8973 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8976 PR tree-optimization/84687
8977 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8979 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8981 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8983 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8984 Rename to ppc_speculation_barrier.
8985 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8986 __builtin_ppc_speculation_barrier.
8988 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8991 * combine.c (combine_simplify_rtx): Don't try to simplify if
8992 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8995 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8997 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8998 to 32 bytes when compiling for POWER9.
9000 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9003 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9004 regparm >= 3 with no arg reg available also for calls with
9005 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9008 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9010 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9012 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9015 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9017 2018-03-05 Richard Biener <rguenther@suse.de>
9019 PR tree-optimization/84486
9020 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9021 When inserting a __builtin_assume_aligned call set the LHS
9022 SSA name alignment info accordingly.
9024 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9026 PR tree-optimization/84114
9027 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9028 Avoid reassociation of FLOAT_MODE addition.
9030 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9032 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9033 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9034 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9035 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9036 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9037 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9038 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9040 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9041 __builtin_ia32_wbinvd): New builtins.
9042 (SPECIAL_ARGS2): New.
9043 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9044 (SPECIAL_ARGS2): New.
9045 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9046 (ix86_valid_target_attribute_inner_p): Ditto.
9047 (ix86_init_mmx_sse_builtins): Add special_args2.
9048 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9049 TARGET_WBNOINVD_P): New.
9050 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9051 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9052 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9053 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9054 * config/i386/pconfigintrin.h: New file.
9055 * config/i386/wbnoinvdintrin.h: Ditto.
9056 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9058 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9060 2018-03-05 Richard Biener <rguenther@suse.de>
9062 PR tree-optimization/84670
9063 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9065 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9066 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9067 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9068 doesn't grow if all successors (recursively) were visited at least
9071 2018-03-05 Richard Biener <rguenther@suse.de>
9073 PR tree-optimization/84650
9074 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9075 if executed in the loop pipeline.
9077 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9079 * doc/configfiles.texi (Configuration Files): Move info about
9080 conditionalizing $target-protos.h to...
9081 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9082 differs from $target-protos.h.
9084 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9085 Chung-Ju Wu <jasonwucj@gmail.com>
9087 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9088 * config/nds32/nds32-multiple.md (setmemsi): Define.
9089 * config/nds32/nds32-memory-manipulation.c
9090 (nds32_gen_dup_4_byte_to_word_value): New.
9091 (emit_setmem_word_loop): New.
9092 (emit_setmem_byte_loop): New.
9093 (nds32_expand_setmem_loop): New.
9094 (nds32_expand_setmem_loop_v3m): New.
9095 (nds32_expand_setmem_unroll): New.
9096 (nds32_expand_setmem): New.
9098 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9099 Chung-Ju Wu <jasonwucj@gmail.com>
9101 * config/nds32/nds32-memory-manipulation.c
9102 (nds32_emit_load_store): New.
9103 (nds32_emit_post_inc_load_store): New.
9104 (nds32_emit_mem_move): New.
9105 (nds32_emit_mem_move_block): New.
9106 (nds32_expand_movmemsi_loop_unknown_size): New.
9107 (nds32_expand_movmemsi_loop_known_size): New.
9108 (nds32_expand_movmemsi_loop): New.
9109 (nds32_expand_movmemsi_unroll): New.
9110 (nds32_expand_movmemqi): Rename ...
9111 (nds32_expand_movmemsi): ... to this.
9112 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9113 (movmemsi): ... to this.
9114 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9115 (nds32_expand_movmemsi): ... to this.
9117 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9118 Monk Chiang <sh.chiang04@gmail.com>
9119 Chung-Ju Wu <jasonwucj@gmail.com>
9121 * config/nds32/nds32-protos.h
9122 (nds32_expand_load_multiple): New arguments.
9123 (nds32_expand_store_multiple): Ditto.
9124 (nds32_valid_multiple_load_store): Rename ...
9125 (nds32_valid_multiple_load_store_p): ... to this.
9126 * config/nds32/nds32-memory-manipulation.c
9127 (nds32_expand_load_multiple): Refine implementation.
9128 (nds32_expand_store_multiple): Ditto.
9129 * config/nds32/nds32-multiple.md
9130 (load_multiple): Update nds32_expand_load_multiple interface.
9131 (store_multiple): Update nds32_expand_store_multiple interface.
9132 * config/nds32/nds32-predicates.c
9133 (nds32_valid_multiple_load_store): Rename ...
9134 (nds32_valid_multiple_load_store_p): ... to this and refine
9136 * config/nds32/predicates.md
9137 (nds32_load_multiple_and_update_address_operation): New predicate.
9138 (nds32_store_multiple_and_update_address_operation): New predicate.
9140 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9141 Chung-Ju Wu <jasonwucj@gmail.com>
9143 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9144 (combo): New attribute.
9145 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9147 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9149 * config/nds32/nds32.opt: Change -mcmodel= default value.
9151 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9152 Monk Chiang <sh.chiang04@gmail.com>
9153 Chung-Ju Wu <jasonwucj@gmail.com>
9155 * config/nds32/constants.md (unspec_element): New enum.
9156 * config/nds32/constraints.md (Umw): New constraint.
9157 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9158 * config/nds32/nds32-intrinsic.md: Likewise.
9159 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9160 (nds32_valid_smw_lwm_base_p): New.
9161 (nds32_output_smw_single_word): New.
9162 (nds32_output_lmw_single_word): New.
9163 (nds32_expand_unaligned_load): New.
9164 (nds32_expand_unaligned_store): New.
9165 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9166 (nds32_output_smw_single_word): Declare.
9167 (nds32_output_lmw_single_word): Declare.
9168 (nds32_expand_unaligned_load): Declare.
9169 (nds32_expand_unaligned_store): Declare.
9170 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9171 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9172 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9173 NDS32_BUILTIN_UASTORE_DW.
9174 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9177 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9178 Kito Cheng <kito.cheng@gmail.com>
9179 Chung-Ju Wu <jasonwucj@gmail.com>
9181 * config/nds32/nds32-intrinsic.c
9182 (nds32_expand_builtin_null_ftype_reg): Delete.
9183 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9184 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9185 (nds32_read_argument): New.
9186 (nds32_legitimize_target): Ditto.
9187 (nds32_legitimize_argument): Ditto.
9188 (nds32_check_constant_argument): Ditto.
9189 (nds32_expand_unop_builtin): Ditto.
9190 (nds32_expand_unopimm_builtin): Ditto.
9191 (nds32_expand_binop_builtin): Ditto.
9192 (nds32_builtin_decl_impl): Ditto.
9193 (builtin_description): Ditto.
9194 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9195 (nds32_init_builtins_impl): Ditto.
9196 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9197 (nds32_builtin_decl): New.
9198 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9199 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9201 2018-03-02 Jeff Law <law@redhat.com>
9203 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9204 (redundant_insn, fill_simple_delay_slots): Likewise.
9205 (fill_slots_from_thread): Likewise.
9206 * resource.c (mark_referenced_resources): Likewise.
9207 (mark_set_resources, find_dead_or_set_registers): Likewise.
9209 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9211 * substring-locations.h (format_warning_va): Formatting fix for
9213 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9215 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9216 * substring-locations.c: Include intl.h.
9217 (format_warning_va): Turned into small wrapper around
9218 format_warning_n_va, renamed to ...
9219 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9220 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9222 (format_warning_at_substring_n): New function.
9223 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9224 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9225 format_warning_at_substring with just a shorter name instead of
9226 const function pointer.
9227 (fmtwarn_n): New function.
9228 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9229 appropriate, get rid of all the fmtstr temporaries, move conditionals
9230 with G_() wrapped string literals directly into fmtwarn arguments,
9231 cast dir.len to (int), formatting fixes.
9233 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9235 * doc/invoke.texi: Remove "Cilk Plus" references.
9237 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9238 Richard Biener <rguenther@suse.de>
9241 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9242 for error or warning attributes if CALL_FROM_THUNK_P is set.
9245 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9248 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9249 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9252 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9253 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9255 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9256 instead of -1U in last predictors element's probability member.
9258 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9261 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9262 arguments if they are comparable.
9264 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9266 PR tree-optimization/84634
9267 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9268 masks and masked_loop_p with a single loop_masks, making sure it's
9269 null for bb vectorization.
9271 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9273 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9274 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9275 loop->force_vectorize to check whether there is no alias.
9277 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9280 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9282 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9284 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9285 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9286 instead of a loop around prev_real_insn.
9287 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9291 * config/i386/i386.c (ix86_print_operand): Use conditional
9292 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9295 2018-03-02 Richard Biener <rguenther@suse.de>
9297 PR tree-optimization/84427
9298 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9299 (bitmap_set_subtract_values): Rewrite to handle multiple
9302 (prune_clobbered_mems): Likewise.
9303 (phi_translate): Take edge instead of pred/phiblock.
9304 (phi_translate_1): Likewise.
9305 (phi_translate_set): Likewise. Insert all translated
9306 exprs for a value into the set, keeping possibly multiple
9307 expressions per value.
9308 (compute_antic_aux): Adjust for phi_translate changes.
9309 When intersecting union the expressions and prune those
9310 not in the final value set, keeping possibly multiple
9311 expressions per value. Do not use value-insertion
9312 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9313 all expressions. Add verification that the value-sets
9314 only shrink during iteration.
9315 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9316 (do_pre_regular_insertion): Likewise.
9317 (do_pre_partial_partial_insertion): Likewise.
9319 2018-03-02 Richard Biener <rguenther@suse.de>
9322 * config/darwin.c (saved_debug_info_level): New static global.
9323 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9324 (darwin_asm_lto_end): Restore debug info generation settings.
9326 2018-03-01 Martin Liska <mliska@suse.cz>
9329 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9332 2018-03-01 Richard Biener <rguenther@suse.de>
9335 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9336 type annotation with LTO when debug was disabled at compile-time.
9338 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9340 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9342 (mips_final_postscan_insn): Likewise.
9344 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9346 PR rtl-optimization/84528
9347 * alias.c (init_alias_target): Add commentary.
9348 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9349 a unique base value if the frame pointer is not eliminated
9350 to the stack pointer.
9352 2018-03-01 Tom de Vries <tom@codesourcery.com>
9354 PR rtl-optimization/83327
9355 * lra-int.h (hard_regs_spilled_into): Declare.
9356 * lra.c (hard_regs_spilled_into): Define.
9357 (init_reg_info): Init hard_regs_spilled_into.
9358 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9359 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9360 (process_bb_lives): Handle hard_regs_spilled_into.
9361 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9362 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9364 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9366 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9367 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9368 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9369 * config/rs6000/aix72.h: New file.
9371 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9373 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9374 instead of warning_at with conditional singular and plural messages
9378 * stor-layout.c (update_alignment_for_field): For
9379 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9380 && !DECL_PACKED (field), do the alignment update, just use
9381 only desired_align instead of MAX (type_align, desired_align)
9383 (place_field): Don't do known_align < desired_align handling
9384 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9385 is non-NULL, instead do it after rli->prev_field handling and
9386 only if not within a bitfield word. For DECL_PACKED (field)
9387 use type_align of BITS_PER_UNIT.
9389 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9391 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9392 superfluous parentheses and trailing spaces.
9394 2018-02-28 Richard Biener <rguenther@suse.de>
9396 PR tree-optimization/84584
9397 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9398 SCoPs with fake exit edge.
9400 2018-02-28 Martin Liska <mliska@suse.cz>
9403 * timevar.c (timer::print): Fix format to properly print 100%
9406 2018-02-28 Richard Biener <rguenther@suse.de>
9409 * genmatch.c (capture_info::walk_match): Do not mark
9410 captured expressions without operands as expr_p given
9411 they act more like predicates and should be subject to
9412 "lost tail" side-effect preserving.
9414 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9416 PR rtl-optimization/81611
9417 * auto-inc-dec.c (attempt_change): Move dead note from
9418 mem_insn if it's the next use of regno
9419 (find_address): Take address use of reg holding
9420 non-incremented value. Add parm to limit search to the named
9422 (merge_in_block): Attempt to use a mem insn that is the next
9423 use of the original regno.
9425 2018-02-27 Martin Sebor <msebor@redhat.com>
9428 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9429 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9431 2018-02-27 Martin Sebor <msebor@redhat.com>
9433 PR translation/84207
9434 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9435 n argument to unsigned HOST_WIDE_INT.
9436 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9437 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9438 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9441 2018-02-27 Richard Biener <rguenther@suse.de>
9443 PR tree-optimization/84512
9444 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9445 Do not use the estimate returned from record_stmt_cost for
9446 the scalar iteration cost but sum properly using add_stmt_cost.
9448 2018-02-27 Richard Biener <rguenther@suse.de>
9450 PR tree-optimization/84466
9451 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9452 Adjust last change to less strictly validate use operands.
9454 2018-02-27 Martin Liska <mliska@suse.cz>
9456 PR gcov-profile/84548
9457 * gcov.c (process_file): Allow partial overlap and consider it
9458 also as group functions.
9459 (output_lines): Properly calculate range of lines for a group.
9461 2018-02-27 Martin Liska <mliska@suse.cz>
9463 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9464 'ggc' suffixes. Change first column width.
9465 (timer::print): Fix formatting of the column.
9467 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9469 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9470 preserve inline entry blocks for the sake of debug inline
9471 entry point markers alone.
9472 (remove_unused_locals): Suggest in comments a better place to
9473 force the preservation of inline entry blocks that are
9474 otherwise unused, but do not preserve them.
9476 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9478 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9480 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9483 * config/i386/constraints.md (Bs): Replace
9484 ix86_indirect_branch_register with
9485 TARGET_INDIRECT_BRANCH_REGISTER.
9487 * config/i386/i386.md (indirect_jump): Likewise.
9488 (tablejump): Likewise.
9489 (*sibcall_memory): Likewise.
9490 (*sibcall_value_memory): Likewise.
9491 Peepholes of indirect call and jump via memory: Likewise.
9492 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9493 (*sibcall_value_GOT_32): Likewise.
9494 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9495 (GOT_memory_operand): Likewise.
9496 (call_insn_operand): Likewise.
9497 (sibcall_insn_operand): Likewise.
9498 (GOT32_symbol_operand): Likewise.
9499 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9501 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9503 PR rtl-optimization/83496
9504 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9505 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9506 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9507 redundant insn, if any.
9508 (relax_delay_slots): Likewise.
9509 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9511 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9513 PR tree-optimization/83965
9514 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9515 that grouped statements are part of a reduction chain. Return
9516 true if the statement is not marked as a reduction itself but
9518 (vect_recog_dot_prod_pattern): Don't check whether the statement
9519 is part of a group here.
9520 (vect_recog_sad_pattern): Likewise.
9521 (vect_recog_widen_sum_pattern): Likewise.
9523 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9526 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9528 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9531 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9533 (ix86_output_indirect_function_return): New prototype.
9534 (ix86_split_simple_return_pop_internal): Likewise.
9535 * config/i386/i386.c (indirect_return_via_cx): New.
9536 (indirect_return_via_cx_bnd): Likewise.
9537 (indirect_thunk_name): Handle return va CX_REG.
9538 (output_indirect_thunk_function): Create alias for
9539 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9540 (ix86_output_indirect_jmp): Remove the bool argument.
9541 (ix86_output_indirect_function_return): New function.
9542 (ix86_split_simple_return_pop_internal): Likewise.
9543 * config/i386/i386.md (*indirect_jump): Don't pass false
9544 to ix86_output_indirect_jmp.
9545 (*tablejump_1): Likewise.
9546 (simple_return_pop_internal): Change it to define_insn_and_split.
9547 Call ix86_split_simple_return_pop_internal to split it for
9549 (simple_return_indirect_internal): Call
9550 ix86_output_indirect_function_return instead of
9551 ix86_output_indirect_jmp.
9553 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9556 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9557 memset and value initialization afterwards.
9559 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9561 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9563 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9566 * common/config/aarch64/aarch64-common.c
9567 (aarch_option_optimization_table[]): Switch
9568 off fomit-frame-pointer
9570 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9571 Chung-Ju Wu <jasonwucj@gmail.com>
9573 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9575 (store_multiple): Ditto.
9577 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9579 * config.gcc: Add --with-cpu support for nds32 target.
9580 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9581 * config/nds32/nds32.opt: Add -mcpu= option.
9583 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9585 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9586 isel=yes): Warn for these deprecated options.
9588 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9590 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9591 ISA_2_5_MASKS_EMBEDDED.
9593 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9595 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9596 p->max as pointers rather than using iterative_hash_expr.
9598 2018-02-23 Carl Love <cel@us.ibm.com>
9600 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9601 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9603 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9604 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9605 P8V_BUILTIN_VEC_VUNSIGNED2.
9607 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9610 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9611 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9613 * lra-constraints.c (curr_insn_transform): Set up
9614 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9616 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9617 * lra-eliminations.c (spill_pseudos): Ditto.
9618 (process_insn_for_elimination): Ditto.
9619 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9620 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9621 LRA_NON_CLOBBERED_ALT.
9623 2018-02-22 Martin Sebor <msebor@redhat.com>
9625 PR tree-optimization/84480
9626 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9627 to maybe_diag_stxncpy_trunc. Call it.
9628 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9629 from gimple_fold_builtin_strcpy. Print inlining stack.
9630 (handle_builtin_stxncpy): Print inlining stack.
9631 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9633 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9636 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9637 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9638 and -fcheck-pointer-bounds are used together.
9639 (indirect_thunk_prefix): New enum.
9640 (indirect_thunk_need_prefix): New function.
9641 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9642 "_nt" instead of "_bnd" for NOTRACK prefix.
9643 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9644 (output_indirect_thunk_function): Likewise.
9646 (ix86_code_end): Update output_indirect_thunk_function calls.
9647 (ix86_output_indirect_branch_via_reg): Replace
9648 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9649 (ix86_output_indirect_branch_via_push): Likewise.
9650 (ix86_output_function_return): Likewise.
9651 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9652 incompatible with -fcf-protection=branch and
9653 -fcheck-pointer-bounds.
9655 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9658 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9659 Change gcc_assert call to output_operand_lossage.
9661 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9663 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9665 2018-02-22 DJ Delorie <dj@redhat.com>
9666 Sebastian Perta <sebastian.perta@renesas.com>
9667 Oleg Endo <olegendo@gcc.gnu.org>
9669 * config/rx/rx.c (rx_rtx_costs): New function.
9670 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9672 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9674 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9676 2018-02-22 Martin Liska <mliska@suse.cz>
9679 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9680 Add "native" as a possible value.
9682 2018-02-22 Martin Liska <mliska@suse.cz>
9685 * config/i386/i386.c (ix86_option_override_internal):
9686 Add "native" as a possible value for -march and -mtune.
9688 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9691 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9692 to all type variants.
9694 PR tree-optimization/84503
9695 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9696 width as info->bitpos + info->bitsize - start.
9697 (merged_store_group::merge_overlapping): Simplify width computation.
9698 (check_no_overlap): New function.
9699 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9700 start + width and last_order of the group, fail if check_no_overlap
9702 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9703 to group if check_no_overlap fails.
9705 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9707 * config/rs6000/altivec.md: Delete contraint arguments to
9708 define_expand, define_split, and define_peephole2, and in
9709 define_insn_and_split if always unused.
9710 * config/rs6000/darwin.md: Ditto.
9711 * config/rs6000/dfp.md: Ditto.
9712 * config/rs6000/rs6000.md: Ditto.
9713 * config/rs6000/sync.md: Ditto.
9714 * config/rs6000/vector.md: Ditto.
9715 * config/rs6000/vsx.md: Ditto.
9717 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9719 * config/rs6000/altivec.md: Write output control strings as braced
9720 blocks instead of double-quoted strings.
9721 * config/rs6000/darwin.md: Ditto.
9722 * config/rs6000/rs6000.md: Ditto.
9723 * config/rs6000/vector.md: Ditto.
9724 * config/rs6000/vsx.md: Ditto.
9726 2018-02-21 Jason Merrill <jason@redhat.com>
9728 PR c++/84314 - ICE with templates and fastcall attribute.
9729 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9731 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9733 * ipa-cp.c (determine_versionability): Fix comment typos.
9735 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9738 * ipa-cp.c (determine_versionability): Do not version functions caling
9741 2018-02-21 Martin Liska <mliska@suse.cz>
9744 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9745 Add "native" as a possible value.
9746 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9747 the macro when native cpu detection is available.
9749 2018-02-21 Martin Liska <mliska@suse.cz>
9752 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9753 Add "native" as a possible value.
9754 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9755 when native cpu detection is available.
9757 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9758 Martin Sebor <msebor@redhat.com>
9760 PR tree-optimization/84478
9761 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9763 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9764 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9765 support which is conservatively correct, for 2 only stay conservative
9766 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9767 argument to the 2 argument get_range_strlen, adjust 6 arg
9768 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9770 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9771 (gimple_fold_builtin_strlen): Pass true as last argument to
9774 2018-02-20 Martin Sebor <msebor@redhat.com>
9777 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9778 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9779 (builtin_memref::builtin_memref): Factor out parts into
9780 set_base_and_offset and call it.
9782 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9785 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9786 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9787 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9788 search at the associated MODE_INT.
9790 2018-02-20 Jeff Law <law@redhat.com>
9793 PR tree-optimization/81592
9795 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9796 for range data rather than using global data.
9797 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9798 range data rather than using global data.
9799 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9800 pass it to children as needed.
9801 (struct directive::fmtresult): Similarly.
9802 (struct directive::set_width): Similarly.
9803 (struct directive::set_precision): Similarly.
9804 (format_integer, format_directive, parse_directive): Similarly.
9805 (format_none): Accept unnamed vr_values parameter.
9806 (format_percent, format_floating, format_character): Similarly.
9807 (format_string, format_plain): Similarly.
9808 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9809 the EVRP range analyzer for range data rather than using global data.
9810 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9811 gimple-ssa-evrp-analyze.h
9812 (class sprintf_dom_walker): Add after_dom_children member function.
9813 Add evrp_range_analyzer member.
9814 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9815 range analyzer as needed.
9816 (sprintf_dom_walker::after_dom_children): New member function.
9817 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9819 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9820 (evrp_range_analyzer::pop_to_marker): Likewise.
9822 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9824 PR tree-optimization/84419
9825 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9826 with the required type if its current type is compatible but
9829 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9832 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9833 after vectorization.
9835 2018-02-20 Martin Liska <mliska@suse.cz>
9838 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9839 possible values if we don't have a hint.
9841 2018-02-20 Martin Liska <mliska@suse.cz>
9845 * final.c (shorten_branches): Build align_tab array with one
9847 * opts.c (finish_options): Add alignment option limit check.
9848 (MAX_CODE_ALIGN): Likewise.
9849 (MAX_CODE_ALIGN_VALUE): Likewise.
9850 * doc/invoke.texi: Document maximum allowed option value for
9851 all -falign-* options.
9853 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9856 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9857 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9858 * var-tracking.c (emit_note_insn_var_location): Remove all references
9859 to NOTE_INSN_CALL_ARG_LOCATION.
9860 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9861 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9862 Use copy_rtx_if_shared.
9863 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9864 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9865 (dwarf2out_var_location): Remove handling of
9866 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9868 * final.c (final_scan_insn): Remove all references to
9869 NOTE_INSN_CALL_ARG_LOCATION.
9870 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9871 before dumping final insns.
9872 * except.c (emit_note_eh_region_end): Remove all references to
9873 NOTE_INSN_CALL_ARG_LOCATION.
9874 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9875 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9876 * config/arc/arc.c (hwloop_optimize): Likewise.
9877 * config/arm/arm.c (create_fix_barrier): Likewise.
9878 * config/s390/s390.c (s390_chunkify_start): Likewise.
9879 * config/sh/sh.c (find_barrier): Likewise.
9880 * config/i386/i386.c (rest_of_insert_endbranch,
9881 ix86_seh_fixup_eh_fallthru): Likewise.
9882 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9883 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9884 * config/frv/frv.c (frv_function_prologue): Likewise.
9885 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9887 (note_outside_basic_block_p): Remove all references to
9888 NOTE_INSN_CALL_ARG_LOCATION.
9889 * gengtype.c (adjust_field_rtx_def): Likewise.
9890 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9892 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9893 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9896 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9899 PR tree-optimization/84452
9900 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9901 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9904 2018-02-19 Martin Liska <mliska@suse.cz>
9907 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9909 2018-02-19 Martin Liska <mliska@suse.cz>
9910 Richard Sandiford <richard.sandiford@linaro.org>
9912 PR tree-optimization/82491
9913 * gimple-fold.c (get_base_constructor): Make earlier bail out
9916 2018-02-19 Carl Love <cel@us.ibm.com>
9918 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9919 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9921 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9922 P8V_BUILTIN_VEC_NEG.
9924 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9926 * config/rl78/rl78.md (movdf): New define expand.
9928 2018-02-19 Martin Liska <mliska@suse.cz>
9931 * doc/invoke.texi: Fix typo.
9932 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9934 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9936 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9937 handle rs6000_single_float and rs6000_double_float specially for
9940 2018-02-16 Jeff Law <law@redhat.com>
9942 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9945 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9948 * ipa-inline.c (inline_small_functions): Fix a typo.
9950 2018-02-16 Nathan Sidwell <nathan@acm.org>
9952 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9954 2018-02-16 Carl Love <cel@us.ibm.com>
9956 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9957 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9958 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9959 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9960 expansion to P8V_BUILTIN_VEC_FLOAT2.
9962 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9964 PR rtl-optimization/70023
9965 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9966 src_regno into account.
9968 2018-02-16 Carl Love <cel@us.ibm.com>
9970 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9971 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9972 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9973 * config/rs6000/rs6000.c: Remove case statements for
9974 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9975 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9976 and P9V_BUILTIN_VEC_VINSERT4B.
9977 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9978 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9979 * config/rs6000/vsx.md:
9980 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9983 2018-02-16 Carl Love <cel@us.ibm.com>
9985 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9987 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9989 * config/rs6000/rs6000-c.c: Add the definitions for
9990 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9991 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9992 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9993 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9994 definition for insert4b and define insn *insert3b_internal.
9995 * doc/extend.texi: Add documentation for vec_extract4b.
9997 2018-02-16 Nathan Sidwell <nathan@acm.org>
9999 * doc/extend.texi (Backwards Compatibility): Mention friend
10000 injection. Note for-scope is deprecated.
10001 * doc/invoke.texi (-ffriend-injection): Deprecate.
10003 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10005 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10006 that moved to I2, also allow destinations that are a paradoxical
10007 subreg (instead of a normal reg).
10009 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10012 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10015 2018-02-16 Richard Biener <rguenther@suse.de>
10017 PR tree-optimization/84037
10018 PR tree-optimization/84016
10020 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10021 Adjust vec_construct for the fact we need additional higher latency
10022 128bit inserts for AVX256 and AVX512 vector builds.
10023 (ix86_add_stmt_cost): Scale vector construction cost for
10026 2018-02-16 Richard Biener <rguenther@suse.de>
10028 PR tree-optimization/84417
10029 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10030 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10031 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10033 2018-02-16 Martin Liska <mliska@suse.cz>
10036 * internal-fn.def (ASAN_CHECK): Set proper flags.
10037 (ASAN_MARK): Likewise.
10039 2018-02-16 Julia Koval <julia.koval@intel.com>
10041 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10042 from PTA_CANNONLAKE.
10044 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10047 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10048 Use ++iter rather than iter++ for std::list iterators.
10049 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10050 defer deleting them until all nodes in the forest are processed. Do
10051 free even leaf nodes. Change to_process into auto_vec.
10054 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10055 * vec.h (vec_default_construct): Use memset instead of placement new
10056 if BROKEN_VALUE_INITIALIZATION is defined.
10057 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10058 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10061 PR rtl-optimization/83723
10062 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10063 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10064 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10066 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10068 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10070 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10072 PR rtl-optimization/81443
10073 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10074 from inner REGs to paradoxical SUBREGs.
10076 2018-02-16 Richard Biener <rguenther@suse.de>
10078 PR tree-optimization/84399
10079 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10080 For operands we can analyze at their definition make sure we can
10081 analyze them at each use as well.
10083 2018-02-16 Richard Biener <rguenther@suse.de>
10085 PR tree-optimization/84190
10086 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10087 volatile accesses if the decl isn't volatile.
10089 2018-02-15 Jason Merrill <jason@redhat.com>
10091 PR c++/84314 - ICE with templates and fastcall attribute.
10092 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10093 TYPE_CANONICAL on an existing type.
10095 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10097 PR tree-optimization/84383
10098 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10099 dstoff nor call operand_equal_p if dstbase is NULL.
10101 PR tree-optimization/84334
10102 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10103 also a CONSTANT_CLASS_P, punt.
10105 2018-02-14 Jim Wilson <jimw@sifive.com>
10107 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10108 first SMALL_OPERAND check. New local min_second_step. Move assert
10109 to where locals are set. Add TARGET_RVC support.
10110 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10112 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10114 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10116 2018-02-14 Martin Sebor <msebor@redhat.com>
10118 PR tree-optimization/83698
10119 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10120 arrays constrain the offset range to their bounds.
10121 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10122 (builtin_access::overlap): Avoid setting the size of overlap if it's
10124 (maybe_diag_overlap): Also consider arrays when deciding what values
10125 of offsets to include in diagnostics.
10127 2018-02-14 Martin Sebor <msebor@redhat.com>
10130 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10131 that correspond to the kind of a declaration.
10133 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10136 * config/pa/pa.md: Load address of PIC label using the linkage table
10137 if the label is nonlocal.
10139 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10141 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10142 warning message if user requests -maltivec=be.
10143 * doc/invoke.texi: Document deprecation of -maltivec=be.
10145 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10148 * config/rs6000/rs6000-c.c: Update definitions for
10149 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10150 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10152 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10155 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10156 add _get_ssp intrinsics. Remove argument from
10157 __builtin_ia32_rdssp[d|q].
10158 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10159 * config/i386/i386-builtin.def: Remove argument from
10160 __builtin_ia32_rdssp[d|q].
10161 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10162 ix86_expand_special_args_builtin for _rdssp[d|q].
10163 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10164 Clear register before usage.
10165 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10166 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10168 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10170 PR tree-optimization/84357
10171 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10172 operand 1 of an ARRAY_REF too.
10174 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10177 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10178 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10180 (set_of_reg): New struct.
10181 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10182 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10183 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10185 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10186 Split into bitclr, bitset, bitinvert patterns if appropriate.
10187 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10188 use rx_fuse_in_memory_bitop.
10189 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10190 to named insn, correct maximum insn length.
10192 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10195 * machmode.def: Define a complex mode for PARTIAL_INT.
10196 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10198 * doc/rtl.texi: Document CSPImode.
10199 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10201 (msp430_hard_regno_nregs_with_padding): Likewise.
10203 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10206 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10208 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10210 PR rtl-optimization/84169
10211 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10212 we generated a parallel as new i3 and we split that to new i2 and i3
10213 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10214 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10215 those to i2, not i1. Partially rewrite this scan code.
10217 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10220 * stor-layout.c (place_field): For variable length fields, adjust
10221 offset_align afterwards not just based on the field's alignment,
10222 but also on the size.
10224 PR middle-end/84309
10225 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10226 of exps and logs in the use_exp2 case.
10228 2018-02-13 Jeff Law <law@redhat.com>
10230 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10231 entry for "vector".
10233 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10236 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10240 * common.opt (gas-loc-support, gas-locview-support): New.
10241 (ginline-points, ginternal-reset-location-views): New.
10242 * doc/invoke.texi: Document them. Use @itemx where intended.
10243 (gvariable-location-views): Adjust.
10244 * target.def (reset_location_view): New.
10245 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10246 (TARGET_RESET_LOCATION_VIEW): New.
10247 * doc/tm.texi: Rebuilt.
10248 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10249 (dwarf2out_default_as_locview_support): New.
10250 (output_asm_line_debug_info): Use option variables.
10251 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10252 (output_loc_list): Likewise.
10253 (add_high_low_attributes): Check option variables.
10254 Don't output entry view attribute in strict mode.
10255 (gen_inlined_subroutine_die): Check option variables.
10256 (dwarf2out_inline_entry): Likewise.
10257 (init_sections_and_labels): Likewise.
10258 (dwarf2out_early_finish): Likewise.
10259 (maybe_reset_location_view): New, from...
10260 (dwarf2out_var_location): ... here. Call it.
10261 * debug.h (dwarf2out_default_as_loc_support): Declare.
10262 (dwarf2out_default_as_locview_support): Declare.
10263 * hooks.c (hook_int_rtx_insn_0): New.
10264 * hooks.h (hook_int_rtx_insn_0): Declare.
10265 * toplev.c (process_options): Take -gas-loc-support and
10266 -gas-locview-support from dwarf2out. Enable
10267 -gvariable-location-views by default only with locview
10268 assembler support. Enable -ginternal-reset-location-views by
10269 default only if the target defines the corresponding hook.
10270 Enable -ginline-points by default if location views are
10271 enabled; force it disabled if statement frontiers are
10273 * tree-inline.c (expand_call_inline): Check option variables.
10274 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10276 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10278 PR tree-optimization/84321
10279 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10280 handling. Also check whether the anti-range contains any values
10281 that satisfy the mask; switch to a VR_RANGE if not.
10283 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10286 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10288 2018-02-13 Martin Jambor <mjambor@suse.cz>
10291 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10292 of call statements, also set location of a load to a temporary.
10294 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10296 * config/rl78/rl78.c (add_vector_labels): New function.
10297 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10298 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10299 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10300 which checks that no arguments are passed.
10301 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10302 * doc/extend.texi: Documentation for the new attribute.
10304 2018-02-13 Andreas Schwab <schwab@suse.de>
10306 * config/riscv/linux.h (CPP_SPEC): Define.
10308 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10311 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10312 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10313 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10314 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10315 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10316 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10317 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10318 temporarily for AES and PCLMUL builtins.
10320 PR tree-optimization/84339
10321 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10322 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10325 PR middle-end/84309
10326 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10327 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10328 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10330 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10332 * omp-simd-clone.h: New file.
10333 * omp-simd-clone.c: Include omp-simd-clone.h.
10334 (expand_simd_clones): No longer static.
10335 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10336 cgraph.h and omp-simd-clone.h.
10337 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10338 (vect_recog_widen_shift_pattern): Formatting fix.
10339 (vect_pattern_recog_1): Don't check optab for calls.
10342 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10343 operands[2] into a REG before using gen_lowpart on it.
10345 2018-02-12 Jeff Law <law@redhat.com>
10348 * config/sh/sh.c (find_barrier): Consider a sibling call
10351 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10352 successfully back substituting a reg.
10354 2018-02-12 Richard Biener <rguenther@suse.de>
10356 PR tree-optimization/84037
10357 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10358 parameter, move visited init to caller.
10359 (vect_slp_analyze_operations): Separate cost from validity
10360 check, initialize visited once for all instances.
10361 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10363 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10365 (vect_model_promotion_demotion_cost): Likewise.
10366 (vectorizable_bswap): Guard cost modeling with !slp_node
10367 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10369 (vectorizable_call): Likewise.
10370 (vectorizable_conversion): Likewise.
10371 (vectorizable_assignment): Likewise.
10372 (vectorizable_shift): Likewise.
10373 (vectorizable_operation): Likewise.
10374 (vectorizable_store): Likewise.
10375 (vectorizable_load): Likewise.
10376 (vectorizable_condition): Likewise.
10377 (vectorizable_comparison): Likewise.
10379 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10382 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10383 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10384 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10386 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10388 PR middle-end/83665
10389 * params.def (inline-min-speedup): Increase from 8 to 15.
10390 (max-inline-insns-auto): Decrease from 40 to 30.
10391 * ipa-split.c (consider_split): Add some buffer for function to
10392 be considered inlining candidate.
10393 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10396 2018-02-12 Richard Biener <rguenther@suse.de>
10398 PR tree-optimization/84037
10399 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10400 matched stmts if we cannot swap the non-matched ones.
10402 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10404 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10405 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10406 _mm_maskz_scalef_round_ss): New intrinsics.
10407 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10408 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10409 __builtin_ia32_scalefss_round): Remove.
10410 (__builtin_ia32_scalefsd_mask_round,
10411 __builtin_ia32_scalefss_mask_round): New intrinsics.
10412 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10413 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10414 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10415 "<round_constraint>")): Changed to ...
10416 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10417 "<round_scalar_constraint>")): ... this.
10418 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10419 %0, %1, %2<round_op3>}"): Changed to ...
10420 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10421 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10422 %2<round_scalar_mask_op3>}"): ... this.
10423 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10425 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10427 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10428 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10429 (_mm_maskz_sqrt_round_ss): New intrinsics.
10430 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10431 (__builtin_ia32_sqrtsd_mask_round)
10432 (__builtin_ia32_sqrtss_mask_round): New builtins.
10433 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10434 (__builtin_ia32_sqrtss_round): Remove.
10435 (__builtin_ia32_sqrtsd_mask_round)
10436 (__builtin_ia32_sqrtss_mask_round): New builtins.
10437 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10438 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10439 ((match_operand:VF_128 1 "vector_operand"
10440 "xBm,<round_constraint>")): Changed to ...
10441 ((match_operand:VF_128 1 "vector_operand"
10442 "xBm,<round_scalar_constraint>")): ... this.
10443 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10444 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10445 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10446 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10447 %<iptr>1<round_scalar_mask_op3>}): ... this.
10448 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10449 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10451 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10454 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10455 Cast vec_cmpeq result to correct type.
10456 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10457 Cast vec_cmpgt result to correct type.
10459 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10461 * final.c (final_scan_insn_1): Renamed from...
10462 (final_scan_insn): ... this. New wrapper, to recover
10463 seen from the outermost call in recursive ones.
10464 * config/sparc/sparc.c (output_return): Drop seen from call.
10465 (output_sibcall): Likewise.
10466 * config/visium/visium.c (output_branch): Likewise.
10468 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10470 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10473 2018-02-10 Alan Modra <amodra@gmail.com>
10476 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10477 Specify LR as an input.
10479 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10482 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10483 remove_member_access_dummy_vars): New functions.
10484 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10485 lower_omp_1, execute_lower_omp): Use them.
10487 PR rtl-optimization/84308
10488 * shrink-wrap.c (spread_components): Release todo vector.
10490 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10492 PR rtl-optimization/57193
10493 * ira-color.c (struct allocno_color_data): Add member
10494 conflict_allocno_hard_prefs.
10495 (update_conflict_allocno_hard_prefs): New.
10496 (bucket_allocno_compare_func): Add a preference based on
10497 conflict_allocno_hard_prefs.
10498 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10499 (color_allocnos): Remove a dead code. Initiate
10500 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10502 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10505 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10506 constraint from =wa to wa. Avoid a subreg on the output operand,
10507 instead use a pseudo and subreg it in a move.
10508 (p9_xxbrd_<mode>): Changed to ...
10509 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10510 (p9_xxbrd_v2df): New expander.
10511 (p9_xxbrw_<mode>): Changed to ...
10512 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10513 (p9_xxbrw_v4sf): New expander.
10515 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10517 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10519 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10522 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10523 multiply in 32-bit mode.
10524 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10525 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10528 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10530 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10531 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10532 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10533 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10535 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10538 * dwarf2out.c (is_trivial_indirect_ref): New function.
10539 (dwarf2out_late_global_decl): Do not generate a location
10540 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10541 and that are not defined in the current unit.
10543 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10545 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10546 instead of a libcall for UNORDERED.
10548 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10551 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10552 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10554 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10557 * config/s390/s390.c (s390_set_current_function): Invoke
10558 s390_indirect_branch_settings also if fndecl didn't change.
10560 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10562 * config/rs6000/rs6000.md (blockage): Set length to zero.
10564 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10566 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10568 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10571 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10572 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10576 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10577 PARALLEL incoming that failed vt_get_decl_and_offset check.
10579 PR middle-end/84237
10580 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10581 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10583 (get_variable_section): For decls in named .bss* sections pass true as
10584 second argument to bss_initializer_p.
10586 2018-02-09 Marek Polacek <polacek@redhat.com>
10587 Jakub Jelinek <jakub@redhat.com>
10590 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10591 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10592 Sync some changes from cxx_fold_indirect_ref.
10594 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10596 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10598 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10599 (BLOCK_INLINE_ENTRY_LABEL): New.
10600 (dwarf2out_var_location): Disregard inline entry markers.
10601 (inline_entry_data): New struct.
10602 (inline_entry_data_hasher): New hashtable type.
10603 (inline_entry_data_hasher::hash): New.
10604 (inline_entry_data_hasher::equal): New.
10605 (inline_entry_data_table): New variable.
10606 (add_high_low_attributes): Add DW_AT_entry_pc and
10607 DW_AT_GNU_entry_view attributes if a pending entry is found
10608 in inline_entry_data_table. Add old entry_pc attribute only
10609 if debug nonbinding markers are disabled.
10610 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10611 markers are enabled.
10612 (block_within_block_p, dwarf2out_inline_entry): New.
10613 (dwarf2out_finish): Check that no entries remained in
10614 inline_entry_data_table.
10615 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10616 (final_scan_insn, notice_source_line): Likewise.
10617 (rest_of_clean_state): Skip inline entry markers.
10618 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10620 * gimple.c (gimple_build_debug_inline_entry): New.
10621 * gimple.h (enum gimple_debug_subcode): Add
10622 GIMPLE_DEBUG_INLINE_ENTRY.
10623 (gimple_build_debug_inline_entry): Declare.
10624 (gimple_debug_inline_entry_p): New.
10625 (gimple_debug_nonbind_marker_p): Adjust.
10626 * insn-notes.def (INLINE_ENTRY): New.
10627 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10628 inline entry marker notes.
10629 (print_insn): Likewise.
10630 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10631 (INSN_DEBUG_MARKER_KIND): Likewise.
10632 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10633 * tree-inline.c (expand_call_inline): Build and insert
10634 debug_inline_entry stmt.
10635 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10636 inline entry blocks early, if nonbind markers are enabled.
10637 (dump_scope_block): Dump fragment info.
10638 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10639 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10640 (gimple_build_debug_inline_entry): New.
10641 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10642 Enable/disable inline entry points too.
10643 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10644 (DEBUG_INSN): Describe inline entry markers.
10646 * common.opt (gvariable-location-views): New.
10647 (gvariable-location-views=incompat5): New.
10648 * config.in: Rebuilt.
10649 * configure: Rebuilt.
10650 * configure.ac: Test assembler for view support.
10651 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10652 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10653 * dwarf2out.c (var_loc_view): New typedef.
10654 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10655 (dwarf2out_locviews_in_attribute): New.
10656 (dwarf2out_locviews_in_loclist): New.
10657 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10658 (enum dw_line_info_opcode): Add LI_adv_address.
10659 (struct dw_line_info_table): Add view.
10660 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10661 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10662 (zero_view_p): New variable.
10663 (ZERO_VIEW_P): New macro.
10664 (output_asm_line_debug_info): New.
10665 (struct var_loc_node): Add view.
10666 (add_AT_view_list, AT_loc_list): New.
10667 (add_var_loc_to_decl): Add view param. Test it against last.
10668 (new_loc_list): Add view params. Record them.
10669 (AT_loc_list_ptr): Handle loc and view lists.
10670 (view_list_to_loc_list_val_node): New.
10671 (print_dw_val): Handle dw_val_class_view_list.
10672 (size_of_die): Likewise.
10673 (value_format): Likewise.
10674 (loc_list_has_views): New.
10675 (gen_llsym): Set vl_symbol too.
10676 (maybe_gen_llsym, skip_loc_list_entry): New.
10677 (dwarf2out_maybe_output_loclist_view_pair): New.
10678 (output_loc_list): Output view list or entries too.
10679 (output_view_list_offset): New.
10680 (output_die): Handle dw_val_class_view_list.
10681 (output_dwarf_version): New.
10682 (output_compilation_unit_header): Use it.
10683 (output_skeleton_debug_sections): Likewise.
10684 (output_rnglists, output_line_info): Likewise.
10685 (output_pubnames, output_aranges): Update version comments.
10686 (output_one_line_info_table): Output view numbers in asm comments.
10687 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10688 Call maybe_gen_llsym.
10689 (loc_list_from_tree_1): Adjust.
10690 (add_AT_location_description): Create view list attribute if
10691 needed, check it's absent otherwise.
10692 (convert_cfa_to_fb_loc_list): Adjust.
10693 (maybe_emit_file): Call output_asm_line_debug_info for test.
10694 (dwarf2out_var_location): Reset views as needed. Precompute
10695 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10696 attribute. Set view.
10697 (new_line_info_table): Reset next view.
10698 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10699 (dwarf2out_source_line): Likewise. Output view resets and labels to
10700 the assembler, or select appropriate line info opcodes.
10701 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10702 (optimize_string_length): Catch it. Adjust.
10703 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10704 dw_val_class_view_list, and remove it if no longer needed.
10705 (hash_loc_list): Hash view numbers.
10706 (loc_list_hasher::equal): Compare them.
10707 (optimize_location_lists): Check whether a view list symbol is
10708 needed, and whether the locview attribute is present, and
10709 whether they match. Remove the locview attribute if no longer
10711 (index_location_lists): Call skip_loc_list_entry for test.
10712 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10713 Use output_dwarf_version.
10714 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10715 (struct dw_val_node): Add val_view_list.
10716 * final.c (SEEN_NEXT_VIEW): New.
10717 (set_next_view_needed): New.
10718 (clear_next_view_needed): New.
10719 (maybe_output_next_view): New.
10720 (final_start_function): Rename to...
10721 (final_start_function_1): ... this. Take pointer to FIRST,
10722 add SEEN parameter. Emit param bindings in the initial view.
10723 (final_start_function): Reintroduce SEEN-less interface.
10724 (final): Rename to...
10725 (final_1): ... this. Take SEEN parameter. Output final pending
10726 next view at the end.
10727 (final): Reintroduce seen-less interface.
10728 (final_scan_insn): Output pending next view before switching
10729 sections or ending a block. Mark the next view as needed when
10730 outputting variable locations. Notify debug backend of section
10731 changes, and of location view changes.
10732 (rest_of_handle_final): Adjust.
10733 * toplev.c (process_options): Autodetect value for debug variable
10734 location views option. Warn on incompat5 without -gdwarf-5.
10735 * doc/invoke.texi (gvariable-location-views): New.
10736 (gvariable-location-views=incompat5): New.
10737 (gno-variable-location-views): New.
10739 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10741 PR tree-optimization/84136
10742 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10743 that the result of find_edge is non-NULL.
10745 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10748 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10749 storing integer register in SImode. Fix cost of 256 and 512
10750 byte aligned SSE register store.
10752 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10754 * config/i386/i386.c (ix86_multiplication_cost): Fix
10755 multiplication cost for TARGET_AVX512DQ.
10757 2018-02-08 Marek Polacek <polacek@redhat.com>
10759 PR tree-optimization/84238
10760 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10763 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10765 PR tree-optimization/84265
10766 * tree-vect-stmts.c (vectorizable_store): Don't treat
10767 VMAT_CONTIGUOUS accesses as grouped.
10768 (vectorizable_load): Likewise.
10770 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10772 PR tree-optimization/81635
10773 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10774 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10775 (test_round_for_mask): New functions.
10776 (wide_int_cc_tests): Call test_round_for_mask.
10777 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10778 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10779 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10780 range returned by get_range_info.
10782 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10785 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10786 * symtab.c: Include builtins.h
10787 (symtab_node::output_to_lto_symbol_table_p): Move here
10788 from lto-streamer-out.c:output_symbol_p.
10789 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10790 (output_symbol_p): Move all logic to symtab.c
10791 (produce_symtab): Update.
10793 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10795 * config/s390/s390-opts.h (enum indirect_branch): Define.
10796 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10797 (s390_indirect_branch_via_thunk)
10798 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10799 (enum s390_indirect_branch_type): Define.
10800 * config/s390/s390.c (struct s390_frame_layout, struct
10801 machine_function): Remove.
10802 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10803 (indirect_branch_table_label_no, indirect_branch_table_name):
10805 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10806 (enum s390_indirect_branch_option): Define.
10807 (s390_return_addr_from_memory): New function.
10808 (s390_handle_string_attribute): New function.
10809 (s390_attribute_table): Add new attribute handler.
10810 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10811 (s390_indirect_branch_via_thunk): New function.
10812 (s390_indirect_branch_via_inline_thunk): New function.
10813 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10814 sibling call optimization for non z10 compiles.
10815 (s390_emit_call): Force indirect branch target to be a single
10816 register. Add r1 clobber for non-z10 compiles.
10817 (s390_emit_epilogue): Emit return jump via return_use expander.
10818 (s390_reorg): Handle JUMP_INSNs as execute targets.
10819 (s390_option_override_internal): Perform validity checks for the
10820 new command line options.
10821 (s390_indirect_branch_attrvalue): New function.
10822 (s390_indirect_branch_settings): New function.
10823 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10824 (s390_output_indirect_thunk_function): New function.
10825 (s390_code_end): Implement target hook.
10826 (s390_case_values_threshold): Implement target hook.
10827 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10829 * config/s390/s390.h (struct s390_frame_layout)
10830 (struct machine_function): Move here from s390.c.
10831 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10832 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10833 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10834 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10835 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10836 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10837 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10838 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10839 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10840 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10841 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10842 (mnemonic attribute): Add values which aren't recognized
10844 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10845 pattern for branch conversion. Fix mnemonic attribute.
10846 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10847 indirect branch via thunk if requested.
10848 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10849 ("*indirect_jump"): Disable for branch conversion using out of
10851 ("indirect_jump_via_thunk<mode>_z10")
10852 ("indirect_jump_via_thunk<mode>")
10853 ("indirect_jump_via_inlinethunk<mode>_z10")
10854 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10855 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10856 ("casesi_jump_via_inlinethunk<mode>_z10")
10857 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10858 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10859 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10860 ("*indirect2_jump"): Disable for branch conversion.
10861 ("casesi_jump"): Turn into expander and expand patterns for branch
10863 ("return_use"): New expander.
10864 ("*return"): Emit return via thunk and rename it to ...
10865 ("*return<mode>"): ... this one.
10866 * config/s390/s390.opt: Add new options and and enum for the
10869 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10871 * lra-constraints.c (match_reload): Unconditionally use
10872 gen_lowpart_SUBREG, rather than selecting between that
10873 and equivalent gen_rtx_SUBREG code.
10875 2018-02-08 Richard Biener <rguenther@suse.de>
10877 PR tree-optimization/84233
10878 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10879 changed flag instead of boguously re-using phi_inserted.
10881 2018-02-08 Martin Jambor <mjambor@suse.cz>
10883 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10884 static local variables.
10886 2018-02-08 Richard Biener <rguenther@suse.de>
10888 PR tree-optimization/84278
10889 * tree-vect-stmts.c (vectorizable_store): When looking for
10890 smaller vector types to perform grouped strided loads/stores
10891 make sure the mode is supported by the target.
10892 (vectorizable_load): Likewise.
10894 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10896 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10897 Increase LDP/STP opportunities by adding adjacent callee-saves.
10899 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10901 PR rtl-optimization/84068
10902 PR rtl-optimization/83459
10903 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10905 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10907 PR tree-optimization/84224
10908 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10909 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10910 non-zero arguments.
10912 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10915 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10916 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10917 count, remove one USE.
10919 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10921 * doc/install.texi (Configuration): Document the
10922 --with-long-double-format={ibm,ieee} PowerPC configuration
10926 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10927 Convert from define_expand to be define_insn_and_split. Rework
10928 float/double/_Float128 conversions to QI/HI/SImode to work with
10929 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10930 conversions to QI/HImode types did a store and then a load to
10931 truncate the value. For conversions to VSX registers, don't split
10932 the insn, instead emit the code directly. Use the code iterator
10933 any_fix to combine signed and unsigned conversions.
10934 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10935 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10936 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10937 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10938 (fix_<mode>di2_hw): Likewise.
10939 (fixuns_<mode>di2_hw): Likewise.
10940 (fix_<mode>si2_hw): Likewise.
10941 (fixuns_<mode>si2_hw): Likewise.
10942 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10943 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10944 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10945 fix<uns>_trunc<SFDF:mode>si2_p8.
10946 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10947 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10948 (fix<uns>_<mode>_mem): Likewise.
10949 (fctiw<u>z_<mode>_mem): Likewise.
10950 (fix<uns>_<mode>_mem): Likewise.
10951 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10952 the register allocator from doing a direct move to the GPRs to do
10953 a store, and instead use the ISA 3.0 store byte/half-word from
10954 vector register instruction. For IEEE 128-bit floating point,
10955 also optimize stores of 32-bit ints.
10956 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10958 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10960 * genextract.c (push_pathstr_operand): New function to support
10962 (walk_rtx): Call push_pathstr_operand.
10963 (print_path): Support [a-zA-Z].
10965 2018-02-07 Richard Biener <rguenther@suse.de>
10967 PR tree-optimization/84037
10968 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10969 (cse_and_gimplify_to_preheader): Declare.
10970 (vect_get_place_in_interleaving_chain): Likewise.
10971 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10973 (_loop_vec_info::~_loop_vec_info): Delete it.
10974 (cse_and_gimplify_to_preheader): New function.
10975 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10976 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10977 (vectorizable_load): Likewise. For grouped stores always base
10978 the IV on the first element.
10979 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10980 condition before gimplifying.
10982 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10984 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10985 *DIV_EXPR and *MOD_EXPR.
10987 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10990 * config/i386/i386.c (ix86_option_override_internal): Mask out
10991 the CF_SET bit when checking -fcf-protection.
10993 2018-02-07 Tom de Vries <tom@codesourcery.com>
10996 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10999 2018-02-07 Richard Biener <rguenther@suse.de>
11001 PR tree-optimization/84204
11002 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11005 PR tree-optimization/84205
11006 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11007 special-case isl_ast_op_zdiv_r.
11009 PR tree-optimization/84223
11010 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11011 Only add conditions from within the region.
11012 (gather_bbs::after_dom_children): Adjust.
11014 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11017 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11018 * config/avr/avr.md: Only post-reload split REG-REG moves if
11019 either register is GENERAL_REG_P.
11021 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11023 PR tree-optimization/84235
11024 * tree-ssa-scopedtables.c
11025 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11026 if the subtraction is performed in floating point type where NaNs are
11027 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11028 build 1. Formatting fix.
11030 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11033 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11034 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11035 and skip it regardless of bb boundaries. Use CALL_P macro,
11036 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11037 unnecessarily, formatting fix.
11039 2018-02-06 Michael Collison <michael.collison@arm.com>
11041 * config/arm/thumb2.md:
11042 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11043 (*thumb_mov_notscc): Ditto.
11045 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11048 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11049 unsigned_fix, not "s".
11051 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11053 * configure.ac (gcc_fn_eh_frame_ro): New function.
11054 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11055 correct .eh_frame permissions.
11056 * configure: Regenerate.
11058 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11060 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11061 irrelevant options.
11063 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11065 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11066 Display warning message for -mno-speculate-indirect-jumps.
11068 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11070 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11072 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11074 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11076 PR tree-optimization/84225
11077 * tree-eh.c (find_trapping_overflow): Only call
11078 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11080 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11083 * config/i386/i386.c: Reimplement the check of possible options
11084 -mibt/-mshstk conbination. Change error messages.
11085 * doc/invoke.texi: Fix a typo: remove extra '='.
11087 2018-02-06 Marek Polacek <polacek@redhat.com>
11089 PR tree-optimization/84228
11090 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11092 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11095 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11096 emitted arch directives.
11097 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11098 __ARM_FEATURE_COPROC before changing architectures.
11100 2018-02-06 Richard Biener <rguenther@suse.de>
11102 * config/i386/i386.c (print_reg): Fix typo.
11103 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11105 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11107 * configure: Regenerate.
11109 2018-02-05 Martin Sebor <msebor@redhat.com>
11111 PR tree-optimization/83369
11112 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11115 2018-02-05 Martin Liska <mliska@suse.cz>
11117 * doc/invoke.texi: Cherry-pick upstream r323995.
11119 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11121 * ira.c (ira_init_register_move_cost): Adjust comment.
11123 2018-02-05 Martin Liska <mliska@suse.cz>
11125 PR gcov-profile/84137
11126 * doc/gcov.texi: Fix typo in documentation.
11128 2018-02-05 Martin Liska <mliska@suse.cz>
11130 PR gcov-profile/83879
11131 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11132 using dlopen functionality.
11134 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11136 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11137 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11138 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11139 _mm_maskz_range_round_ss): New intrinsics.
11140 (__builtin_ia32_rangesd128_round)
11141 (__builtin_ia32_rangess128_round): Remove.
11142 (__builtin_ia32_rangesd128_mask_round,
11143 __builtin_ia32_rangess128_mask_round): New builtins.
11144 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11145 __builtin_ia32_rangess128_round): Remove.
11146 (__builtin_ia32_rangesd128_mask_round,
11147 __builtin_ia32_rangess128_mask_round): New builtins.
11148 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11149 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11150 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11151 "<round_saeonly_constraint>")): Changed to ...
11152 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11153 "<round_saeonly_scalar_constraint>")): ... this.
11154 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11155 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11156 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11157 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11158 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11160 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11162 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11164 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11165 Remove all values except native, 8540 and 8548.
11167 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11169 * config/i386/i386.c (ix86_output_function_return): Pass
11170 INVALID_REGNUM, instead of -1, as invalid register number to
11171 indirect_thunk_name and output_indirect_thunk.
11173 2018-02-02 Julia Koval <julia.koval@intel.com>
11175 * config.gcc: Add -march=icelake.
11176 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11177 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11178 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11179 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11180 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11181 (processor_target_table): Add icelake.
11182 (ix86_option_override_internal): Handle new PTAs.
11183 (get_builtin_code_for_version): Handle icelake.
11184 (M_INTEL_COREI7_ICELAKE): New.
11185 (fold_builtin_cpu): Handle icelake.
11186 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11187 * doc/invoke.texi: Add -march=icelake.
11189 2018-02-02 Julia Koval <julia.koval@intel.com>
11191 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11192 to wide_int_bitmask.
11193 * wide-int-bitmask.h: New.
11195 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11198 * config/i386/i386.md: Replace Pmode with word_mode in
11199 builtin_setjmp_setup and builtin_longjmp to support x32.
11201 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11205 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11207 (rs6000_supported_cpu_names): New static variable.
11208 (linux_cpu_translation_table): Likewise.
11209 (elf_platform) <cpu>: Define new static variable and use it.
11210 Translate kernel AT_PLATFORM name to canonical name if needed.
11211 Error if platform name is unknown.
11213 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11216 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11218 2018-02-01 Jeff Law <law@redhat.com>
11221 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11222 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11223 the scratch if RELEASE_VIA_POP is false.
11224 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11225 If we have to save a temporary register, decrement SIZE appropriately.
11226 Pass new arguments to release_scratch_register_on_entry.
11227 (ix86_adjust_stack_and_probe): Likewise.
11228 (ix86_emit_probe_stack_range): Pass new arguments to
11229 release_scratch_register_on_entry.
11231 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11233 PR rtl-optimization/84157
11234 * combine.c (change_zero_ext): Use REG_P predicate in
11235 front of HARD_REGISTER_P predicate.
11237 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11239 * config/avr/avr.c (avr_option_override): Move disabling of
11240 -fdelete-null-pointer-checks to...
11241 * common/config/avr/avr-common.c (avr_option_optimization_table):
11244 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11246 PR tree-optimization/81635
11247 * tree-data-ref.c (split_constant_offset_1): For types that
11248 wrap on overflow, try to use range info to prove that wrapping
11251 2018-02-01 Renlin Li <renlin.li@arm.com>
11254 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11255 TAILCALL_ADDR_REGS.
11256 (aarch64_register_move_cost): Likewise.
11257 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11258 TAILCALL_ADDR_REGS.
11259 (REG_CLASS_NAMES): Likewise.
11260 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11261 TAILCALL_ADDR_REGS. Remove IP registers.
11262 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11264 2018-02-01 Richard Biener <rguenther@suse.de>
11266 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11267 for specifying RPO order and allow NULL for that.
11268 * domwalk.c (dom_walker::dom_walker): Likewise.
11269 (dom_walker::walk): Handle NULL RPO order.
11270 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11272 (rewrite_update_dom_walker): Likewise.
11273 (mark_def_dom_walker): Likewise.
11275 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11277 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11278 (aarch64_maybe_expand_sve_subreg_move): Declare.
11279 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11280 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11282 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11283 that are semantically a reverse operation.
11284 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11285 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11286 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11288 (aarch64_can_change_mode_class): For big-endian, forbid changes
11289 between two SVE modes if they have different element sizes.
11291 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11293 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11294 the TImode handling for big-endian targets.
11296 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11298 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11299 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11301 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11302 Remove BSWAP handing for big-endian targets and use the form of
11303 LD1RQ appropariate for the mode.
11305 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11307 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11308 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11309 duplicated element.
11311 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11314 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11315 check for operands that need to go through aarch64_sve_reload_be.
11317 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11319 PR tree-optimization/81661
11320 PR tree-optimization/84117
11321 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11322 * tree-eh.c: Include gimplify.h.
11323 (find_trapping_overflow, replace_trapping_overflow,
11324 rewrite_to_non_trapping_overflow): New functions.
11325 * tree-vect-loop.c: Include tree-eh.h.
11326 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11327 * tree-data-ref.c: Include tree-eh.h.
11328 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11330 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11332 PR rtl-optimization/84123
11333 * combine.c (change_zero_ext): Check if hard register satisfies
11334 can_change_dest_mode before calling gen_lowpart_SUBREG.
11336 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11339 * ira.c (ira_init_register_move_cost): Remove assert.
11341 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11343 PR rtl-optimization/84071
11344 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11345 * doc/tm.texi: Regenerate.
11347 2018-01-31 Richard Biener <rguenther@suse.de>
11349 PR tree-optimization/84132
11350 * tree-data-ref.c (analyze_miv_subscript): Properly
11351 check whether evolution_function_is_affine_multivariate_p
11352 before calling gcd_of_steps_may_divide_p.
11354 2018-01-31 Julia Koval <julia.koval@intel.com>
11357 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11358 * config/i386/i386.md (rdpid_rex64) New.
11359 (rdpid): Make 32bit only.
11361 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11364 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11365 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11367 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11370 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11372 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11374 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11376 PR rtl-optimization/84071
11377 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11378 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11380 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11382 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11383 (arc_attribute_table): Add 'aux' attribute.
11384 (arc_in_small_data_p): Consider aux like variables.
11385 (arc_is_aux_reg_p): New function.
11386 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11387 (arc_get_aux_arg): New function.
11388 (prepare_move_operands): Handle aux-register access.
11389 (arc_handle_aux_attribute): New function.
11390 * doc/extend.texi (ARC Variable attributes): Add subsection.
11392 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11394 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11395 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11396 (arc_attribute_table): Add 'uncached' attribute.
11397 (arc_print_operand): Print '.di' flag for uncached memory
11399 (arc_in_small_data_p): Do not consider for small data the uncached
11401 (arc_is_uncached_mem_p): New function.
11402 * config/arc/predicates.md (compact_store_memory_operand): Check
11403 for uncached memory accesses.
11404 (nonvol_nonimm_operand): Likewise.
11405 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11407 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11410 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11411 falign-loops=): Add Optimization flag.
11413 2018-01-30 Jeff Law <law@redhat.com>
11416 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11417 INT_REGISTERS_SAVED. Check it prior to calling
11418 get_scratch_register_on_entry.
11419 (ix86_adjust_stack_and_probe): Similarly.
11420 (ix86_emit_probe_stack_range): Similarly.
11421 (ix86_expand_prologue): Corresponding changes.
11423 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11426 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11427 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11429 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11432 * lra-constraints.c (curr_insn_transform): Process AND in the
11435 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11437 PR rtl-optimization/83986
11438 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11439 dependence against last_pending_memory_flush in addition to
11440 pending_jump_insns.
11442 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11444 PR tree-optimization/81611
11445 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11448 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11451 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11454 2018-01-30 Richard Biener <rguenther@suse.de>
11455 Jakub Jelinek <jakub@redhat.com>
11457 PR tree-optimization/84111
11458 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11459 inner loops added during recursion, as they don't have up-to-date
11462 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11465 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11466 (can_inline_edge_by_limits_p): ... here.
11467 (can_early_inline_edge_p, check_callers,
11468 update_caller_keys, update_callee_keys, recursive_inlining,
11469 add_new_edges_to_heap, speculation_useful_p,
11470 inline_small_functions,
11471 inline_small_functions, flatten_function,
11472 inline_to_all_callers_1): Update.
11474 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11476 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11479 2018-01-30 Richard Biener <rguenther@suse.de>
11481 PR tree-optimization/83008
11482 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11483 invariant and constant vector uses in stmts when they need
11484 more than one stmt.
11486 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11489 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11490 * configure: Regenerate.
11492 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11494 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11496 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11497 Use gen_rtx_REG rather than gen_lowpart.
11499 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11501 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11502 rather than 0 when creating partial subregs.
11504 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11506 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11509 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11512 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11513 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11514 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11515 flags. This restores the settings used before the 2017-07-24.
11516 Turning off pre increment/decrement/modify allows IVOPTS to
11517 optimize DF/SF loops where the index is an int.
11519 2018-01-29 Richard Biener <rguenther@suse.de>
11520 Kelvin Nilsen <kelvin@gcc.gnu.org>
11523 * tree-vect-stmts.c (vectorizable_call): Don't call
11524 targetm.vectorize_builtin_md_vectorized_function if callee is
11527 2018-01-22 Carl Love <cel@us.ibm.com>
11529 * doc/extend.tex: Fix typo in second arg in
11530 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11532 2018-01-29 Richard Biener <rguenther@suse.de>
11534 PR tree-optimization/84086
11535 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11536 (flush_ssaname_freelist): When SSA names were released reset
11537 the SCEV hash table.
11539 2018-01-29 Richard Biener <rguenther@suse.de>
11541 PR tree-optimization/84057
11542 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11543 removed paths when removing edges.
11545 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11547 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11548 -mfunction-return=@var{choice}.
11550 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11552 PR diagnostic/84034
11553 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11554 Handle CR like TAB.
11555 (layout::print_source_line): Likewise.
11556 (test_get_line_width_without_trailing_whitespace): Add test cases.
11558 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11560 PR middle-end/84040
11561 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11564 2018-01-26 Jim Wilson <jimw@sifive.com>
11566 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11568 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11571 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11573 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11574 and CMP + SUB-immediate -> SUBS.
11576 2018-01-26 Martin Sebor <msebor@redhat.com>
11578 PR tree-optimization/83896
11579 * tree-ssa-strlen.c (get_string_len): Rename...
11580 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11581 Avoid assuming length is constant.
11582 (handle_char_store): Use HOST_WIDE_INT for string length.
11584 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11587 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11588 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11590 2018-01-26 Richard Biener <rguenther@suse.de>
11592 PR rtl-optimization/84003
11593 * dse.c (record_store): Only record redundant stores when
11594 the earlier store aliases at least all accesses the later one does.
11596 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11598 PR rtl-optimization/83985
11599 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11600 REG_CFA_RESTORE insns.
11601 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11602 REG_CFA_RESTORE insns here.
11605 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11606 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11608 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11610 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11611 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11612 (arc_init): Likewise.
11613 (arc_override_options): Likewise.
11614 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11616 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11618 * config/arc/arc.h (TARGET_DBNZ): Define.
11619 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11620 properly set the tune attribute.
11621 (dbnz): Use TARGET_DBNZ guard.
11622 * config/arc/arc.opt (mtune): Add core3 option.
11624 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11626 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11627 recognize new pic like addresses.
11628 (arc_delegitimize_address): Clean up.
11630 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11632 * config/arc/arc-arches.def: Option mrf16 valid for all
11634 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11635 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11636 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11637 * config/arc/arc-tables.opt: Regenerate.
11638 * config/arc/arc.c (arc_conditional_register_usage): Handle
11639 reduced register file case.
11640 (arc_file_start): Set must have build attributes.
11641 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11642 mrf16 option value.
11643 * config/arc/arc.opt (mrf16): Add new option.
11644 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11645 * config/arc/genmultilib.awk: Handle new mrf16 option.
11646 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11647 * config/arc/t-multilib: Regenerate.
11648 * doc/invoke.texi (ARC Options): Document mrf16 option.
11650 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11652 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11653 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11654 (arc_attribute_table): Add 'secure_call' attribute.
11655 (arc_print_operand): Print secure call operand.
11656 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11658 (arc_is_secure_call_p): New function. * config/arc/arc.md
11659 (call_i): Add support for sjli instruction.
11660 (call_value_i): Likewise.
11661 * config/arc/constraints.md (Csc): New constraint.
11663 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11664 John Eric Martin <John.Martin@emmicro-us.com>
11666 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11667 * config/arc/arc.c (_arc_jli_section): New struct.
11668 (arc_jli_section): New type.
11669 (rc_jli_sections): New static variable.
11670 (arc_handle_jli_attribute): New function.
11671 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11672 (arc_file_end): New function.
11673 (TARGET_ASM_FILE_END): Define.
11674 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11675 (arc_add_jli_section): New function.
11676 (jli_call_scan): Likewise.
11677 (arc_reorg): Call jli_call_scan.
11678 (arc_output_addsi): Remove 'S' from printing asm operand.
11679 (arc_is_jli_call_p): New function.
11680 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11682 (movhi_insn): Likewise.
11683 (movsi_insn): Likewise.
11684 (movsi_set_cc_insn): Likewise.
11685 (loadqi_update): Likewise.
11686 (load_zeroextendqisi_update): Likewise.
11687 (load_signextendqisi_update): Likewise.
11688 (loadhi_update): Likewise.
11689 (load_zeroextendhisi_update): Likewise.
11690 (load_signextendhisi_update): Likewise.
11691 (loadsi_update): Likewise.
11692 (loadsf_update): Likewise.
11693 (movsicc_insn): Likewise.
11694 (bset_insn): Likewise.
11695 (bxor_insn): Likewise.
11696 (bclr_insn): Likewise.
11697 (bmsk_insn): Likewise.
11698 (bicsi3_insn): Likewise.
11699 (cmpsi_cc_c_insn): Likewise.
11700 (movsi_ne): Likewise.
11701 (movsi_cond_exec): Likewise.
11702 (clrsbsi2): Likewise.
11703 (norm_f): Likewise.
11713 (call_i): Remove 'S' asm letter, add jli instruction.
11714 (call_value_i): Likewise.
11715 * config/arc/arc.op (mjli-always): New option.
11716 * config/arc/constraints.md (Cji): New constraint.
11717 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11719 (subsf3_fpx): Likewise.
11720 (mulsf3_fpx): Likewise.
11721 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11723 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11724 function attrbutes.
11725 * doc/invoke.texi (ARC): Document mjli-always option.
11727 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11729 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11730 avoid addition with 0 and use incw and decw where possible.
11732 2018-01-26 Richard Biener <rguenther@suse.de>
11734 PR tree-optimization/81082
11735 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11736 association if it requires casting to unsigned.
11737 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11738 from fold_plusminus_mult_expr to catch important cases late when
11739 range info is available.
11741 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11743 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11744 * configure.ac (hidden_linkonce): New test.
11745 * configure: Regenerate.
11746 * config.in: Regenerate.
11748 2018-01-26 Julia Koval <julia.koval@intel.com>
11750 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11751 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11752 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11753 _mm_mask_bitshuffle_epi64_mask): Fix type.
11754 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11755 USI_FTYPE_V4DI_V4DI_USI): Remove.
11756 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11757 __builtin_ia32_vpshufbitqmb256_mask,
11758 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11759 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11760 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11762 2018-01-26 Alan Modra <amodra@gmail.com>
11765 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11766 UNSPEC_VBPERMQ. Sort other unspecs.
11768 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11770 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11772 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11774 PR middle-end/83055
11775 * predict.c (drop_profile): Do not push/pop cfun; update also
11777 (handle_missing_profiles): Fix logic looking for zero profiles.
11779 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11781 PR middle-end/83977
11782 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11783 on functions with #pragma omp declare simd or functions with simd
11785 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11786 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11787 Remove trailing \n from warning_at calls.
11789 2018-01-25 Tom de Vries <tom@codesourcery.com>
11792 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11793 for neutered workers.
11795 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11798 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11799 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11801 2018-01-24 Jeff Law <law@redhat.com>
11804 * i386.c (get_probe_interval): Move to earlier point.
11805 (ix86_compute_frame_layout): If -fstack-clash-protection and
11806 the frame is larger than the probe interval, then use pushes
11807 to save registers rather than reg->mem moves.
11808 (ix86_expand_prologue): Remove conditional for int_registers_saved
11811 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11814 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11815 min/max for never referenced object.
11817 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11819 PR middle-end/83977
11820 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11822 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11823 attributes from DECL_ATTRIBUTES (decl) without affecting
11824 DECL_ATTRIBUTES (current_function_decl).
11825 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11826 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11828 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11830 PR tree-optimization/83979
11831 * fold-const.c (fold_comparison): Use constant_boolean_node
11832 instead of boolean_{true,false}_node.
11834 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11836 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11839 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11841 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11842 Simplify the clause that sets the length attribute.
11843 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11844 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11845 clause that sets the length attribute.
11846 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11848 2018-01-24 Tom de Vries <tom@codesourcery.com>
11851 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11852 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11853 Add strict parameter.
11854 (prevent_branch_around_nothing): Insert dummy insn between branch to
11855 label and label with no ptx insn inbetween.
11856 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11858 2018-01-24 Tom de Vries <tom@codesourcery.com>
11861 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11862 for neutered threads in warp.
11863 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11865 2018-01-24 Richard Biener <rguenther@suse.de>
11867 PR tree-optimization/83176
11868 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11871 2018-01-24 Richard Biener <rguenther@suse.de>
11873 PR tree-optimization/82819
11874 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11875 code generating pluses that are no-ops in the target precision.
11877 2018-01-24 Richard Biener <rguenther@suse.de>
11879 PR middle-end/84000
11880 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11882 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11884 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11885 to merge probabilities.
11886 * predict.c (probably_never_executed): Also mark as cold functions
11887 with global 0 profile and guessed local profile.
11888 * profile-count.c (profile_probability::combine_with_count): New
11890 * profile-count.h (profile_probability::operator*,
11891 profile_probability::operator*=, profile_probability::operator/,
11892 profile_probability::operator/=): Reduce precision to adjusted
11893 and set value to guessed on contradictory divisions.
11894 (profile_probability::combine_with_freq): Remove.
11895 (profile_probability::combine_wiht_count): Declare.
11896 (profile_count::force_nonzero):: Set to adjusted.
11897 (profile_count::probability_in):: Set quality to adjusted.
11898 * tree-ssa-tail-merge.c (replace_block_by): Use
11899 combine_with_count.
11901 2018-01-23 Andrew Waterman <andrew@sifive.com>
11902 Jim Wilson <jimw@sifive.com>
11904 * config/riscv/riscv.c (riscv_stack_boundary): New.
11905 (riscv_option_override): Set riscv_stack_boundary. Handle
11906 riscv_preferred_stack_boundary_arg.
11907 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11908 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11909 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11910 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11911 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11912 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11914 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11917 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11918 of struct ix86_frame.
11919 (ix86_expand_epilogue): Likewise. Add a local variable for
11920 the reg_save_offset field in struct ix86_frame.
11922 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11924 PR tree-optimization/82604
11925 * tree-loop-distribution.c (enum partition_kind): New enum item
11926 PKIND_PARTIAL_MEMSET.
11927 (partition_builtin_p): Support above new enum item.
11928 (generate_code_for_partition): Ditto.
11929 (compute_access_range): Differentiate cases that equality can be
11930 proven at all loops, the innermost loops or no loops.
11931 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11932 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11933 (finalize_partitions, distribute_loop): Don't fuse partition of
11934 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11935 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11936 parloop is enabled.
11938 2018-01-23 Martin Liska <mliska@suse.cz>
11940 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11941 order to ignore the predictor.
11942 (PRED_POLYMORPHIC_CALL): Likewise.
11943 (PRED_RECURSIVE_CALL): Likewise.
11945 2018-01-23 Martin Liska <mliska@suse.cz>
11947 * tree-profile.c (tree_profiling): Print function header to
11948 aware reader which function we are working on.
11949 * value-prof.c (gimple_find_values_to_profile): Do not print
11950 not interesting value histograms.
11952 2018-01-23 Martin Liska <mliska@suse.cz>
11954 * profile-count.h (enum profile_quality): Add
11955 profile_uninitialized as the first value. Do not number values
11956 as they are zero based.
11957 (profile_count::verify): Update sanity check.
11958 (profile_probability::verify): Likewise.
11960 2018-01-23 Nathan Sidwell <nathan@acm.org>
11962 * doc/invoke.texi (ffor-scope): Deprecate.
11964 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11966 PR tree-optimization/83510
11967 * domwalk.c (set_all_edges_as_executable): New function.
11968 (dom_walker::dom_walker): Convert bool param
11969 "skip_unreachable_blocks" to enum reachability. Move setup of
11970 edge flags to set_all_edges_as_executable and only do it when
11971 reachability is REACHABLE_BLOCKS.
11972 * domwalk.h (enum dom_walker::reachability): New enum.
11973 (dom_walker::dom_walker): Convert bool param
11974 "skip_unreachable_blocks" to enum reachability.
11975 (set_all_edges_as_executable): New decl.
11976 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11977 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11979 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11980 but converting true to REACHABLE_BLOCKS.
11981 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11983 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11984 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11985 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11987 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11988 if check_all_array_refs will be called.
11990 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11992 * tree.c (selftest::test_location_wrappers): Add more test
11995 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11997 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11998 (selftest::test_bit_in_range): Likewise.
12000 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12003 * doc/sourcebuild.texi (vect_float): Say that the selector
12004 only describes the situation when -funsafe-math-optimizations is on.
12005 (vect_float_strict): Document.
12007 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12009 PR tree-optimization/83965
12010 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12011 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12012 instead of checking only for a reduction.
12013 (vect_recog_widen_sum_pattern): Likewise.
12015 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12017 * predict.c (probably_never_executed): Only use precise profile info.
12018 (compute_function_frequency): Skip after inlining hack since we now
12019 have quality checking.
12021 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12023 * profile-count.h (profile_probability::very_unlikely,
12024 profile_probability::unlikely, profile_probability::even): Set
12025 precision to guessed.
12027 2018-01-23 Richard Biener <rguenther@suse.de>
12029 PR tree-optimization/83963
12030 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12031 Properly terminate dominator walk when crossing the exit edge not
12032 when visiting its source block.
12034 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12037 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12038 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12040 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12042 PR tree-optimization/83957
12043 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12044 semicolon after for body surrounded by braces.
12046 PR tree-optimization/83081
12047 * profile-count.h (profile_probability::split): New method.
12048 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12049 Use profile_probability::split.
12050 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12051 when splitting a single conditional jump into 2.
12053 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12055 PR tree-optimization/69452
12056 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12059 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12061 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12062 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12063 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12065 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12067 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12069 * config/rl78/rl78.md (movdi): New define_expand.
12070 * config/rl78/rl78.c (rl78_split_movdi): New function.
12072 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12075 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12077 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12078 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12079 128-bit to produce an UNSPEC move to get the double word with the
12080 signbit and then a shift directly to do signbit.
12081 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12082 implementation with a new version that just does either a direct
12083 move or a regular move. Move memory interface to separate insns.
12084 Move insns so they are next to the expander.
12085 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12086 with signbit move. Split big and little endian case.
12087 (signbit<mode>2_dm_mem_le): Likewise.
12088 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12089 (signbit<mode>2_dm2): Likewise.
12091 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12093 * config/rl78/rl78.md (anddi3): New define_expand.
12095 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12097 * config/rl78/rl78.md (umindi3): New define_expand.
12099 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12101 * config/rl78/rl78.md (smindi3): New define_expand.
12103 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12105 * config/rl78/rl78.md (smaxdi3): New define_expand.
12107 2018-01-22 Carl Love <cel@us.ibm.com>
12109 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12110 LVX_V1TI): Add macro expansion.
12111 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12112 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12113 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12114 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12115 Change check to determine if the instruction is a byte reversing
12116 entry. Fix typo in comment.
12117 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12118 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12119 Add def_builtin calls for new builtins.
12120 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12121 Add define_insn expansion.
12123 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12125 * config/rl78/rl78.md (umaxdi3): New define_expand.
12127 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12129 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12130 for non-QImode registers.
12132 2018-01-22 Richard Biener <rguenther@suse.de>
12134 PR tree-optimization/83963
12135 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12136 including the loop exit block.
12137 (scop_detection::merge_sese): Likewise.
12138 (scop_detection::add_scop): Do it here instead.
12140 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12142 * doc/sourcebuild.texi (arm_softfloat): Document.
12144 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12147 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12148 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12149 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12151 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12152 David Edelsohn <dje.gcc@gmail.com>
12155 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12156 Change "crset eq" to "crset 2".
12157 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12158 (*call_indirect_aix<mode>_nospec): Likewise.
12159 (*call_value_indirect_aix<mode>_nospec): Likewise.
12160 (*call_indirect_elfv2<mode>_nospec): Likewise.
12161 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12162 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12163 change assembly output from . to $.
12164 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12165 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12166 (*tablejump<mode>_internal1_nospec): Likewise.
12168 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12171 * config/sh/sh_optimize_sett_clrt.cc:
12172 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12174 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12176 PR tree-optimization/83940
12177 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12178 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12179 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12180 use it to pass back the definition type.
12181 (vect_check_store_rhs): Likewise rhs_dt_out.
12182 (vect_build_gather_load_calls): Add a mask_dt argument and use
12183 it instead of a call to vect_is_simple_use.
12184 (vectorizable_store): Update calls to vect_check_load_store_mask
12185 and vect_check_store_rhs. Use the dt returned by the latter instead
12186 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12187 instead of calls to vect_is_simple_use. Pass the scalar rather
12188 than the vector operand to vect_is_simple_use when handling
12189 second and subsequent copies of an rhs value.
12190 (vectorizable_load): Update calls to vect_check_load_store_mask
12191 and vect_build_gather_load_calls. Use the cached mask_dt and
12192 gs_info.offset_dt instead of calls to vect_is_simple_use.
12194 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12196 PR middle-end/83945
12197 * tree-emutls.c: Include gimplify.h.
12198 (lower_emutls_2): New function.
12199 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12200 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12201 it before further processing.
12204 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12205 UINTVAL (trueop1) instead of INTVAL (op1).
12207 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12211 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12212 INCOMING_FRAME_SP_OFFSET if not defined.
12213 (scan_trace): Add ENTRY argument. If true and
12214 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12215 emit a note to adjust the CFA offset.
12216 (create_cfi_notes): Adjust scan_trace callers.
12217 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12218 INCOMING_FRAME_SP_OFFSET in the CIE.
12219 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12220 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12222 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12223 * doc/tm.texi: Regenerated.
12225 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12227 PR rtl-optimization/83147
12228 * lra-constraints.c (remove_inheritance_pseudos): Use
12229 lra_substitute_pseudo_within_insn.
12231 2018-01-19 Tom de Vries <tom@codesourcery.com>
12232 Cesar Philippidis <cesar@codesourcery.com>
12235 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12237 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12240 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12241 spaces for function labels.
12243 2018-01-19 Martin Liska <mliska@suse.cz>
12245 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12246 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12247 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12248 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12249 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12250 (PRED_CONST_RETURN): Change from 69 to 65.
12251 (PRED_NULL_RETURN): Change from 91 to 71.
12252 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12253 (PRED_LOOP_GUARD): Change from 66 to 73.
12255 2018-01-19 Martin Liska <mliska@suse.cz>
12257 * predict.c (predict_insn_def): Add new assert.
12258 (struct branch_predictor): Change type to signed integer.
12259 (test_prediction_value_range): Amend test to cover
12260 PROB_UNINITIALIZED.
12261 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12262 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12263 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12264 (PRED_LOOP_IV_COMPARE): Likewise.
12265 * predict.h (PROB_UNINITIALIZED): Define new constant.
12267 2018-01-19 Martin Liska <mliska@suse.cz>
12269 * predict.c (dump_prediction): Add new format for
12270 analyze_brprob.py script which is enabled with -details
12272 * profile-count.h (precise_p): New function.
12274 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12276 PR tree-optimization/83922
12277 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12278 there are no statements that need masking.
12279 (vect_active_double_reduction_p): New function.
12280 (vect_analyze_loop_operations): Use it when handling phis that
12281 are not in the loop header.
12283 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12285 PR tree-optimization/83914
12286 * tree-vect-loop.c (vectorizable_induction): Don't convert
12287 init_expr or apply the peeling adjustment for inductions
12288 that are nested within the vectorized loop.
12290 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12292 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12295 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12299 * function.h (gimplify_parameters): Add gimple_seq * argument.
12300 * function.c: Include gimple.h and options.h.
12301 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12302 for the added local temporaries if needed.
12303 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12304 if there are any parameter cleanups, wrap whole body into a
12305 try/finally with the cleanups.
12307 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12310 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12311 Use GET_MODE_CLASS for scalar floating point.
12313 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12317 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12318 Fix call of call_cgraph_insertion_hooks.
12320 2018-01-18 Martin Sebor <msebor@redhat.com>
12322 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12324 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12327 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12330 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12333 * common.opt: (-ffile-prefix-map): New option.
12334 * opts.c (common_handle_option): Defer it.
12335 * opts-global.c (handle_common_deferred_options): Handle it.
12336 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12337 * file-prefix-map.h: New file.
12338 (remap_debug_filename, add_debug_prefix_map): ...here.
12339 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12340 * final.c (debug_prefix_map, add_debug_prefix_map
12341 remap_debug_filename): Move to...
12342 * file-prefix-map.c: New file.
12343 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12344 generalize, get rid of alloca(), use strrchr() instead of strchr().
12345 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12346 Implement in terms of add_prefix_map().
12347 (remap_macro_filename, remap_debug_filename): Implement in term of
12349 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12350 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12351 * dbxout.c: Include file-prefix-map.h.
12352 * varasm.c: Likewise.
12353 * vmsdbgout.c: Likewise.
12354 * xcoffout.c: Likewise.
12355 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12356 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12357 * doc/invoke.texi (-ffile-prefix-map): Document.
12358 (-fdebug-prefix-map): Update description.
12360 2018-01-18 Martin Liska <mliska@suse.cz>
12362 * config/i386/i386.c (indirect_thunk_name): Document that also
12364 (output_indirect_thunk): Document why both instructions
12365 (pause and lfence) are generated.
12367 2018-01-18 Richard Biener <rguenther@suse.de>
12369 PR tree-optimization/83887
12370 * graphite-scop-detection.c
12371 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12372 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12373 (scop_detection::merge_sese): Re-implement with a flood-fill
12374 algorithm that properly finds a SESE region if it exists.
12376 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12379 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12380 pointer_diff optimizations use view_convert instead of convert.
12382 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12384 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12385 Generate different code for -mno-speculate-indirect-jumps.
12386 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12387 (*call_indirect_aix<mode>): Disable for
12388 -mno-speculate-indirect-jumps.
12389 (*call_indirect_aix<mode>_nospec): New define_insn.
12390 (*call_value_indirect_aix<mode>): Disable for
12391 -mno-speculate-indirect-jumps.
12392 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12393 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12394 -mno-speculate-indirect-jumps.
12395 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12397 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12399 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12400 long double type, set the flags for noting the default long double
12401 type, even if we don't pass or return a long double type.
12403 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12406 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12409 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12411 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12412 support for merge[hl].
12413 (fold_mergehl_helper): New helper function.
12414 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12415 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12416 (altivec_vmrglw_direct): Add xxmrglw insn.
12418 2018-01-17 Andrew Waterman <andrew@sifive.com>
12420 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12421 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12423 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12426 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12427 call the lto_location_cache before reading the
12428 DECL_SOURCE_LOCATION of the types.
12430 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12431 Richard Sandiford <richard.sandiford@linaro.org>
12433 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12434 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12435 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12436 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12437 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12439 * config/aarch64/constraints.md (aarch64_movti_operand):
12441 * config/aarch64/predicates.md (Uti): Add new constraint.
12443 2018-01-17 Carl Love <cel@us.ibm.com>
12445 * config/rs6000/vsx.md (define_expand xl_len_r,
12446 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12447 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12449 (define_expand, define_insn): Move the shift left from the
12450 define_insn to the define_expand for lxvl and stxvl instructions.
12451 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12452 and XL_LEN_R definitions to PURE.
12454 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12456 * config/i386/i386.c (indirect_thunk_name): Declare regno
12457 as unsigned int. Compare regno with INVALID_REGNUM.
12458 (output_indirect_thunk): Ditto.
12459 (output_indirect_thunk_function): Ditto.
12460 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12461 in the call to output_indirect_thunk_function.
12463 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12465 PR middle-end/83884
12466 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12467 rather than the size of inner_type to determine the stack slot size
12468 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12470 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12473 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12476 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12478 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12479 endian Linux systems to optionally enable multilibs for selecting
12480 the long double type if the user configured an explicit type.
12481 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12482 have no long double multilibs if not defined.
12483 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12484 warn if the user used -mabi={ieee,ibm}longdouble and we built
12485 multilibs for long double.
12486 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12487 appropriate multilib option.
12488 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12490 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12491 for building long double multilibs.
12492 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12494 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12496 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12499 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12501 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12504 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12507 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12510 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12512 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12513 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12515 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12517 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12518 different rtl trees depending on TARGET_64BIT.
12519 (rs6000_gen_lvx): Likewise.
12521 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12523 * config/visium/visium.md (nop): Tweak comment.
12524 (hazard_nop): Likewise.
12526 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12528 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12529 -mspeculate-indirect-jumps.
12530 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12531 for -mno-speculate-indirect-jumps.
12532 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12533 (*call_value_indirect_elfv2<mode>): Disable for
12534 -mno-speculate-indirect-jumps.
12535 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12536 (indirect_jump): Emit different RTL for
12537 -mno-speculate-indirect-jumps.
12538 (*indirect_jump<mode>): Disable for
12539 -mno-speculate-indirect-jumps.
12540 (*indirect_jump<mode>_nospec): New define_insn.
12541 (tablejump): Emit different RTL for
12542 -mno-speculate-indirect-jumps.
12543 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12544 (tablejumpsi_nospec): New define_expand.
12545 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12546 (tablejumpdi_nospec): New define_expand.
12547 (*tablejump<mode>_internal1): Disable for
12548 -mno-speculate-indirect-jumps.
12549 (*tablejump<mode>_internal1_nospec): New define_insn.
12550 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12553 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12555 * caller-save.c (insert_save): Drop unnecessary parameter. All
12558 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12559 Richard Biener <rguenth@suse.de>
12562 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12563 return early, inline manually is_gimple_sizepos. Make sure if we
12564 call gimplify_expr we don't end up with a gimple constant.
12565 * tree.c (variably_modified_type_p): Don't return true for
12566 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12567 * gimplify.h (is_gimple_sizepos): Remove.
12569 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12571 PR tree-optimization/83857
12572 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12573 vectorizable_live_operation for pure SLP statements.
12574 (vectorizable_live_operation): Handle PHIs.
12576 2018-01-16 Richard Biener <rguenther@suse.de>
12578 PR tree-optimization/83867
12579 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12580 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12582 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12585 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12586 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12587 If off is not INTEGER_CST, issue a may not be aligned warning
12588 rather than isn't aligned. Use isn%'t rather than isn't.
12589 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12591 <case MULT_EXPR>: Improve the case when bottom and one of the
12592 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12593 operand, in that case check if the other operand is multiple of
12594 bottom divided by the INTEGER_CST operand.
12596 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12599 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12600 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12601 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12602 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12603 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12604 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12605 (pa_function_arg_size): New function.
12607 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12609 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12610 in a separate statement.
12612 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12614 PR tree-optimization/83847
12615 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12616 group gathers and scatters.
12618 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12620 PR rtl-optimization/86620
12621 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12623 PR rtl-optimization/83213
12624 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12625 to last if both are JUMP_INSNs.
12627 PR tree-optimization/83843
12628 * gimple-ssa-store-merging.c
12629 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12630 store_immediate_info for bswap/nop orig_stores.
12632 2018-01-15 Andrew Waterman <andrew@sifive.com>
12634 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12636 <UDIV>: Increase cost if !TARGET_DIV.
12638 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12640 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12641 (define_attr "cr_logical_3op"): New.
12642 (cceq_ior_compare): Adjust.
12643 (cceq_ior_compare_complement): Adjust.
12644 (*cceq_rev_compare): Adjust.
12645 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12646 (is_cracked_insn): Adjust.
12647 (insn_must_be_first_in_group): Adjust.
12648 * config/rs6000/40x.md: Adjust.
12649 * config/rs6000/440.md: Adjust.
12650 * config/rs6000/476.md: Adjust.
12651 * config/rs6000/601.md: Adjust.
12652 * config/rs6000/603.md: Adjust.
12653 * config/rs6000/6xx.md: Adjust.
12654 * config/rs6000/7450.md: Adjust.
12655 * config/rs6000/7xx.md: Adjust.
12656 * config/rs6000/8540.md: Adjust.
12657 * config/rs6000/cell.md: Adjust.
12658 * config/rs6000/e300c2c3.md: Adjust.
12659 * config/rs6000/e500mc.md: Adjust.
12660 * config/rs6000/e500mc64.md: Adjust.
12661 * config/rs6000/e5500.md: Adjust.
12662 * config/rs6000/e6500.md: Adjust.
12663 * config/rs6000/mpc.md: Adjust.
12664 * config/rs6000/power4.md: Adjust.
12665 * config/rs6000/power5.md: Adjust.
12666 * config/rs6000/power6.md: Adjust.
12667 * config/rs6000/power7.md: Adjust.
12668 * config/rs6000/power8.md: Adjust.
12669 * config/rs6000/power9.md: Adjust.
12670 * config/rs6000/rs64.md: Adjust.
12671 * config/rs6000/titan.md: Adjust.
12673 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12675 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12676 ix86_indirect_branch_register logic.
12678 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12680 * config/i386/constraints.md (Bs): Update
12681 ix86_indirect_branch_register check. Don't check
12682 ix86_indirect_branch_register with GOT_memory_operand.
12684 * config/i386/predicates.md (GOT_memory_operand): Don't check
12685 ix86_indirect_branch_register here.
12686 (GOT32_symbol_operand): Likewise.
12688 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12690 * config/i386/predicates.md (constant_call_address_operand):
12691 Rewrite ix86_indirect_branch_register logic.
12692 (sibcall_insn_operand): Likewise.
12694 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12696 * config/i386/constraints.md (Bs): Replace
12697 ix86_indirect_branch_thunk_register with
12698 ix86_indirect_branch_register.
12700 * config/i386/i386.md (indirect_jump): Likewise.
12701 (tablejump): Likewise.
12702 (*sibcall_memory): Likewise.
12703 (*sibcall_value_memory): Likewise.
12704 Peepholes of indirect call and jump via memory: Likewise.
12705 * config/i386/i386.opt: Likewise.
12706 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12707 (GOT_memory_operand): Likewise.
12708 (call_insn_operand): Likewise.
12709 (sibcall_insn_operand): Likewise.
12710 (GOT32_symbol_operand): Likewise.
12712 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12714 PR middle-end/83837
12715 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12716 type rather than type addr's type points to.
12717 (expand_omp_atomic_mutex): Likewise.
12718 (expand_omp_atomic): Likewise.
12720 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12723 * config/i386/i386.c (output_indirect_thunk_function): Use
12724 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12725 for __x86_return_thunk.
12727 2018-01-15 Richard Biener <rguenther@suse.de>
12729 PR middle-end/83850
12730 * expmed.c (extract_bit_field_1): Fix typo.
12732 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12735 * config/arm/iterators.md (VF): New mode iterator.
12736 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12737 Remove integer-related logic from pattern.
12738 (neon_vabd<mode>_3): Likewise.
12740 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12742 PR middle-end/82694
12743 * common.opt (fstrict-overflow): No longer an alias.
12744 (fwrapv-pointer): New option.
12745 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12746 also for pointer types based on flag_wrapv_pointer.
12747 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12748 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12749 opts->x_flag_wrapv got set.
12750 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12751 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12752 POINTER_TYPE_OVERFLOW_UNDEFINED.
12753 * match.pd: Likewise in address comparison pattern.
12754 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12756 2018-01-15 Richard Biener <rguenther@suse.de>
12759 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12760 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12761 Reset type names to their identifier if their TYPE_DECL doesn't
12762 have linkage (and thus is used for ODR and devirt).
12763 (save_debug_info_for_decl): Remove.
12764 (save_debug_info_for_type): Likewise.
12765 (add_tree_to_fld_list): Adjust.
12766 * tree-pretty-print.c (dump_generic_node): Make dumping of
12767 type names more robust.
12769 2018-01-15 Richard Biener <rguenther@suse.de>
12771 * BASE-VER: Bump to 8.0.1.
12773 2018-01-14 Martin Sebor <msebor@redhat.com>
12776 * builtins.c (check_access): Avoid warning when the no-warning bit
12779 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12781 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12782 * ira-color (allocno_hard_regs_compare): Likewise.
12784 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12787 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12788 Use .pushsection/.popsection.
12790 2018-01-14 Martin Sebor <msebor@redhat.com>
12793 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12795 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12797 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12798 entry from extra_headers.
12799 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12800 extra_headers, make the list bitwise identical to the i?86-*-* one.
12802 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12804 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12805 -mcmodel=large with -mindirect-branch=thunk,
12806 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12807 -mfunction-return=thunk-extern.
12808 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12809 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12810 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12812 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12814 * config/i386/i386.c (print_reg): Print the name of the full
12815 integer register without '%'.
12816 (ix86_print_operand): Handle 'V'.
12817 * doc/extend.texi: Document 'V' modifier.
12819 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12821 * config/i386/constraints.md (Bs): Disallow memory operand for
12822 -mindirect-branch-register.
12824 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12825 (GOT_memory_operand): Likewise.
12826 (call_insn_operand): Likewise.
12827 (sibcall_insn_operand): Likewise.
12828 (GOT32_symbol_operand): Likewise.
12829 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12830 for -mindirect-branch-register.
12831 (tablejump): Likewise.
12832 (*sibcall_memory): Likewise.
12833 (*sibcall_value_memory): Likewise.
12834 Disallow peepholes of indirect call and jump via memory for
12835 -mindirect-branch-register.
12836 (*call_pop): Replace m with Bw.
12837 (*call_value_pop): Likewise.
12838 (*sibcall_pop_memory): Replace m with Bs.
12839 * config/i386/i386.opt (mindirect-branch-register): New option.
12840 * doc/invoke.texi: Document -mindirect-branch-register option.
12842 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12844 * config/i386/i386-protos.h (ix86_output_function_return): New.
12845 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12846 set function_return_type.
12847 (indirect_thunk_name): Add ret_p to indicate thunk for function
12849 (output_indirect_thunk_function): Pass false to
12850 indirect_thunk_name.
12851 (ix86_output_indirect_branch_via_reg): Likewise.
12852 (ix86_output_indirect_branch_via_push): Likewise.
12853 (output_indirect_thunk_function): Create alias for function
12854 return thunk if regno < 0.
12855 (ix86_output_function_return): New function.
12856 (ix86_handle_fndecl_attribute): Handle function_return.
12857 (ix86_attribute_table): Add function_return.
12858 * config/i386/i386.h (machine_function): Add
12859 function_return_type.
12860 * config/i386/i386.md (simple_return_internal): Use
12861 ix86_output_function_return.
12862 (simple_return_internal_long): Likewise.
12863 * config/i386/i386.opt (mfunction-return=): New option.
12864 (indirect_branch): Mention -mfunction-return=.
12865 * doc/extend.texi: Document function_return function attribute.
12866 * doc/invoke.texi: Document -mfunction-return= option.
12868 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12870 * config/i386/i386-opts.h (indirect_branch): New.
12871 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12872 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12873 with local indirect jump when converting indirect call and jump.
12874 (ix86_set_indirect_branch_type): New.
12875 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12876 (indirectlabelno): New.
12877 (indirect_thunk_needed): Likewise.
12878 (indirect_thunk_bnd_needed): Likewise.
12879 (indirect_thunks_used): Likewise.
12880 (indirect_thunks_bnd_used): Likewise.
12881 (INDIRECT_LABEL): Likewise.
12882 (indirect_thunk_name): Likewise.
12883 (output_indirect_thunk): Likewise.
12884 (output_indirect_thunk_function): Likewise.
12885 (ix86_output_indirect_branch_via_reg): Likewise.
12886 (ix86_output_indirect_branch_via_push): Likewise.
12887 (ix86_output_indirect_branch): Likewise.
12888 (ix86_output_indirect_jmp): Likewise.
12889 (ix86_code_end): Call output_indirect_thunk_function if needed.
12890 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12892 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12893 (ix86_attribute_table): Add indirect_branch.
12894 * config/i386/i386.h (machine_function): Add indirect_branch_type
12895 and has_local_indirect_jump.
12896 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12898 (tablejump): Likewise.
12899 (*indirect_jump): Use ix86_output_indirect_jmp.
12900 (*tablejump_1): Likewise.
12901 (simple_return_indirect_internal): Likewise.
12902 * config/i386/i386.opt (mindirect-branch=): New option.
12903 (indirect_branch): New.
12906 (thunk-inline): Likewise.
12907 (thunk-extern): Likewise.
12908 * doc/extend.texi: Document indirect_branch function attribute.
12909 * doc/invoke.texi: Document -mindirect-branch= option.
12911 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12914 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12916 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12918 * ipa-inline.c (want_inline_small_function_p): Return false if
12919 inlining has already failed with CIF_FINAL_ERROR.
12920 (update_caller_keys): Call want_inline_small_function_p before
12922 (update_callee_keys): Likewise.
12924 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12926 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12928 (rs6000_quadword_masked_address_p): Likewise.
12929 (quad_aligned_load_p): Likewise.
12930 (quad_aligned_store_p): Likewise.
12931 (const_load_sequence_p): Add comment to describe the outer-most loop.
12932 (mimic_memory_attributes_and_flags): New function.
12933 (rs6000_gen_stvx): Likewise.
12934 (replace_swapped_aligned_store): Likewise.
12935 (rs6000_gen_lvx): Likewise.
12936 (replace_swapped_aligned_load): Likewise.
12937 (replace_swapped_load_constant): Capitalize argument name in
12938 comment describing this function.
12939 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12940 and stores that access quad-word aligned addresses and replace
12941 with stvx or lvx instructions when appropriate.
12942 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12943 New function prototype.
12944 (rs6000_quadword_masked_address_p): Likewise.
12945 (rs6000_gen_lvx): Likewise.
12946 (rs6000_gen_stvx): Likewise.
12947 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12948 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12949 when memory address is aligned.
12950 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12951 this split to select lvx instruction when memory address is aligned.
12952 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12953 instruction when memory address is aligned.
12954 (*vsx_le_perm_load_v16qi): Likewise.
12955 (four unnamed splitters): Modify to select the stvx instruction
12956 when memory is aligned.
12958 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12960 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12961 which appears in the queue multiple times.
12963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12964 Alan Hayward <alan.hayward@arm.com>
12965 David Sherwood <david.sherwood@arm.com>
12967 * tree-vectorizer.h (vec_lower_bound): New structure.
12968 (_loop_vec_info): Add check_nonzero and lower_bounds.
12969 (LOOP_VINFO_CHECK_NONZERO): New macro.
12970 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12971 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12972 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12973 fields. Make seg_len the distance travelled, not including the
12975 (dr_direction_indicator): Declare.
12976 (dr_zero_step_indicator): Likewise.
12977 (dr_known_forward_stride_p): Likewise.
12978 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12980 (runtime_alias_check_p): Allow runtime alias checks with
12982 (operator ==): Compare access_size and align.
12983 (prune_runtime_alias_test_list): Rework for new distinction between
12984 the access_size and seg_len.
12985 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12987 (get_segment_min_max): New function.
12988 (create_intersect_range_checks): Use it.
12989 (dr_step_indicator): New function.
12990 (dr_direction_indicator): Likewise.
12991 (dr_zero_step_indicator): Likewise.
12992 (dr_known_forward_stride_p): Likewise.
12993 * tree-loop-distribution.c (data_ref_segment_size): Return
12994 DR_STEP * (niters - 1).
12995 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12997 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12998 (vect_preserves_scalar_order_p): New function, split out from...
12999 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13000 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13001 (vect_vfa_access_size): New function.
13002 (vect_vfa_align): Likewise.
13003 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13004 (dump_lower_bound): New function.
13005 (vect_check_lower_bound): Likewise.
13006 (vect_small_gap_p): Likewise.
13007 (vectorizable_with_step_bound_p): Likewise.
13008 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13009 depencies if the vectorization factor is 1. Convert the checks
13010 for nonzero steps into checks on the bounds of DR_STEP. Try using
13011 a bunds check for variable steps if the minimum required step is
13012 relatively small. Update calls to the dr_with_seg_len
13013 constructor and to vect_compile_time_alias.
13014 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13016 (vect_loop_versioning): Call it.
13017 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13019 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13022 Alan Hayward <alan.hayward@arm.com>
13023 David Sherwood <david.sherwood@arm.com>
13025 * doc/sourcebuild.texi (vect_scatter_store): Document.
13026 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13028 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13030 * genopinit.c (main): Add supports_vec_scatter_store and
13031 supports_vec_scatter_store_cached to target_optabs.
13032 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13033 IFN_MASK_SCATTER_STORE.
13034 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13036 * internal-fn.h (internal_store_fn_p): Declare.
13037 (internal_fn_stored_value_index): Likewise.
13038 * internal-fn.c (scatter_store_direct): New macro.
13039 (expand_scatter_store_optab_fn): New function.
13040 (direct_scatter_store_optab_supported_p): New macro.
13041 (internal_store_fn_p): New function.
13042 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13043 IFN_MASK_SCATTER_STORE.
13044 (internal_fn_mask_index): Likewise.
13045 (internal_fn_stored_value_index): New function.
13046 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13047 for scatter stores.
13048 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13049 * optabs-query.c (supports_vec_scatter_store_p): New function.
13050 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13051 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13052 true for scatter stores.
13053 (vect_gather_scatter_fn_p): Handle scatter stores too.
13054 (vect_check_gather_scatter): Consider using scatter stores if
13055 supports_vec_scatter_store_p.
13056 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13057 scatter stores too.
13058 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13059 internal_fn_stored_value_index.
13060 (check_load_store_masking): Handle scatter stores too.
13061 (vect_get_store_rhs): Make public.
13062 (vectorizable_call): Use internal_store_fn_p.
13063 (vectorizable_store): Handle scatter store internal functions.
13064 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13065 when deciding whether the end of the group has been reached.
13066 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13067 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13068 (mask_scatter_store<mode>): New insns.
13070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13071 Alan Hayward <alan.hayward@arm.com>
13072 David Sherwood <david.sherwood@arm.com>
13074 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13075 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13076 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13078 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13079 Use vect_truncate_gather_scatter_offset if we can't treat the
13080 operation as a normal gather load or scatter store.
13081 (get_group_load_store_type): Take the gather_scatter_info
13082 as argument. Try using a gather load or scatter store for
13083 single-element groups.
13084 (get_load_store_type): Update calls to get_group_load_store_type
13085 and vect_use_strided_gather_scatters_p.
13087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13088 Alan Hayward <alan.hayward@arm.com>
13089 David Sherwood <david.sherwood@arm.com>
13091 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13092 optional tree argument.
13093 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13095 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13096 but continue to use the current value as a fallback.
13097 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13098 to compare the updates.
13099 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13100 (get_load_store_type): Use it when handling a strided access.
13101 (vect_get_strided_load_store_ops): New function.
13102 (vect_get_data_ptr_increment): Likewise.
13103 (vectorizable_load): Handle strided gather loads. Always pass
13104 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13107 Alan Hayward <alan.hayward@arm.com>
13108 David Sherwood <david.sherwood@arm.com>
13110 * doc/md.texi (gather_load@var{m}): Document.
13111 (mask_gather_load@var{m}): Likewise.
13112 * genopinit.c (main): Add supports_vec_gather_load and
13113 supports_vec_gather_load_cached to target_optabs.
13114 * optabs-tree.c (init_tree_optimization_optabs): Use
13115 ggc_cleared_alloc to allocate target_optabs.
13116 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13117 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13119 * internal-fn.h (internal_load_fn_p): Declare.
13120 (internal_gather_scatter_fn_p): Likewise.
13121 (internal_fn_mask_index): Likewise.
13122 (internal_gather_scatter_fn_supported_p): Likewise.
13123 * internal-fn.c (gather_load_direct): New macro.
13124 (expand_gather_load_optab_fn): New function.
13125 (direct_gather_load_optab_supported_p): New macro.
13126 (direct_internal_fn_optab): New function.
13127 (internal_load_fn_p): Likewise.
13128 (internal_gather_scatter_fn_p): Likewise.
13129 (internal_fn_mask_index): Likewise.
13130 (internal_gather_scatter_fn_supported_p): Likewise.
13131 * optabs-query.c (supports_at_least_one_mode_p): New function.
13132 (supports_vec_gather_load_p): Likewise.
13133 * optabs-query.h (supports_vec_gather_load_p): Declare.
13134 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13135 and memory_type field.
13136 (NUM_PATTERNS): Bump to 15.
13137 * tree-vect-data-refs.c: Include internal-fn.h.
13138 (vect_gather_scatter_fn_p): New function.
13139 (vect_describe_gather_scatter_call): Likewise.
13140 (vect_check_gather_scatter): Try using internal functions for
13141 gather loads. Recognize existing calls to a gather load function.
13142 (vect_analyze_data_refs): Consider using gather loads if
13143 supports_vec_gather_load_p.
13144 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13145 (vect_get_gather_scatter_offset_type): Likewise.
13146 (vect_convert_mask_for_vectype): Likewise.
13147 (vect_add_conversion_to_patterm): Likewise.
13148 (vect_try_gather_scatter_pattern): Likewise.
13149 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13150 (vect_vect_recog_func_ptrs): Add it.
13151 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13152 internal_fn_mask_index and internal_gather_scatter_fn_p.
13153 (check_load_store_masking): Take the gather_scatter_info as an
13154 argument and handle gather loads.
13155 (vect_get_gather_scatter_ops): New function.
13156 (vectorizable_call): Check internal_load_fn_p.
13157 (vectorizable_load): Likewise. Handle gather load internal
13159 (vectorizable_store): Update call to check_load_store_masking.
13160 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13161 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13162 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13163 (aarch64_gather_scale_operand_d): New predicates.
13164 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13165 (mask_gather_load<mode>): New insns.
13167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13168 Alan Hayward <alan.hayward@arm.com>
13169 David Sherwood <david.sherwood@arm.com>
13171 * optabs.def (fold_left_plus_optab): New optab.
13172 * doc/md.texi (fold_left_plus_@var{m}): Document.
13173 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13174 * internal-fn.c (fold_left_direct): Define.
13175 (expand_fold_left_optab_fn): Likewise.
13176 (direct_fold_left_optab_supported_p): Likewise.
13177 * fold-const-call.c (fold_const_fold_left): New function.
13178 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13179 * tree-parloops.c (valid_reduction_p): New function.
13180 (gather_scalar_reductions): Use it.
13181 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13182 (vect_finish_replace_stmt): Declare.
13183 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13184 (needs_fold_left_reduction_p): New function, split out from...
13185 (vect_is_simple_reduction): ...here. Accept reductions that
13186 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13187 (vect_force_simple_reduction): Also store the reduction type in
13188 the assignment's STMT_VINFO_REDUC_TYPE.
13189 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13190 (merge_with_identity): New function.
13191 (vect_expand_fold_left): Likewise.
13192 (vectorize_fold_left_reduction): Likewise.
13193 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13194 scalar phi in place for it. Check for target support and reject
13195 cases that would reassociate the operation. Defer the transform
13196 phase to vectorize_fold_left_reduction.
13197 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13198 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13199 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13203 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13204 call to ifc_temp_var.
13206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13207 Alan Hayward <alan.hayward@arm.com>
13208 David Sherwood <david.sherwood@arm.com>
13210 * target.def (legitimize_address_displacement): Take the original
13211 offset as a poly_int.
13212 * targhooks.h (default_legitimize_address_displacement): Update
13214 * targhooks.c (default_legitimize_address_displacement): Likewise.
13215 * doc/tm.texi: Regenerate.
13216 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13217 as an argument, moving assert of ad->disp == ad->disp_term to...
13218 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13219 Try calling targetm.legitimize_address_displacement before expanding
13220 the address rather than afterwards, and adjust for the new interface.
13221 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13222 Match the new hook interface. Handle SVE addresses.
13223 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13224 new hook interface.
13226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13228 * Makefile.in (OBJS): Add early-remat.o.
13229 * target.def (select_early_remat_modes): New hook.
13230 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13231 * doc/tm.texi: Regenerate.
13232 * targhooks.h (default_select_early_remat_modes): Declare.
13233 * targhooks.c (default_select_early_remat_modes): New function.
13234 * timevar.def (TV_EARLY_REMAT): New timevar.
13235 * passes.def (pass_early_remat): New pass.
13236 * tree-pass.h (make_pass_early_remat): Declare.
13237 * early-remat.c: New file.
13238 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13240 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13243 Alan Hayward <alan.hayward@arm.com>
13244 David Sherwood <david.sherwood@arm.com>
13246 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13247 vfm1 with a bound_epilog parameter.
13248 (vect_do_peeling): Update calls accordingly, and move the prologue
13249 call earlier in the function. Treat the base bound_epilog as 0 for
13250 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13251 this base when peeling for gaps.
13252 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13253 with fully-masked loops.
13254 (vect_estimate_min_profitable_iters): Handle the single peeled
13255 iteration in that case.
13257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13258 Alan Hayward <alan.hayward@arm.com>
13259 David Sherwood <david.sherwood@arm.com>
13261 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13262 single-element interleaving even if the size is not a power of 2.
13263 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13264 accesses for single-element interleaving if the group size is
13267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13268 Alan Hayward <alan.hayward@arm.com>
13269 David Sherwood <david.sherwood@arm.com>
13271 * doc/md.texi (fold_extract_last_@var{m}): Document.
13272 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13273 * optabs.def (fold_extract_last_optab): New optab.
13274 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13275 * internal-fn.c (fold_extract_direct): New macro.
13276 (expand_fold_extract_optab_fn): Likewise.
13277 (direct_fold_extract_optab_supported_p): Likewise.
13278 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13279 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13280 EXTRACT_LAST_REDUCTION.
13281 (get_initial_def_for_reduction): Do not create an initial vector
13282 for EXTRACT_LAST_REDUCTION reductions.
13283 (vectorizable_reduction): Leave the scalar phi in place for
13284 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13285 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13286 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13287 transform phase to vectorizable_condition.
13288 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13290 (vect_finish_stmt_generation): ...here.
13291 (vect_finish_replace_stmt): New function.
13292 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13293 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13295 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13298 Alan Hayward <alan.hayward@arm.com>
13299 David Sherwood <david.sherwood@arm.com>
13301 * doc/md.texi (extract_last_@var{m}): Document.
13302 * optabs.def (extract_last_optab): New optab.
13303 * internal-fn.def (EXTRACT_LAST): New internal function.
13304 * internal-fn.c (cond_unary_direct): New macro.
13305 (expand_cond_unary_optab_fn): Likewise.
13306 (direct_cond_unary_optab_supported_p): Likewise.
13307 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13308 loops using EXTRACT_LAST.
13309 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13310 (extract_last_<mode>): ...this optab.
13311 (vec_extract<mode><Vel>): Update accordingly.
13313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13314 Alan Hayward <alan.hayward@arm.com>
13315 David Sherwood <david.sherwood@arm.com>
13317 * target.def (empty_mask_is_expensive): New hook.
13318 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13319 * doc/tm.texi: Regenerate.
13320 * targhooks.h (default_empty_mask_is_expensive): Declare.
13321 * targhooks.c (default_empty_mask_is_expensive): New function.
13322 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13323 if the target says that empty masks are expensive.
13324 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13326 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13329 Alan Hayward <alan.hayward@arm.com>
13330 David Sherwood <david.sherwood@arm.com>
13332 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13333 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13334 (vect_use_loop_mask_for_alignment_p): New function.
13335 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13336 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13337 niters_skip argument. Make sure that the first niters_skip elements
13338 of the first iteration are inactive.
13339 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13340 Update call to vect_set_loop_masks_directly.
13341 (get_misalign_in_elems): New function, split out from...
13342 (vect_gen_prolog_loop_niters): ...here.
13343 (vect_update_init_of_dr): Take a code argument that specifies whether
13344 the adjustment should be added or subtracted.
13345 (vect_update_init_of_drs): Likewise.
13346 (vect_prepare_for_masked_peels): New function.
13347 (vect_do_peeling): Skip prologue peeling if we're using a mask
13348 instead. Update call to vect_update_inits_of_drs.
13349 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13351 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13352 alignment. Do not include the number of peeled iterations in
13353 the minimum threshold in that case.
13354 (vectorizable_induction): Adjust the start value down by
13355 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13356 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13357 Take the number of skipped iterations into account when calculating
13359 * tree-vect-stmts.c (vect_gen_while_not): New function.
13361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13362 Alan Hayward <alan.hayward@arm.com>
13363 David Sherwood <david.sherwood@arm.com>
13365 * doc/sourcebuild.texi (vect_fully_masked): Document.
13366 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13367 default value to 0.
13368 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13370 (vect_analyze_loop_2): ...here. Don't check the vectorization
13371 factor against the number of loop iterations if the loop is
13374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13375 Alan Hayward <alan.hayward@arm.com>
13376 David Sherwood <david.sherwood@arm.com>
13378 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13379 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13380 (dump_groups): Update accordingly.
13381 (iv_use::mem_type): New member variable.
13382 (address_p): New function.
13383 (record_use): Add a mem_type argument and initialize the new
13385 (record_group_use): Add a mem_type argument. Use address_p.
13386 Remove obsolete null checks of base_object. Update call to record_use.
13387 (find_interesting_uses_op): Update call to record_group_use.
13388 (find_interesting_uses_cond): Likewise.
13389 (find_interesting_uses_address): Likewise.
13390 (get_mem_type_for_internal_fn): New function.
13391 (find_address_like_use): Likewise.
13392 (find_interesting_uses_stmt): Try find_address_like_use before
13393 calling find_interesting_uses_op.
13394 (addr_offset_valid_p): Use the iv mem_type field as the type
13395 of the addressed memory.
13396 (add_autoinc_candidates): Likewise.
13397 (get_address_cost): Likewise.
13398 (split_small_address_groups_p): Use address_p.
13399 (split_address_groups): Likewise.
13400 (add_iv_candidate_for_use): Likewise.
13401 (autoinc_possible_for_pair): Likewise.
13402 (rewrite_groups): Likewise.
13403 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13404 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13405 (get_alias_ptr_type_for_ptr_address): New function.
13406 (rewrite_use_address): Rewrite address uses in calls that were
13407 identified by find_address_like_use.
13409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13410 Alan Hayward <alan.hayward@arm.com>
13411 David Sherwood <david.sherwood@arm.com>
13413 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13415 * gimple-expr.h (is_gimple_addressable: Likewise.
13416 * gimple-expr.c (is_gimple_address): Likewise.
13417 * internal-fn.c (expand_call_mem_ref): New function.
13418 (expand_mask_load_optab_fn): Use it.
13419 (expand_mask_store_optab_fn): Likewise.
13421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13422 Alan Hayward <alan.hayward@arm.com>
13423 David Sherwood <david.sherwood@arm.com>
13425 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13426 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13427 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13428 (cond_umax@var{mode}): Document.
13429 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13430 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13431 (cond_umin_optab, cond_umax_optab): New optabs.
13432 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13433 (COND_IOR, COND_XOR): New internal functions.
13434 * internal-fn.h (get_conditional_internal_fn): Declare.
13435 * internal-fn.c (cond_binary_direct): New macro.
13436 (expand_cond_binary_optab_fn): Likewise.
13437 (direct_cond_binary_optab_supported_p): Likewise.
13438 (get_conditional_internal_fn): New function.
13439 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13440 Cope with reduction statements that are vectorized as calls rather
13442 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13443 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13444 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13445 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13446 (UNSPEC_COND_EOR): New unspecs.
13447 (optab): Add mappings for them.
13448 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13449 (sve_int_op, sve_fp_op): New int attributes.
13451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13452 Alan Hayward <alan.hayward@arm.com>
13453 David Sherwood <david.sherwood@arm.com>
13455 * optabs.def (while_ult_optab): New optab.
13456 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13457 * internal-fn.def (WHILE_ULT): New internal function.
13458 * internal-fn.h (direct_internal_fn_supported_p): New override
13459 that takes two types as argument.
13460 * internal-fn.c (while_direct): New macro.
13461 (expand_while_optab_fn): New function.
13462 (convert_optab_supported_p): Likewise.
13463 (direct_while_optab_supported_p): New macro.
13464 * wide-int.h (wi::udiv_ceil): New function.
13465 * tree-vectorizer.h (rgroup_masks): New structure.
13466 (vec_loop_masks): New typedef.
13467 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13468 and fully_masked_p.
13469 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13470 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13471 (vect_max_vf): New function.
13472 (slpeel_make_loop_iterate_ntimes): Delete.
13473 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13474 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13475 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13476 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13477 internal-fn.h, stor-layout.h and optabs-query.h.
13478 (vect_set_loop_mask): New function.
13479 (add_preheader_seq): Likewise.
13480 (add_header_seq): Likewise.
13481 (interleave_supported_p): Likewise.
13482 (vect_maybe_permute_loop_masks): Likewise.
13483 (vect_set_loop_masks_directly): Likewise.
13484 (vect_set_loop_condition_masked): Likewise.
13485 (vect_set_loop_condition_unmasked): New function, split out from
13486 slpeel_make_loop_iterate_ntimes.
13487 (slpeel_make_loop_iterate_ntimes): Rename to..
13488 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13489 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13490 (vect_do_peeling): Update call accordingly.
13491 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13493 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13494 mask_compare_type, can_fully_mask_p and fully_masked_p.
13495 (release_vec_loop_masks): New function.
13496 (_loop_vec_info): Use it to free the loop masks.
13497 (can_produce_all_loop_masks_p): New function.
13498 (vect_get_max_nscalars_per_iter): Likewise.
13499 (vect_verify_full_masking): Likewise.
13500 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13501 retries, and free the mask rgroups before retrying. Check loop-wide
13502 reasons for disallowing fully-masked loops. Make the final decision
13503 about whether use a fully-masked loop or not.
13504 (vect_estimate_min_profitable_iters): Do not assume that peeling
13505 for the number of iterations will be needed for fully-masked loops.
13506 (vectorizable_reduction): Disable fully-masked loops.
13507 (vectorizable_live_operation): Likewise.
13508 (vect_halve_mask_nunits): New function.
13509 (vect_double_mask_nunits): Likewise.
13510 (vect_record_loop_mask): Likewise.
13511 (vect_get_loop_mask): Likewise.
13512 (vect_transform_loop): Handle the case in which the final loop
13513 iteration might handle a partial vector. Call vect_set_loop_condition
13514 instead of slpeel_make_loop_iterate_ntimes.
13515 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13516 (check_load_store_masking): New function.
13517 (prepare_load_store_mask): Likewise.
13518 (vectorizable_store): Handle fully-masked loops.
13519 (vectorizable_load): Likewise.
13520 (supportable_widening_operation): Use vect_halve_mask_nunits for
13522 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13523 (vect_gen_while): New function.
13524 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13525 (aarch64_uqdec<mode>): New insn.
13527 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13528 Alan Hayward <alan.hayward@arm.com>
13529 David Sherwood <david.sherwood@arm.com>
13531 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13532 (reduc_xor_scal_optab): New optabs.
13533 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13534 (reduc_xor_scal_@var{m}): Document.
13535 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13536 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13537 internal functions.
13538 * fold-const-call.c (fold_const_call): Handle them.
13539 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13540 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13541 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13542 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13543 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13544 (UNSPEC_XORV): New unspecs.
13545 (optab): Add entries for them.
13546 (BITWISEV): New int iterator.
13547 (bit_reduc_op): New int attributes.
13549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13550 Alan Hayward <alan.hayward@arm.com>
13551 David Sherwood <david.sherwood@arm.com>
13553 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13554 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13555 * optabs.def (vec_shl_insert_optab): New optab.
13556 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13557 (duplicate_and_interleave): Likewise.
13558 * tree-vect-loop.c: Include internal-fn.h.
13559 (neutral_op_for_slp_reduction): New function, split out from
13560 get_initial_defs_for_reduction.
13561 (get_initial_def_for_reduction): Handle option 2 for variable-length
13562 vectors by loading the neutral value into a vector and then shifting
13563 the initial value into element 0.
13564 (get_initial_defs_for_reduction): Replace the code argument with
13565 the neutral value calculated by neutral_op_for_slp_reduction.
13566 Use gimple_build_vector for constant-length vectors.
13567 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13568 but the first group_size elements have a neutral value.
13569 Use duplicate_and_interleave otherwise.
13570 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13571 Update call to get_initial_defs_for_reduction. Handle SLP
13572 reductions for variable-length vectors by creating one vector
13573 result for each scalar result, with the elements associated
13574 with other scalar results stubbed out with the neutral value.
13575 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13576 Require IFN_VEC_SHL_INSERT for double reductions on
13577 variable-length vectors, or SLP reductions that have
13578 a neutral value. Require can_duplicate_and_interleave_p
13579 support for variable-length unchained SLP reductions if there
13580 is no neutral value, such as for MIN/MAX reductions. Also require
13581 the number of vector elements to be a multiple of the number of
13582 SLP statements when doing variable-length unchained SLP reductions.
13583 Update call to vect_create_epilog_for_reduction.
13584 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13585 and remove initial values.
13586 (duplicate_and_interleave): Make public.
13587 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13588 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13591 Alan Hayward <alan.hayward@arm.com>
13592 David Sherwood <david.sherwood@arm.com>
13594 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13595 (can_duplicate_and_interleave_p): New function.
13596 (vect_get_and_check_slp_defs): Take the vector of statements
13597 rather than just the current one. Remove excess parentheses.
13598 Restriction rejectinon of vect_constant_def and vect_external_def
13599 for variable-length vectors to boolean types, or types for which
13600 can_duplicate_and_interleave_p is false.
13601 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13602 (duplicate_and_interleave): New function.
13603 (vect_get_constant_vectors): Use gimple_build_vector for
13604 constant-length vectors and suitable variable-length constant
13605 vectors. Use duplicate_and_interleave for other variable-length
13606 vectors. Don't defer the update when inserting new statements.
13608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13609 Alan Hayward <alan.hayward@arm.com>
13610 David Sherwood <david.sherwood@arm.com>
13612 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13613 min_profitable_iters doesn't go negative.
13615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13616 Alan Hayward <alan.hayward@arm.com>
13617 David Sherwood <david.sherwood@arm.com>
13619 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13620 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13621 * optabs.def (vec_mask_load_lanes_optab): New optab.
13622 (vec_mask_store_lanes_optab): Likewise.
13623 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13624 (MASK_STORE_LANES): Likewise.
13625 * internal-fn.c (mask_load_lanes_direct): New macro.
13626 (mask_store_lanes_direct): Likewise.
13627 (expand_mask_load_optab_fn): Handle masked operations.
13628 (expand_mask_load_lanes_optab_fn): New macro.
13629 (expand_mask_store_optab_fn): Handle masked operations.
13630 (expand_mask_store_lanes_optab_fn): New macro.
13631 (direct_mask_load_lanes_optab_supported_p): Likewise.
13632 (direct_mask_store_lanes_optab_supported_p): Likewise.
13633 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13635 (vect_load_lanes_supported): Likewise.
13636 * tree-vect-data-refs.c (strip_conversion): New function.
13637 (can_group_stmts_p): Likewise.
13638 (vect_analyze_data_ref_accesses): Use it instead of checking
13639 for a pair of assignments.
13640 (vect_store_lanes_supported): Take a masked_p parameter.
13641 (vect_load_lanes_supported): Likewise.
13642 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13643 vect_store_lanes_supported and vect_load_lanes_supported.
13644 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13645 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13646 parameter. Don't allow gaps for masked accesses.
13647 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13648 and vect_load_lanes_supported.
13649 (get_load_store_type): Take a masked_p parameter and update
13650 call to get_group_load_store_type.
13651 (vectorizable_store): Update call to get_load_store_type.
13652 Handle IFN_MASK_STORE_LANES.
13653 (vectorizable_load): Update call to get_load_store_type.
13654 Handle IFN_MASK_LOAD_LANES.
13656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13657 Alan Hayward <alan.hayward@arm.com>
13658 David Sherwood <david.sherwood@arm.com>
13660 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13662 * config/aarch64/aarch64-protos.h
13663 (aarch64_sve_struct_memory_operand_p): Declare.
13664 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13665 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13666 (VPRED, vpred): Handle SVE structure modes.
13667 * config/aarch64/constraints.md (Utx): New constraint.
13668 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13669 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13670 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13671 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13672 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13673 structure modes. Split into pieces after RA.
13674 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13675 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13677 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13678 SVE structure modes.
13679 (aarch64_classify_address): Likewise.
13680 (sizetochar): Move earlier in file.
13681 (aarch64_print_operand): Handle SVE register lists.
13682 (aarch64_array_mode): New function.
13683 (aarch64_sve_struct_memory_operand_p): Likewise.
13684 (TARGET_ARRAY_MODE): Redefine.
13686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13687 Alan Hayward <alan.hayward@arm.com>
13688 David Sherwood <david.sherwood@arm.com>
13690 * target.def (array_mode): New target hook.
13691 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13692 * doc/tm.texi: Regenerate.
13693 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13694 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13695 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13696 targetm.array_mode.
13697 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13701 Alan Hayward <alan.hayward@arm.com>
13702 David Sherwood <david.sherwood@arm.com>
13704 * fold-const.c (fold_binary_loc): Check the argument types
13705 rather than the result type when testing for a vector operation.
13707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13709 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13710 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13717 (sve): Document new AArch64 extension.
13718 * doc/md.texi (w): Extend the description of the AArch64
13719 constraint to include SVE vectors.
13720 (Upl, Upa): Document new AArch64 predicate constraints.
13721 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13723 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13724 (msve-vector-bits=): New option.
13725 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13726 SVE when these are disabled.
13727 (sve): New extension.
13728 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13729 modes. Adjust their number of units based on aarch64_sve_vg.
13730 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13731 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13732 aarch64_addr_query_type.
13733 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13734 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13735 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13736 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13737 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13738 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13739 (aarch64_simd_imm_zero_p): Delete.
13740 (aarch64_check_zero_based_sve_index_immediate): Declare.
13741 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13742 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13743 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13744 (aarch64_sve_float_mul_immediate_p): Likewise.
13745 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13746 rather than an rtx.
13747 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13748 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13749 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13750 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13751 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13752 (aarch64_regmode_natural_size): Likewise.
13753 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13754 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13756 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13757 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13758 for VG and the SVE predicate registers.
13759 (V_ALIASES): Add a "z"-prefixed alias.
13760 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13761 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13762 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13763 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13764 (REG_CLASS_NAMES): Add entries for them.
13765 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13766 and the predicate registers.
13767 (aarch64_sve_vg): Declare.
13768 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13769 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13770 (REGMODE_NATURAL_SIZE): Define.
13771 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13773 * config/aarch64/aarch64.c: Include cfgrtl.h.
13774 (simd_immediate_info): Add a constructor for series vectors,
13775 and an associated step field.
13776 (aarch64_sve_vg): New variable.
13777 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13778 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13779 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13780 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13781 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13782 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13783 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13784 (aarch64_get_mask_mode): New functions.
13785 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13786 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13787 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13788 predicate modes and predicate registers. Explicitly restrict
13789 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13790 to store a vector mode if it is recognized by
13791 aarch64_classify_vector_mode.
13792 (aarch64_regmode_natural_size): New function.
13793 (aarch64_hard_regno_caller_save_mode): Return the original mode
13795 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13796 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13797 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13798 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13800 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13801 does not overlap dest if the function is frame-related. Handle
13803 (aarch64_split_add_offset): New function.
13804 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13805 them aarch64_add_offset.
13806 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13807 and update call to aarch64_sub_sp.
13808 (aarch64_add_cfa_expression): New function.
13809 (aarch64_expand_prologue): Pass extra temporary registers to the
13810 functions above. Handle the case in which we need to emit new
13811 DW_CFA_expressions for registers that were originally saved
13812 relative to the stack pointer, but now have to be expressed
13813 relative to the frame pointer.
13814 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13816 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13817 IP0 and IP1 values for SVE frames.
13818 (aarch64_expand_vec_series): New function.
13819 (aarch64_expand_sve_widened_duplicate): Likewise.
13820 (aarch64_expand_sve_const_vector): Likewise.
13821 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13822 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13823 into the register, rather than emitting a SET directly.
13824 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13825 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13826 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13827 (offset_9bit_signed_scaled_p): New functions.
13828 (aarch64_replicate_bitmask_imm): New function.
13829 (aarch64_bitmask_imm): Use it.
13830 (aarch64_cannot_force_const_mem): Reject expressions involving
13831 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13832 (aarch64_classify_index): Handle SVE indices, by requiring
13833 a plain register index with a scale that matches the element size.
13834 (aarch64_classify_address): Handle SVE addresses. Assert that
13835 the mode of the address is VOIDmode or an integer mode.
13836 Update call to aarch64_classify_symbol.
13837 (aarch64_classify_symbolic_expression): Update call to
13838 aarch64_classify_symbol.
13839 (aarch64_const_vec_all_in_range_p): New function.
13840 (aarch64_print_vector_float_operand): Likewise.
13841 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13842 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13843 and the FP immediates 1.0 and 0.5.
13844 (aarch64_print_address_internal): Handle SVE addresses.
13845 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13846 (aarch64_regno_regclass): Handle predicate registers.
13847 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13849 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13850 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13851 (aarch64_convert_sve_vector_bits): New function.
13852 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13853 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13854 rather than an rtx.
13855 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13856 Handle SVE vector and predicate modes. Accept VL-based constants
13857 that need only one temporary register, and VL offsets that require
13858 no temporary registers.
13859 (aarch64_conditional_register_usage): Mark the predicate registers
13860 as fixed if SVE isn't available.
13861 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13862 Return true for SVE vector and predicate modes.
13863 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13864 rather than an unsigned int. Handle SVE modes.
13865 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13867 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13869 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13870 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13871 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13872 (aarch64_sve_float_mul_immediate_p): New functions.
13873 (aarch64_sve_valid_immediate): New function.
13874 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13875 Explicitly reject structure modes. Check for INDEX constants.
13876 Handle PTRUE and PFALSE constants.
13877 (aarch64_check_zero_based_sve_index_immediate): New function.
13878 (aarch64_simd_imm_zero_p): Delete.
13879 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13880 vector modes. Accept constants in the range of CNT[BHWD].
13881 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13882 ask for an Advanced SIMD mode.
13883 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13884 (aarch64_simd_vector_alignment): Handle SVE predicates.
13885 (aarch64_vectorize_preferred_vector_alignment): New function.
13886 (aarch64_simd_vector_alignment_reachable): Use it instead of
13888 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13889 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13891 (MAX_VECT_LEN): Delete.
13892 (expand_vec_perm_d): Add a vec_flags field.
13893 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13894 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13895 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13897 (aarch64_evpc_rev): Rename to...
13898 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13899 (aarch64_evpc_rev_global): New function.
13900 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13901 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13903 (aarch64_evpc_sve_tbl): New function.
13904 (aarch64_expand_vec_perm_const_1): Update after rename of
13905 aarch64_evpc_rev. Handle SVE permutes too, trying
13906 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13907 than aarch64_evpc_tbl.
13908 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13909 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13910 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13911 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13912 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13913 (aarch64_expand_sve_vcond): New functions.
13914 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13915 of aarch64_vector_mode_p.
13916 (aarch64_dwarf_poly_indeterminate_value): New function.
13917 (aarch64_compute_pressure_classes): Likewise.
13918 (aarch64_can_change_mode_class): Likewise.
13919 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13920 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13921 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13922 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13923 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13924 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13925 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13926 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13928 (Dn, Dl, Dr): Accept const as well as const_vector.
13929 (Dz): Likewise. Compare against CONST0_RTX.
13930 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13931 of "vector" where appropriate.
13932 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13933 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13934 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13935 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13936 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13937 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13938 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13939 (v_int_equiv): Extend to SVE modes.
13940 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13942 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13943 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13944 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13945 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13946 (SVE_COND_FP_CMP): New int iterators.
13947 (perm_hilo): Handle the new unpack unspecs.
13948 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13950 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13951 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13952 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13953 (aarch64_equality_operator, aarch64_constant_vector_operand)
13954 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13955 (aarch64_sve_nonimmediate_operand): Likewise.
13956 (aarch64_sve_general_operand): Likewise.
13957 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13958 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13959 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13960 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13961 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13962 (aarch64_sve_float_arith_immediate): Likewise.
13963 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13964 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13965 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13966 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13967 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13968 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13969 (aarch64_sve_float_arith_operand): Likewise.
13970 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13971 (aarch64_sve_float_mul_operand): Likewise.
13972 (aarch64_sve_vec_perm_operand): Likewise.
13973 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13974 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13975 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13976 as well as const_vector.
13977 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13978 in file. Use CONST0_RTX and CONSTM1_RTX.
13979 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13980 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13981 Use aarch64_simd_imm_zero.
13982 * config/aarch64/aarch64-sve.md: New file.
13983 * config/aarch64/aarch64.md: Include it.
13984 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13985 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13986 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13987 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13988 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13989 (sve): New attribute.
13990 (enabled): Disable instructions with the sve attribute unless
13992 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13993 aarch64_expand_mov_immediate.
13994 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13995 CNT[BHSD] immediates.
13996 (movti): Split CONST_POLY_INT moves into two halves.
13997 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13998 Split additions that need a temporary here if the destination
13999 is the stack pointer.
14000 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14001 (*add<mode>3_poly_1): New instruction.
14002 (set_clobber_cc): New expander.
14004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14006 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14007 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14008 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14009 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14010 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14011 Change innermode from fixed_mode_size to machine_mode.
14012 (simplify_subreg): Update call accordingly. Handle a constant-sized
14013 subreg of a variable-length CONST_VECTOR.
14015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14016 Alan Hayward <alan.hayward@arm.com>
14017 David Sherwood <david.sherwood@arm.com>
14019 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14020 (add_offset_to_base): New function, split out from...
14021 (create_mem_ref): ...here. When handling a scale other than 1,
14022 check first whether the address is valid without the offset.
14023 Add it into the base if so, leaving the index and scale as-is.
14025 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14028 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14029 fold_for_warn before checking if arg2 is INTEGER_CST.
14031 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14033 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14034 (store_multiple_operation): Delete.
14035 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14036 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14037 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14038 guarded by TARGET_STRING.
14039 (rs6000_output_load_multiple): Delete.
14040 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14041 OPTION_MASK_STRING / TARGET_STRING handling.
14042 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14043 (const rs6000_opt_masks) <"string">: Change mask to 0.
14044 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14045 (MASK_STRING): Delete.
14046 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14048 (load_multiple): Delete.
14055 (store_multiple): Delete.
14062 (movmemsi_8reg): Delete.
14063 (corresponding unnamed define_insn): Delete.
14064 (movmemsi_6reg): Delete.
14065 (corresponding unnamed define_insn): Delete.
14066 (movmemsi_4reg): Delete.
14067 (corresponding unnamed define_insn): Delete.
14068 (movmemsi_2reg): Delete.
14069 (corresponding unnamed define_insn): Delete.
14070 (movmemsi_1reg): Delete.
14071 (corresponding unnamed define_insn): Delete.
14072 * config/rs6000/rs6000.opt (mno-string): New.
14073 (mstring): Replace by deprecation warning stub.
14074 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14076 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14078 * regrename.c (regrename_do_replace): If replacing the same
14079 reg multiple times, try to reuse last created gen_raw_REG.
14082 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14083 main to workaround a bug in GDB.
14085 2018-01-12 Tom de Vries <tom@codesourcery.com>
14088 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14090 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14092 PR rtl-optimization/80481
14093 * ira-color.c (get_cap_member): New function.
14094 (allocnos_conflict_by_live_ranges_p): Use it.
14095 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14096 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14098 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14101 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14102 (*saddl_se_1): Ditto.
14103 (*ssubsi_1): Ditto.
14104 (*ssubl_se_1): Ditto.
14106 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14108 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14109 rather than wi::to_widest for DR_INITs.
14110 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14111 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14112 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14114 (vect_analyze_group_access_1): Note that here.
14116 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14118 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14119 polynomial type sizes.
14121 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14123 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14124 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14125 (gimple_add_tmp_var): Likewise.
14127 2018-01-12 Martin Liska <mliska@suse.cz>
14129 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14130 (gimple_alloc_sizes): Likewise.
14131 (dump_gimple_statistics): Use PRIu64 in printf format.
14132 * gimple.h: Change uint64_t to int.
14134 2018-01-12 Martin Liska <mliska@suse.cz>
14136 * tree-core.h: Use uint64_t instead of int.
14137 * tree.c (tree_node_counts): Likewise.
14138 (tree_node_sizes): Likewise.
14139 (dump_tree_statistics): Use PRIu64 in printf format.
14141 2018-01-12 Martin Liska <mliska@suse.cz>
14143 * Makefile.in: As qsort_chk is implemented in vec.c, add
14144 vec.o to linkage of gencfn-macros.
14145 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14146 passing the info to record_node_allocation_statistics.
14147 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14149 * ggc-common.c (struct ggc_usage): Add operator== and use
14150 it in operator< and compare function.
14151 * mem-stats.h (struct mem_usage): Likewise.
14152 * vec.c (struct vec_usage): Remove operator< and compare
14153 function. Can be simply inherited.
14155 2018-01-12 Martin Jambor <mjambor@suse.cz>
14158 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14159 * tree-ssa-math-opts.c: Include domwalk.h.
14160 (convert_mult_to_fma_1): New function.
14161 (fma_transformation_info): New type.
14162 (fma_deferring_state): Likewise.
14163 (cancel_fma_deferring): New function.
14164 (result_of_phi): Likewise.
14165 (last_fma_candidate_feeds_initial_phi): Likewise.
14166 (convert_mult_to_fma): Added deferring logic, split actual
14167 transformation to convert_mult_to_fma_1.
14168 (math_opts_dom_walker): New type.
14169 (math_opts_dom_walker::after_dom_children): New method, body moved
14170 here from pass_optimize_widening_mul::execute, added deferring logic
14172 (pass_optimize_widening_mul::execute): Moved most of code to
14173 math_opts_dom_walker::after_dom_children.
14174 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14175 * config/i386/i386.c (ix86_option_override_internal): Added
14176 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14178 2018-01-12 Richard Biener <rguenther@suse.de>
14181 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14182 inline instance vars.
14184 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14187 * config/rx/rx.c (rx_is_restricted_memory_address):
14188 Handle SUBREG case.
14190 2018-01-12 Richard Biener <rguenther@suse.de>
14192 PR tree-optimization/80846
14193 * target.def (split_reduction): New target hook.
14194 * targhooks.c (default_split_reduction): New function.
14195 * targhooks.h (default_split_reduction): Declare.
14196 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14197 target requests first reduce vectors by combining low and high
14199 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14200 (get_vectype_for_scalar_type_and_size): Export.
14201 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14202 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14203 * doc/tm.texi: Regenerate.
14204 * config/i386/i386.c (ix86_split_reduction): Implement
14205 TARGET_VECTORIZE_SPLIT_REDUCTION.
14207 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14210 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14211 in PIC mode except for TARGET_VXWORKS_RTP.
14212 * config/sparc/sparc.c: Include cfgrtl.h.
14213 (TARGET_INIT_PIC_REG): Define.
14214 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14215 (sparc_pic_register_p): New predicate.
14216 (sparc_legitimate_address_p): Use it.
14217 (sparc_legitimize_pic_address): Likewise.
14218 (sparc_delegitimize_address): Likewise.
14219 (sparc_mode_dependent_address_p): Likewise.
14220 (gen_load_pcrel_sym): Remove 4th parameter.
14221 (load_got_register): Adjust call to above. Remove obsolete stuff.
14222 (sparc_expand_prologue): Do not call load_got_register here.
14223 (sparc_flat_expand_prologue): Likewise.
14224 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14225 (sparc_use_pseudo_pic_reg): New function.
14226 (sparc_init_pic_reg): Likewise.
14227 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14228 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14230 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14232 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14233 Add item for branch_cost.
14235 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14237 PR rtl-optimization/83565
14238 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14239 not extend the result to a larger mode for rotate operations.
14240 (num_sign_bit_copies1): Likewise.
14242 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14245 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14247 Use values-Xc.o for -pedantic.
14248 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14250 2018-01-12 Martin Liska <mliska@suse.cz>
14253 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14255 (possible_polymorphic_call_targets): Use it.
14256 (ipa_devirt): Likewise.
14258 2018-01-12 Martin Liska <mliska@suse.cz>
14260 * profile-count.h (enum profile_quality): Use 0 as invalid
14261 enum value of profile_quality.
14263 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14265 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14266 -mext-string options.
14268 2018-01-12 Richard Biener <rguenther@suse.de>
14270 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14271 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14272 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14274 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14276 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14278 * configure.ac (--with-long-double-format): Add support for the
14279 configuration option to change the default long double format on
14281 * config.gcc (powerpc*-linux*-*): Likewise.
14282 * configure: Regenerate.
14283 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14284 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14285 used without modification.
14287 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14289 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14290 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14291 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14292 MISC_BUILTIN_SPEC_BARRIER.
14293 (rs6000_init_builtins): Likewise.
14294 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14296 (speculation_barrier): New define_insn.
14297 * doc/extend.texi: Document __builtin_speculation_barrier.
14299 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14302 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14303 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14304 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14306 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14307 integral modes instead of "ss" and "sd".
14308 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14309 vectors with 32-bit and 64-bit elements.
14310 (vecdupssescalarmodesuffix): New mode attribute.
14311 (vec_dup<mode>): Use it.
14313 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14316 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14317 frame if argument is passed on stack.
14319 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14322 * ree.c (combine_reaching_defs): Optimize also
14323 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14324 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14326 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14328 PR middle-end/83189
14329 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14331 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14333 PR middle-end/83718
14334 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14335 after they are computed.
14337 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14339 PR tree-optimization/83695
14340 * gimple-loop-linterchange.cc
14341 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14342 reset cached scev information after interchange.
14343 (pass_linterchange::execute): Remove call to scev_reset_htab.
14345 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14347 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14348 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14349 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14350 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14351 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14352 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14353 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14354 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14355 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14356 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14357 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14358 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14359 (V_lane_reg): Likewise.
14360 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14362 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14363 (vfmal_lane_low<mode>_intrinsic,
14364 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14365 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14366 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14367 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14368 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14369 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14371 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14373 * config/arm/arm-cpus.in (fp16fml): New feature.
14374 (ALL_SIMD): Add fp16fml.
14375 (armv8.2-a): Add fp16fml as an option.
14376 (armv8.3-a): Likewise.
14377 (armv8.4-a): Add fp16fml as part of fp16.
14378 * config/arm/arm.h (TARGET_FP16FML): Define.
14379 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14381 * config/arm/arm-modes.def (V2HF): Define.
14382 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14383 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14384 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14385 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14386 vfmsl_low, vfmsl_high): New set of builtins.
14387 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14388 (vfml_op): New code attribute.
14389 (VFMLHALVES): New int iterator.
14390 (VFML, VFMLSEL): New mode attributes.
14391 (V_reg): Define mapping for V2HF.
14392 (V_hi, V_lo): New mode attributes.
14393 (VF_constraint): Likewise.
14394 (vfml_half, vfml_half_selector): New int attributes.
14395 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14397 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14398 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14400 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14401 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14402 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14403 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14405 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14406 Document new effective target and option set.
14408 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14410 * config/arm/arm-cpus.in (armv8_4): New feature.
14411 (ARMv8_4a): New fgroup.
14412 (armv8.4-a): New arch.
14413 * config/arm/arm-tables.opt: Regenerate.
14414 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14415 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14416 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14417 Add matching rules for -march=armv8.4-a and extensions.
14418 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14420 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14423 * config/rx/rx.md (BW): New mode attribute.
14424 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14426 2018-01-11 Richard Biener <rguenther@suse.de>
14428 PR tree-optimization/83435
14429 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14430 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14431 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14433 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14434 Alan Hayward <alan.hayward@arm.com>
14435 David Sherwood <david.sherwood@arm.com>
14437 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14439 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14440 (aarch64_print_address_internal): Use it to check for a zero offset.
14442 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14443 Alan Hayward <alan.hayward@arm.com>
14444 David Sherwood <david.sherwood@arm.com>
14446 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14447 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14448 Return a poly_int64 rather than a HOST_WIDE_INT.
14449 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14450 rather than a HOST_WIDE_INT.
14451 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14452 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14453 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14454 final_offset from HOST_WIDE_INT to poly_int64.
14455 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14456 to_constant when getting the number of units in an Advanced SIMD
14458 (aarch64_builtin_vectorized_function): Check for a constant number
14460 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14462 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14463 attribute instead of GET_MODE_NUNITS.
14464 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14465 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14466 GET_MODE_SIZE for fixed-size registers.
14467 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14468 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14469 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14470 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14471 (aarch64_print_operand, aarch64_print_address_internal)
14472 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14473 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14474 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14475 Handle polynomial GET_MODE_SIZE.
14476 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14477 wider than SImode without modification.
14478 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14479 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14480 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14481 passing and returning SVE modes.
14482 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14483 rather than GEN_INT.
14484 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14485 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14486 (aarch64_allocate_and_probe_stack_space): Likewise.
14487 (aarch64_layout_frame): Cope with polynomial offsets.
14488 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14489 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14490 polynomial offsets.
14491 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14492 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14493 poly_int64 rather than a HOST_WIDE_INT.
14494 (aarch64_get_separate_components, aarch64_process_components)
14495 (aarch64_expand_prologue, aarch64_expand_epilogue)
14496 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14497 (aarch64_anchor_offset): New function, split out from...
14498 (aarch64_legitimize_address): ...here.
14499 (aarch64_builtin_vectorization_cost): Handle polynomial
14500 TYPE_VECTOR_SUBPARTS.
14501 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14503 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14504 number of elements from the PARALLEL rather than the mode.
14505 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14506 rather than GET_MODE_BITSIZE.
14507 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14508 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14509 (aarch64_expand_vec_perm_const_1): Handle polynomial
14510 d->perm.length () and d->perm elements.
14511 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14512 Apply to_constant to d->perm elements.
14513 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14514 polynomial CONST_VECTOR_NUNITS.
14515 (aarch64_move_pointer): Take amount as a poly_int64 rather
14517 (aarch64_progress_pointer): Avoid temporary variable.
14518 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14519 the mode attribute instead of GET_MODE.
14521 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14522 Alan Hayward <alan.hayward@arm.com>
14523 David Sherwood <david.sherwood@arm.com>
14525 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14526 x exists before using it.
14527 (aarch64_add_constant_internal): Rename to...
14528 (aarch64_add_offset_1): ...this. Replace regnum with separate
14529 src and dest rtxes. Handle the case in which they're different,
14530 including when the offset is zero. Replace scratchreg with an rtx.
14531 Use 2 additions if there is no spare register into which we can
14532 move a 16-bit constant.
14533 (aarch64_add_constant): Delete.
14534 (aarch64_add_offset): Replace reg with separate src and dest
14535 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14536 Use aarch64_add_offset_1.
14537 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14538 an rtx rather than an int. Take the delta as a poly_int64
14539 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14540 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14541 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14542 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14543 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14544 and aarch64_add_sp.
14545 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14546 aarch64_add_constant.
14548 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14550 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14551 Use scalar_float_mode.
14553 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14555 * config/aarch64/aarch64-simd.md
14556 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14557 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14558 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14559 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14560 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14561 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14562 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14563 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14564 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14565 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14567 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14570 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14571 targ_options->x_arm_arch_string is non NULL.
14573 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14575 * config/aarch64/aarch64.h
14576 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14578 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14581 * expmed.c (emit_store_flag_force): Swap if const op0
14582 and change VOIDmode to mode of op0.
14584 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14586 PR rtl-optimization/83761
14587 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14588 than bytes to mode_for_size.
14590 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14592 PR middle-end/83189
14593 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14594 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14597 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14599 PR middle-end/83575
14600 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14601 when in layout mode.
14602 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14603 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14606 2018-01-10 Michael Collison <michael.collison@arm.com>
14608 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14609 * config/aarch64/aarch64-option-extension.def: Add
14610 AARCH64_OPT_EXTENSION of 'fp16fml'.
14611 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14612 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14613 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14614 * config/aarch64/constraints.md (Ui7): New constraint.
14615 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14616 (VFMLA_SEL_W): Ditto.
14619 (VFMLA16_LOW): New int iterator.
14620 (VFMLA16_HIGH): Ditto.
14621 (UNSPEC_FMLAL): New unspec.
14622 (UNSPEC_FMLSL): Ditto.
14623 (UNSPEC_FMLAL2): Ditto.
14624 (UNSPEC_FMLSL2): Ditto.
14625 (f16mac): New code attribute.
14626 * config/aarch64/aarch64-simd-builtins.def
14627 (aarch64_fmlal_lowv2sf): Ditto.
14628 (aarch64_fmlsl_lowv2sf): Ditto.
14629 (aarch64_fmlalq_lowv4sf): Ditto.
14630 (aarch64_fmlslq_lowv4sf): Ditto.
14631 (aarch64_fmlal_highv2sf): Ditto.
14632 (aarch64_fmlsl_highv2sf): Ditto.
14633 (aarch64_fmlalq_highv4sf): Ditto.
14634 (aarch64_fmlslq_highv4sf): Ditto.
14635 (aarch64_fmlal_lane_lowv2sf): Ditto.
14636 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14637 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14638 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14639 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14640 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14641 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14642 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14643 (aarch64_fmlal_lane_highv2sf): Ditto.
14644 (aarch64_fmlsl_lane_highv2sf): Ditto.
14645 (aarch64_fmlal_laneq_highv2sf): Ditto.
14646 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14647 (aarch64_fmlalq_lane_highv4sf): Ditto.
14648 (aarch64_fmlsl_lane_highv4sf): Ditto.
14649 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14650 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14651 * config/aarch64/aarch64-simd.md:
14652 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14653 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14654 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14655 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14656 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14657 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14658 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14659 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14660 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14661 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14662 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14663 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14664 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14665 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14666 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14667 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14668 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14669 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14670 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14671 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14672 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14673 (vfmlsl_low_u32): Ditto.
14674 (vfmlalq_low_u32): Ditto.
14675 (vfmlslq_low_u32): Ditto.
14676 (vfmlal_high_u32): Ditto.
14677 (vfmlsl_high_u32): Ditto.
14678 (vfmlalq_high_u32): Ditto.
14679 (vfmlslq_high_u32): Ditto.
14680 (vfmlal_lane_low_u32): Ditto.
14681 (vfmlsl_lane_low_u32): Ditto.
14682 (vfmlal_laneq_low_u32): Ditto.
14683 (vfmlsl_laneq_low_u32): Ditto.
14684 (vfmlalq_lane_low_u32): Ditto.
14685 (vfmlslq_lane_low_u32): Ditto.
14686 (vfmlalq_laneq_low_u32): Ditto.
14687 (vfmlslq_laneq_low_u32): Ditto.
14688 (vfmlal_lane_high_u32): Ditto.
14689 (vfmlsl_lane_high_u32): Ditto.
14690 (vfmlal_laneq_high_u32): Ditto.
14691 (vfmlsl_laneq_high_u32): Ditto.
14692 (vfmlalq_lane_high_u32): Ditto.
14693 (vfmlslq_lane_high_u32): Ditto.
14694 (vfmlalq_laneq_high_u32): Ditto.
14695 (vfmlslq_laneq_high_u32): Ditto.
14696 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14697 (AARCH64_FL_FOR_ARCH8_4): New.
14698 (AARCH64_ISA_F16FML): New ISA flag.
14699 (TARGET_F16FML): New feature flag for fp16fml.
14700 (doc/invoke.texi): Document new fp16fml option.
14702 2018-01-10 Michael Collison <michael.collison@arm.com>
14704 * config/aarch64/aarch64-builtins.c:
14705 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14706 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14707 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14708 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14709 (AARCH64_ISA_SHA3): New ISA flag.
14710 (TARGET_SHA3): New feature flag for sha3.
14711 * config/aarch64/iterators.md (sha512_op): New int attribute.
14712 (CRYPTO_SHA512): New int iterator.
14713 (UNSPEC_SHA512H): New unspec.
14714 (UNSPEC_SHA512H2): Ditto.
14715 (UNSPEC_SHA512SU0): Ditto.
14716 (UNSPEC_SHA512SU1): Ditto.
14717 * config/aarch64/aarch64-simd-builtins.def
14718 (aarch64_crypto_sha512hqv2di): New builtin.
14719 (aarch64_crypto_sha512h2qv2di): Ditto.
14720 (aarch64_crypto_sha512su0qv2di): Ditto.
14721 (aarch64_crypto_sha512su1qv2di): Ditto.
14722 (aarch64_eor3qv8hi): Ditto.
14723 (aarch64_rax1qv2di): Ditto.
14724 (aarch64_xarqv2di): Ditto.
14725 (aarch64_bcaxqv8hi): Ditto.
14726 * config/aarch64/aarch64-simd.md:
14727 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14728 (aarch64_crypto_sha512su0qv2di): Ditto.
14729 (aarch64_crypto_sha512su1qv2di): Ditto.
14730 (aarch64_eor3qv8hi): Ditto.
14731 (aarch64_rax1qv2di): Ditto.
14732 (aarch64_xarqv2di): Ditto.
14733 (aarch64_bcaxqv8hi): Ditto.
14734 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14735 (vsha512h2q_u64): Ditto.
14736 (vsha512su0q_u64): Ditto.
14737 (vsha512su1q_u64): Ditto.
14738 (veor3q_u16): Ditto.
14739 (vrax1q_u64): Ditto.
14740 (vxarq_u64): Ditto.
14741 (vbcaxq_u16): Ditto.
14742 * config/arm/types.md (crypto_sha512): New type attribute.
14743 (crypto_sha3): Ditto.
14744 (doc/invoke.texi): Document new sha3 option.
14746 2018-01-10 Michael Collison <michael.collison@arm.com>
14748 * config/aarch64/aarch64-builtins.c:
14749 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14751 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14752 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14753 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14754 (AARCH64_ISA_SM4): New ISA flag.
14755 (TARGET_SM4): New feature flag for sm4.
14756 * config/aarch64/aarch64-simd-builtins.def
14757 (aarch64_sm3ss1qv4si): Ditto.
14758 (aarch64_sm3tt1aq4si): Ditto.
14759 (aarch64_sm3tt1bq4si): Ditto.
14760 (aarch64_sm3tt2aq4si): Ditto.
14761 (aarch64_sm3tt2bq4si): Ditto.
14762 (aarch64_sm3partw1qv4si): Ditto.
14763 (aarch64_sm3partw2qv4si): Ditto.
14764 (aarch64_sm4eqv4si): Ditto.
14765 (aarch64_sm4ekeyqv4si): Ditto.
14766 * config/aarch64/aarch64-simd.md:
14767 (aarch64_sm3ss1qv4si): Ditto.
14768 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14769 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14770 (aarch64_sm4eqv4si): Ditto.
14771 (aarch64_sm4ekeyqv4si): Ditto.
14772 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14773 (sm3part_op): Ditto.
14774 (CRYPTO_SM3TT): Ditto.
14775 (CRYPTO_SM3PART): Ditto.
14776 (UNSPEC_SM3SS1): New unspec.
14777 (UNSPEC_SM3TT1A): Ditto.
14778 (UNSPEC_SM3TT1B): Ditto.
14779 (UNSPEC_SM3TT2A): Ditto.
14780 (UNSPEC_SM3TT2B): Ditto.
14781 (UNSPEC_SM3PARTW1): Ditto.
14782 (UNSPEC_SM3PARTW2): Ditto.
14783 (UNSPEC_SM4E): Ditto.
14784 (UNSPEC_SM4EKEY): Ditto.
14785 * config/aarch64/constraints.md (Ui2): New constraint.
14786 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14787 * config/arm/types.md (crypto_sm3): New type attribute.
14788 (crypto_sm4): Ditto.
14789 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14790 (vsm3tt1aq_u32): Ditto.
14791 (vsm3tt1bq_u32): Ditto.
14792 (vsm3tt2aq_u32): Ditto.
14793 (vsm3tt2bq_u32): Ditto.
14794 (vsm3partw1q_u32): Ditto.
14795 (vsm3partw2q_u32): Ditto.
14796 (vsm4eq_u32): Ditto.
14797 (vsm4ekeyq_u32): Ditto.
14798 (doc/invoke.texi): Document new sm4 option.
14800 2018-01-10 Michael Collison <michael.collison@arm.com>
14802 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14803 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14804 (AARCH64_FL_FOR_ARCH8_4): New.
14805 (AARCH64_FL_V8_4): New flag.
14806 (doc/invoke.texi): Document new armv8.4-a option.
14808 2018-01-10 Michael Collison <michael.collison@arm.com>
14810 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14811 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14812 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14813 * config/aarch64/aarch64-option-extension.def: Add
14814 AARCH64_OPT_EXTENSION of 'sha2'.
14815 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14816 (crypto): Disable sha2 and aes if crypto disabled.
14817 (crypto): Enable aes and sha2 if enabled.
14818 (simd): Disable sha2 and aes if simd disabled.
14819 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14821 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14822 (TARGET_SHA2): New feature flag for sha2.
14823 (TARGET_AES): New feature flag for aes.
14824 * config/aarch64/aarch64-simd.md:
14825 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14826 conditional on TARGET_AES.
14827 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14828 (aarch64_crypto_sha1hsi): Make pattern conditional
14830 (aarch64_crypto_sha1hv4si): Ditto.
14831 (aarch64_be_crypto_sha1hv4si): Ditto.
14832 (aarch64_crypto_sha1su1v4si): Ditto.
14833 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14834 (aarch64_crypto_sha1su0v4si): Ditto.
14835 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14836 (aarch64_crypto_sha256su0v4si): Ditto.
14837 (aarch64_crypto_sha256su1v4si): Ditto.
14838 (doc/invoke.texi): Document new aes and sha2 options.
14840 2018-01-10 Martin Sebor <msebor@redhat.com>
14842 PR tree-optimization/83781
14843 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14846 2018-01-11 Martin Sebor <msebor@gmail.com>
14847 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14849 PR tree-optimization/83501
14850 PR tree-optimization/81703
14852 * tree-ssa-strlen.c (get_string_cst): Rename...
14853 (get_string_len): ...to this. Handle global constants.
14854 (handle_char_store): Adjust.
14856 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14857 Jim Wilson <jimw@sifive.com>
14859 * config/riscv/riscv-protos.h (riscv_output_return): New.
14860 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14861 (riscv_attribute_table, riscv_output_return),
14862 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14863 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14864 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14866 (riscv_expand_prologue): Add early return for naked function.
14867 (riscv_expand_epilogue): Likewise.
14868 (riscv_function_ok_for_sibcall): Return false for naked function.
14869 (riscv_set_current_function): New.
14870 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14871 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14872 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14873 * doc/extend.texi (RISC-V Function Attributes): New.
14875 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14877 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14878 check for 128-bit long double before checking TCmode.
14879 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14880 128-bit long doubles before checking TFmode or TCmode.
14881 (FLOAT128_IBM_P): Likewise.
14883 2018-01-10 Martin Sebor <msebor@redhat.com>
14885 PR tree-optimization/83671
14886 * builtins.c (c_strlen): Unconditionally return zero for the empty
14888 Use -Warray-bounds for warnings.
14889 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14890 for non-constant array indices with COMPONENT_REF, arrays of
14891 arrays, and pointers to arrays.
14892 (gimple_fold_builtin_strlen): Determine and set length range for
14893 non-constant character arrays.
14895 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14897 PR middle-end/81897
14898 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14901 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14903 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14905 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14908 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14909 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14910 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14911 indexed_or_indirect_operand predicate.
14912 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14913 (*vsx_le_perm_load_v8hi): Likewise.
14914 (*vsx_le_perm_load_v16qi): Likewise.
14915 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14916 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14917 (*vsx_le_perm_store_v8hi): Likewise.
14918 (*vsx_le_perm_store_v16qi): Likewise.
14919 (eight unnamed splitters): Likewise.
14921 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14923 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14924 * config/rs6000/emmintrin.h: Likewise.
14925 * config/rs6000/mmintrin.h: Likewise.
14926 * config/rs6000/xmmintrin.h: Likewise.
14928 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14931 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14933 * tree.c (tree_nop_conversion): Return true for location wrapper
14935 (maybe_wrap_with_location): New function.
14936 (selftest::check_strip_nops): New function.
14937 (selftest::test_location_wrappers): New function.
14938 (selftest::tree_c_tests): Call it.
14939 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14940 (maybe_wrap_with_location): New decl.
14941 (EXPR_LOCATION_WRAPPER_P): New macro.
14942 (location_wrapper_p): New inline function.
14943 (tree_strip_any_location_wrapper): New inline function.
14945 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14948 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14949 stack_realign_offset for the largest alignment of stack slot
14951 (ix86_find_max_used_stack_alignment): New function.
14952 (ix86_finalize_stack_frame_flags): Use it. Set
14953 max_used_stack_alignment if we don't realign stack.
14954 * config/i386/i386.h (machine_function): Add
14955 max_used_stack_alignment.
14957 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14959 * config/arm/arm.opt (-mbranch-cost): New option.
14960 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14963 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14966 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14967 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14969 2018-01-10 Richard Biener <rguenther@suse.de>
14972 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14973 early out so it also covers the case where we have a non-NULL
14976 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14978 PR tree-optimization/83753
14979 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14980 for non-strided grouped accesses if the number of elements is 1.
14982 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14985 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14986 * i386.h (TARGET_USE_GATHER): Define.
14987 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14989 2018-01-10 Martin Liska <mliska@suse.cz>
14992 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14993 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14995 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14996 CLEANUP_NO_PARTITIONING is not set.
14998 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15000 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15001 for vectors, as a partial revert of r254296.
15002 * rtl.h (const_vec_p): Delete.
15003 (const_vec_duplicate_p): Don't test for vector CONSTs.
15004 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15005 * expmed.c (make_tree): Likewise.
15008 * common.md (E, F): Use CONSTANT_P instead of checking for
15010 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15011 checking for CONST_VECTOR.
15013 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15015 PR middle-end/83575
15016 * predict.c (force_edge_cold): Handle in more sane way edges
15017 with no prediction.
15019 2018-01-09 Carl Love <cel@us.ibm.com>
15021 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15023 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15024 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15025 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15026 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15027 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15028 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15029 * config/rs6000/rs6000-protos.h: Add extern defition for
15030 rs6000_generate_float2_double_code.
15031 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15033 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15034 (float2_v2df): Add define_expand.
15036 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15039 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15040 op_mode in the force_to_mode call.
15042 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15044 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15045 instead of checking each element individually.
15046 (aarch64_evpc_uzp): Likewise.
15047 (aarch64_evpc_zip): Likewise.
15048 (aarch64_evpc_ext): Likewise.
15049 (aarch64_evpc_rev): Likewise.
15050 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15051 instead of checking each element individually. Return true without
15053 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15054 whether all selected elements come from the same input, instead of
15055 checking each element individually. Remove calls to gen_rtx_REG,
15056 start_sequence and end_sequence and instead assert that no rtl is
15059 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15061 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15062 order of HIGH and CONST checks.
15064 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15066 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15067 if the destination isn't an SSA_NAME.
15069 2018-01-09 Richard Biener <rguenther@suse.de>
15071 PR tree-optimization/83668
15072 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15074 (canonicalize_loop_form): ... here, renamed from ...
15075 (canonicalize_loop_closed_ssa_form): ... this and amended to
15076 swap successor edges for loop exit blocks to make us use
15077 the RPO order we need for initial schedule generation.
15079 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15081 PR tree-optimization/64811
15082 * match.pd: When optimizing comparisons with Inf, avoid
15083 introducing or losing exceptions from comparisons with NaN.
15085 2018-01-09 Martin Liska <mliska@suse.cz>
15088 * asan.c (shadow_mem_size): Add gcc_assert.
15090 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15092 Don't save registers in main().
15095 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15096 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15097 * config/avr/avr.c (avr_set_current_function): Don't error if
15098 naked, OS_task or OS_main are specified at the same time.
15099 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15101 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15103 * common/config/avr/avr-common.c (avr_option_optimization_table):
15104 Switch on -mmain-is-OS_task for optimizing compilations.
15106 2018-01-09 Richard Biener <rguenther@suse.de>
15108 PR tree-optimization/83572
15109 * graphite.c: Include cfganal.h.
15110 (graphite_transform_loops): Connect infinite loops to exit
15111 and remove fake edges at the end.
15113 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15115 * ipa-inline.c (edge_badness): Revert accidental checkin.
15117 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15120 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15121 symbols; not inline clones.
15123 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15126 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15127 hard registers. Formatting fixes.
15129 PR preprocessor/83722
15130 * gcc.c (try_generate_repro): Pass
15131 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15132 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15135 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15136 Kito Cheng <kito.cheng@gmail.com>
15138 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15139 (riscv_leaf_function_p): Delete.
15140 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15142 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15144 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15146 (do_ifelse): New function.
15147 (do_isel): New function.
15148 (do_sub3): New function.
15149 (do_add3): New function.
15150 (do_load_mask_compare): New function.
15151 (do_overlap_load_compare): New function.
15152 (expand_compare_loop): New function.
15153 (expand_block_compare): Call expand_compare_loop() when appropriate.
15154 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15155 option description.
15156 (-mblock-compare-inline-loop-limit): New option.
15158 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15161 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15162 Reverse order of second and third operands in first alternative.
15163 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15164 of first and second elements in UNSPEC_VPERMR vector.
15165 (altivec_expand_vec_perm_le): Likewise.
15167 2018-01-08 Jeff Law <law@redhat.com>
15169 PR rtl-optimizatin/81308
15170 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15171 (process_switch): If group_case_labels makes a change, then set
15173 (pass_convert_switch::execute): If a switch is converted, then
15174 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15176 PR rtl-optimization/81308
15177 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15180 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15182 PR target/83663 - Revert r255946
15183 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15184 generation for cases where splatting a value is not useful.
15185 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15186 across a vec_duplicate and a paradoxical subreg forming a vector
15187 mode to a vec_concat.
15189 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15191 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15192 -march=armv8.3-a variants.
15193 * config/arm/t-multilib: Likewise.
15194 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15196 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15198 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15200 (cceq_ior_compare_complement): Give it a name so I can use it, and
15201 change boolean_or_operator predicate to boolean_operator so it can
15202 be used to generate a crand.
15203 (eqne): New code iterator.
15204 (bd/bd_neg): New code_attrs.
15205 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15206 a single define_insn.
15207 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15208 decrement (bdnzt/bdnzf/bdzt/bdzf).
15209 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15210 with the new names of the branch decrement patterns, and added the
15211 names of the branch decrement conditional patterns.
15213 2018-01-08 Richard Biener <rguenther@suse.de>
15215 PR tree-optimization/83563
15216 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15219 2018-01-08 Richard Biener <rguenther@suse.de>
15221 PR middle-end/83713
15222 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15224 2018-01-08 Richard Biener <rguenther@suse.de>
15226 PR tree-optimization/83685
15227 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15228 references to abnormals.
15230 2018-01-08 Richard Biener <rguenther@suse.de>
15233 * dwarf2out.c (output_indirect_strings): Handle empty
15234 skeleton_debug_str_hash.
15235 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15237 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15239 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15240 (emit_store_direct): Likewise.
15241 (arc_trampoline_adjust_address): Likewise.
15242 (arc_asm_trampoline_template): New function.
15243 (arc_initialize_trampoline): Use asm_trampoline_template.
15244 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15245 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15246 * config/arc/arc.md (flush_icache): Delete pattern.
15248 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15250 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15251 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15254 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15257 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15258 by not USED_FOR_TARGET.
15259 (make_pass_resolve_sw_modes): Likewise.
15261 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15263 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15266 2018-01-08 Richard Biener <rguenther@suse.de>
15268 PR middle-end/83580
15269 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15271 2018-01-08 Richard Biener <rguenther@suse.de>
15273 PR middle-end/83517
15274 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15276 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15278 PR middle-end/81897
15279 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15280 basic blocks with a small number of successors.
15281 (convert_control_dep_chain_into_preds): Improve handling of
15283 (dump_predicates): Split apart into...
15284 (dump_pred_chain): ...here...
15285 (dump_pred_info): ...and here.
15286 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15287 (can_chain_union_be_invalidated_p): Improve check for invalidation
15289 (uninit_uses_cannot_happen): Avoid unnecessary if
15290 convert_control_dep_chain_into_preds yielded nothing.
15292 2018-01-06 Martin Sebor <msebor@redhat.com>
15294 PR tree-optimization/83640
15295 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15296 subtracting negative offset from size.
15297 (builtin_access::overlap): Adjust offset bounds of the access to fall
15298 within the size of the object if possible.
15300 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15302 PR rtl-optimization/83699
15303 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15304 extract_bit_field_as_subreg to cases in which the extracted
15305 value is also a vector.
15307 * lra-constraints.c (process_alt_operands): Test for the equivalence
15308 substitutions when detecting a possible reload cycle.
15310 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15313 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15314 by default if flag_selective_schedling{,2}. Formatting fixes.
15316 PR rtl-optimization/83682
15317 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15318 if it has non-VECTOR_MODE element mode.
15319 (vec_duplicate_p): Likewise.
15321 PR middle-end/83694
15322 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15323 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15325 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15328 * config/i386/i386-builtin.def
15329 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15330 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15331 Require also OPTION_MASK_ISA_AVX512F in addition to
15332 OPTION_MASK_ISA_GFNI.
15333 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15334 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15335 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15336 to OPTION_MASK_ISA_GFNI.
15337 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15338 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15339 OPTION_MASK_ISA_AVX512BW.
15340 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15341 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15342 addition to OPTION_MASK_ISA_GFNI.
15343 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15344 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15345 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15346 to OPTION_MASK_ISA_GFNI.
15347 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15348 a requirement for all ISAs rather than any of them with a few
15350 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15352 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15353 bitmasks to be enabled with 3 exceptions, instead of requiring any
15354 enabled ISA with lots of exceptions.
15355 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15356 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15357 Change avx512bw in isa attribute to avx512f.
15358 * config/i386/sgxintrin.h: Add license boilerplate.
15359 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15360 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15361 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15362 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15364 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15365 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15366 temporarily sse2 rather than sse if not enabled already.
15369 * config/i386/sse.md (VI248_VLBW): Rename to ...
15370 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15371 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15372 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15373 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15374 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15375 mode iterator instead of VI248_VLBW.
15377 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15379 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15380 (record_modified): Skip clobbers; add debug output.
15381 (param_change_prob): Use sreal frequencies.
15383 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15385 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15386 punt for user-aligned variables.
15388 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15390 * tree-chrec.c (chrec_contains_symbols): Return true for
15393 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15396 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15397 of (x|y) == x for BICS pattern.
15399 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15401 PR tree-optimization/83605
15402 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15403 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15406 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15408 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15409 * config/epiphany/rtems.h: New file.
15411 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15412 Uros Bizjak <ubizjak@gmail.com>
15415 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15416 QIreg_operand instead of register_operand predicate.
15417 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15418 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15419 comments instead of -fmitigate[-_]rop.
15421 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15424 * cgraphunit.c (symbol_table::compile): Switch to text_section
15425 before calling assembly_start debug hook.
15426 * run-rtl-passes.c (run_rtl_passes): Likewise.
15429 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15431 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15432 range_int_cst_p rather than !symbolic_range_p before calling
15433 extract_range_from_multiplicative_op_1.
15435 2018-01-04 Jeff Law <law@redhat.com>
15437 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15438 redundant test in assertion.
15440 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15442 * doc/rtl.texi: Document machine_mode wrapper classes.
15444 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15446 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15447 using tree_to_uhwi.
15449 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15451 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15452 the VEC_PERM_EXPR fold to fail.
15454 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15457 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15458 to switched_sections.
15460 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15463 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15464 test for d.testing.
15466 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15469 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15470 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15472 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15475 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15476 is BLKmode and bitpos not zero or mode change is needed.
15478 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15481 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15484 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15487 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15488 instead of MULT rtx. Update all corresponding splitters.
15489 (*saddl_se): Ditto.
15490 (*ssub<modesuffix>): Ditto.
15491 (*ssubl_se): Ditto.
15492 (*cmp_sadd_di): Update split patterns.
15493 (*cmp_sadd_si): Ditto.
15494 (*cmp_sadd_sidi): Ditto.
15495 (*cmp_ssub_di): Ditto.
15496 (*cmp_ssub_si): Ditto.
15497 (*cmp_ssub_sidi): Ditto.
15498 * config/alpha/predicates.md (const23_operand): New predicate.
15499 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15500 Look for ASHIFT, not MULT inner operand.
15501 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15503 2018-01-04 Martin Liska <mliska@suse.cz>
15505 PR gcov-profile/83669
15506 * gcov.c (output_intermediate_file): Add version to intermediate
15508 * doc/gcov.texi: Document new field 'version' in intermediate
15509 file format. Fix location of '-k' option of gcov command.
15511 2018-01-04 Martin Liska <mliska@suse.cz>
15514 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15516 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15518 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15520 2018-01-03 Martin Sebor <msebor@redhat.com>
15522 PR tree-optimization/83655
15523 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15524 checking calls with invalid arguments.
15526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15528 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15529 (vectorizable_mask_load_store): Delete.
15530 (vectorizable_call): Return false for masked loads and stores.
15531 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15532 instead of gimple_assign_rhs1.
15533 (vectorizable_load): Handle IFN_MASK_LOAD.
15534 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15538 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15540 (vectorizable_mask_load_store): ...here.
15541 (vectorizable_load): ...and here.
15543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15545 * tree-vect-stmts.c (vect_build_all_ones_mask)
15546 (vect_build_zero_merge_argument): New functions, split out from...
15547 (vectorizable_load): ...here.
15549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15551 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15553 (vectorizable_mask_load_store): ...here.
15554 (vectorizable_store): ...and here.
15556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15558 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15560 (vectorizable_mask_load_store): ...here.
15562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15564 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15565 (vect_model_store_cost): Take a vec_load_store_type instead of a
15567 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15568 (vect_model_store_cost): Take a vec_load_store_type instead of a
15570 (vectorizable_mask_load_store): Update accordingly.
15571 (vectorizable_store): Likewise.
15572 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15576 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15577 IFN_MASK_LOAD calls here rather than...
15578 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15581 Alan Hayward <alan.hayward@arm.com>
15582 David Sherwood <david.sherwood@arm.com>
15584 * expmed.c (extract_bit_field_1): For vector extracts,
15585 fall back to extract_bit_field_as_subreg if vec_extract
15588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15589 Alan Hayward <alan.hayward@arm.com>
15590 David Sherwood <david.sherwood@arm.com>
15592 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15593 they are variable or constant sized.
15594 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15595 slots for constant-sized data.
15597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15598 Alan Hayward <alan.hayward@arm.com>
15599 David Sherwood <david.sherwood@arm.com>
15601 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15602 handling COND_EXPRs with boolean comparisons, try to find a better
15603 basis for the mask type than the boolean itself.
15605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15607 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15608 is calculated and how it can be overridden.
15609 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15610 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15612 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15616 Alan Hayward <alan.hayward@arm.com>
15617 David Sherwood <david.sherwood@arm.com>
15619 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15620 Remove the mode argument.
15621 (aarch64_simd_valid_immediate): Remove the mode and inverse
15623 * config/aarch64/iterators.md (bitsize): New iterator.
15624 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15625 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15626 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15627 aarch64_simd_valid_immediate.
15628 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15629 (aarch64_reg_or_bic_imm): Likewise.
15630 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15631 with an insn_type enum and msl with a modifier_type enum.
15632 Replace element_width with a scalar_mode. Change the shift
15633 to unsigned int. Add constructors for scalar_float_mode and
15634 scalar_int_mode elements.
15635 (aarch64_vect_float_const_representable_p): Delete.
15636 (aarch64_can_const_movi_rtx_p)
15637 (aarch64_simd_scalar_immediate_valid_for_move)
15638 (aarch64_simd_make_constant): Update call to
15639 aarch64_simd_valid_immediate.
15640 (aarch64_advsimd_valid_immediate_hs): New function.
15641 (aarch64_advsimd_valid_immediate): Likewise.
15642 (aarch64_simd_valid_immediate): Remove mode and inverse
15643 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15644 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15645 and aarch64_float_const_representable_p on the result.
15646 (aarch64_output_simd_mov_immediate): Remove mode argument.
15647 Update call to aarch64_simd_valid_immediate and use of
15648 simd_immediate_info.
15649 (aarch64_output_scalar_simd_mov_immediate): Update call
15652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15653 Alan Hayward <alan.hayward@arm.com>
15654 David Sherwood <david.sherwood@arm.com>
15656 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15657 (mode_nunits): Likewise CONST_MODE_NUNITS.
15658 * machmode.def (ADJUST_NUNITS): Document.
15659 * genmodes.c (mode_data::need_nunits_adj): New field.
15660 (blank_mode): Update accordingly.
15661 (adj_nunits): New variable.
15662 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15664 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15665 listed in adj_nunits.
15666 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15667 listed in adj_nunits. Don't emit case statements for such modes.
15668 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15669 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15670 nothing if adj_nunits is nonnull.
15671 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15672 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15673 (emit_mode_fbit): Update use of print_maybe_const_decl.
15674 (emit_move_size): Likewise. Treat the array as non-const
15676 (emit_mode_adjustments): Handle adj_nunits.
15678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15680 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15681 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15682 (VECTOR_MODES): Use it.
15683 (make_vector_modes): Take the prefix as an argument.
15685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15686 Alan Hayward <alan.hayward@arm.com>
15687 David Sherwood <david.sherwood@arm.com>
15689 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15690 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15691 for MODE_VECTOR_BOOL.
15692 * machmode.def (VECTOR_BOOL_MODE): Document.
15693 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15694 (make_vector_bool_mode): New function.
15695 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15697 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15698 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15700 * stor-layout.c (int_mode_for_mode): Likewise.
15701 * tree.c (build_vector_type_for_mode): Likewise.
15702 * varasm.c (output_constant_pool_2): Likewise.
15703 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15704 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15705 for MODE_VECTOR_BOOL.
15706 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15707 of mode class checks.
15708 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15709 instead of a list of mode class checks.
15710 (expand_vector_scalar_condition): Likewise.
15711 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15714 Alan Hayward <alan.hayward@arm.com>
15715 David Sherwood <david.sherwood@arm.com>
15717 * machmode.h (mode_size): Change from unsigned short to
15719 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15720 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15721 or if measurement_type is not polynomial.
15722 (fixed_size_mode::includes_p): Check for constant-sized modes.
15723 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15724 return a poly_uint16 rather than an unsigned short.
15725 (emit_mode_size): Change the type of mode_size from unsigned short
15726 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15727 (emit_mode_adjustments): Cope with polynomial vector sizes.
15728 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15730 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15732 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15733 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15734 * caller-save.c (setup_save_areas): Likewise.
15735 (replace_reg_with_saved_mem): Likewise.
15736 * calls.c (emit_library_call_value_1): Likewise.
15737 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15738 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15739 (gen_lowpart_for_combine): Likewise.
15740 * convert.c (convert_to_integer_1): Likewise.
15741 * cse.c (equiv_constant, cse_insn): Likewise.
15742 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15743 (cselib_subst_to_values): Likewise.
15744 * dce.c (word_dce_process_block): Likewise.
15745 * df-problems.c (df_word_lr_mark_ref): Likewise.
15746 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15747 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15748 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15749 (rtl_for_decl_location): Likewise.
15750 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15751 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15752 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15753 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15754 (expand_expr_real_1): Likewise.
15755 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15756 (pad_below): Likewise.
15757 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15758 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15759 * ira.c (get_subreg_tracking_sizes): Likewise.
15760 * ira-build.c (ira_create_allocno_objects): Likewise.
15761 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15762 (ira_sort_regnos_for_alter_reg): Likewise.
15763 * ira-costs.c (record_operand_costs): Likewise.
15764 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15765 (resolve_simple_move): Likewise.
15766 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15767 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15768 (lra_constraints): Likewise.
15769 (CONST_POOL_OK_P): Reject variable-sized modes.
15770 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15771 (add_pseudo_to_slot, lra_spill): Likewise.
15772 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15773 * optabs-query.c (get_best_extraction_insn): Likewise.
15774 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15775 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15776 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15777 * recog.c (offsettable_address_addr_space_p): Likewise.
15778 * regcprop.c (maybe_mode_change): Likewise.
15779 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15780 * regrename.c (build_def_use): Likewise.
15781 * regstat.c (dump_reg_info): Likewise.
15782 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15783 (find_reloads, find_reloads_subreg_address): Likewise.
15784 * reload1.c (eliminate_regs_1): Likewise.
15785 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15786 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15787 (simplify_binary_operation_1, simplify_subreg): Likewise.
15788 * targhooks.c (default_function_arg_padding): Likewise.
15789 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15790 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15791 (verify_gimple_assign_ternary): Likewise.
15792 * tree-inline.c (estimate_move_cost): Likewise.
15793 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15794 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15795 (get_address_cost_ainc): Likewise.
15796 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15797 (vect_supportable_dr_alignment): Likewise.
15798 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15799 (vectorizable_reduction): Likewise.
15800 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15801 (vectorizable_operation, vectorizable_load): Likewise.
15802 * tree.c (build_same_sized_truth_vector_type): Likewise.
15803 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15804 * var-tracking.c (emit_note_insn_var_location): Likewise.
15805 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15806 (ADDR_VEC_ALIGN): Likewise.
15808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15809 Alan Hayward <alan.hayward@arm.com>
15810 David Sherwood <david.sherwood@arm.com>
15812 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15814 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15815 or if measurement_type is polynomial.
15816 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15817 * combine.c (make_extraction): Likewise.
15818 * dse.c (find_shift_sequence): Likewise.
15819 * dwarf2out.c (mem_loc_descriptor): Likewise.
15820 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15821 (extract_bit_field, extract_low_bits): Likewise.
15822 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15823 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15824 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15825 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15826 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15827 * reload.c (find_reloads): Likewise.
15828 * reload1.c (alter_reg): Likewise.
15829 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15830 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15831 * tree-if-conv.c (predicate_mem_writes): Likewise.
15832 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15833 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15834 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15835 * valtrack.c (dead_debug_insert_temp): Likewise.
15836 * varasm.c (mergeable_constant_section): Likewise.
15837 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15840 Alan Hayward <alan.hayward@arm.com>
15841 David Sherwood <david.sherwood@arm.com>
15843 * expr.c (expand_assignment): Cope with polynomial mode sizes
15844 when assigning to a CONCAT.
15846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15847 Alan Hayward <alan.hayward@arm.com>
15848 David Sherwood <david.sherwood@arm.com>
15850 * machmode.h (mode_precision): Change from unsigned short to
15852 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15854 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15855 or if measurement_type is not polynomial.
15856 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15857 in which the mode is already known to be a scalar_int_mode.
15858 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15859 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15861 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15862 for GET_MODE_PRECISION.
15863 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15864 for GET_MODE_PRECISION.
15865 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15867 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15868 (expand_field_assignment, make_extraction): Likewise.
15869 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15870 (get_last_value): Likewise.
15871 * convert.c (convert_to_integer_1): Likewise.
15872 * cse.c (cse_insn): Likewise.
15873 * expr.c (expand_expr_real_1): Likewise.
15874 * lra-constraints.c (simplify_operand_subreg): Likewise.
15875 * optabs-query.c (can_atomic_load_p): Likewise.
15876 * optabs.c (expand_atomic_load): Likewise.
15877 (expand_atomic_store): Likewise.
15878 * ree.c (combine_reaching_defs): Likewise.
15879 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15880 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15881 * tree.h (type_has_mode_precision_p): Likewise.
15882 * ubsan.c (instrument_si_overflow): Likewise.
15884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15885 Alan Hayward <alan.hayward@arm.com>
15886 David Sherwood <david.sherwood@arm.com>
15888 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15889 polynomial numbers of units.
15890 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15891 (valid_vector_subparts_p): New function.
15892 (build_vector_type): Remove temporary shim and take the number
15893 of units as a poly_uint64 rather than an int.
15894 (build_opaque_vector_type): Take the number of units as a
15895 poly_uint64 rather than an int.
15896 * tree.c (build_vector_from_ctor): Handle polynomial
15897 TYPE_VECTOR_SUBPARTS.
15898 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15899 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15900 (build_vector_from_val): If the number of units is variable,
15901 use build_vec_duplicate_cst for constant operands and
15902 VEC_DUPLICATE_EXPR otherwise.
15903 (make_vector_type): Remove temporary is_constant ().
15904 (build_vector_type, build_opaque_vector_type): Take the number of
15905 units as a poly_uint64 rather than an int.
15906 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15908 * cfgexpand.c (expand_debug_expr): Likewise.
15909 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15910 (store_constructor, expand_expr_real_1): Likewise.
15911 (const_scalar_mask_from_tree): Likewise.
15912 * fold-const-call.c (fold_const_reduction): Likewise.
15913 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15914 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15915 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15916 (fold_relational_const): Likewise.
15917 (native_interpret_vector): Likewise. Change the size from an
15918 int to an unsigned int.
15919 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15920 TYPE_VECTOR_SUBPARTS.
15921 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15922 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15923 duplicating a non-constant operand into a variable-length vector.
15924 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15925 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15926 * ipa-icf.c (sem_variable::equals): Likewise.
15927 * match.pd: Likewise.
15928 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15929 * print-tree.c (print_node): Likewise.
15930 * stor-layout.c (layout_type): Likewise.
15931 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15932 * tree-cfg.c (verify_gimple_comparison): Likewise.
15933 (verify_gimple_assign_binary): Likewise.
15934 (verify_gimple_assign_ternary): Likewise.
15935 (verify_gimple_assign_single): Likewise.
15936 * tree-pretty-print.c (dump_generic_node): Likewise.
15937 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15938 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15939 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15940 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15941 (vect_shift_permute_load_chain): Likewise.
15942 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15943 (expand_vector_condition, optimize_vector_constructor): Likewise.
15944 (lower_vec_perm, get_compute_type): Likewise.
15945 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15946 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15947 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15948 (vect_recog_mask_conversion_pattern): Likewise.
15949 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15950 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15951 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15952 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15953 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15954 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15955 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15956 (supportable_widening_operation): Likewise.
15957 (supportable_narrowing_operation): Likewise.
15958 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15960 * varasm.c (output_constant): Likewise.
15962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15963 Alan Hayward <alan.hayward@arm.com>
15964 David Sherwood <david.sherwood@arm.com>
15966 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15967 so that both the length == 3 and length != 3 cases set up their
15968 own permute vectors. Add comments explaining why we know the
15969 number of elements is constant.
15970 (vect_permute_load_chain): Likewise.
15972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15973 Alan Hayward <alan.hayward@arm.com>
15974 David Sherwood <david.sherwood@arm.com>
15976 * machmode.h (mode_nunits): Change from unsigned char to
15978 (ONLY_FIXED_SIZE_MODES): New macro.
15979 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15980 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15981 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15983 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15984 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15985 or if measurement_type is not polynomial.
15986 * genmodes.c (ZERO_COEFFS): New macro.
15987 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15989 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15990 Use ZERO_COEFFS when emitting initializers.
15991 * data-streamer.h (bp_pack_poly_value): New function.
15992 (bp_unpack_poly_value): Likewise.
15993 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15994 for GET_MODE_NUNITS.
15995 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15996 for GET_MODE_NUNITS.
15997 * tree.c (make_vector_type): Remove temporary shim and make
15998 the real function take the number of units as a poly_uint64
15999 rather than an int.
16000 (build_vector_type_for_mode): Handle polynomial nunits.
16001 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16002 * emit-rtl.c (const_vec_series_p_1): Likewise.
16003 (gen_rtx_CONST_VECTOR): Likewise.
16004 * fold-const.c (test_vec_duplicate_folding): Likewise.
16005 * genrecog.c (validate_pattern): Likewise.
16006 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16007 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16008 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16009 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16010 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16011 * rtlanal.c (subreg_get_info): Likewise.
16012 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16013 (vect_grouped_load_supported): Likewise.
16014 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16015 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16016 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16017 (simplify_const_unary_operation, simplify_binary_operation_1)
16018 (simplify_const_binary_operation, simplify_ternary_operation)
16019 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16020 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16021 instead of CONST_VECTOR_NUNITS.
16022 * varasm.c (output_constant_pool_2): Likewise.
16023 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16024 explicit-encoded elements in the XVEC for variable-length vectors.
16026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16028 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16031 Alan Hayward <alan.hayward@arm.com>
16032 David Sherwood <david.sherwood@arm.com>
16034 * coretypes.h (fixed_size_mode): Declare.
16035 (fixed_size_mode_pod): New typedef.
16036 * builtins.h (target_builtins::x_apply_args_mode)
16037 (target_builtins::x_apply_result_mode): Change type to
16038 fixed_size_mode_pod.
16039 * builtins.c (apply_args_size, apply_result_size, result_vector)
16040 (expand_builtin_apply_args_1, expand_builtin_apply)
16041 (expand_builtin_return): Update accordingly.
16043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16045 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16046 * cselib.c (cselib_hash_rtx): Likewise.
16047 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16048 CONST_VECTOR encoding.
16050 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16051 Jeff Law <law@redhat.com>
16054 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16055 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16056 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16057 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16060 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16061 explicitly probe *sp in a noreturn function if there were any callee
16062 register saves or frame pointer is needed.
16064 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16067 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16068 BLKmode for ternary, binary or unary expressions.
16071 * var-tracking.c (delete_vta_debug_insn): New inline function.
16072 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16073 insns from get_insns () to NULL instead of each bb separately.
16074 Use delete_vta_debug_insn. No longer static.
16075 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16076 delete_vta_debug_insns callers.
16077 * rtl.h (delete_vta_debug_insns): Declare.
16078 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16079 instead of variable_tracking_main.
16081 2018-01-03 Martin Sebor <msebor@redhat.com>
16083 PR tree-optimization/83603
16084 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16085 arguments past the endof the argument list in functions declared
16086 without a prototype.
16087 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16088 Avoid checking when arguments are null.
16090 2018-01-03 Martin Sebor <msebor@redhat.com>
16093 * doc/extend.texi (attribute const): Fix a typo.
16094 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16095 issuing -Wsuggest-attribute for void functions.
16097 2018-01-03 Martin Sebor <msebor@redhat.com>
16099 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16100 offset_int::from instead of wide_int::to_shwi.
16101 (maybe_diag_overlap): Remove assertion.
16102 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16103 * gimple-ssa-sprintf.c (format_directive): Same.
16104 (parse_directive): Same.
16105 (sprintf_dom_walker::compute_format_length): Same.
16106 (try_substitute_return_value): Same.
16108 2018-01-03 Jeff Law <law@redhat.com>
16110 PR middle-end/83654
16111 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16112 non-constant residual for zero at runtime and avoid probing in
16113 that case. Reorganize code for trailing problem to mirror handling
16116 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16118 PR tree-optimization/83501
16119 * tree-ssa-strlen.c (get_string_cst): New.
16120 (handle_char_store): Call get_string_cst.
16122 2018-01-03 Martin Liska <mliska@suse.cz>
16124 PR tree-optimization/83593
16125 * tree-ssa-strlen.c: Include tree-cfg.h.
16126 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16127 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16128 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16130 (strlen_dom_walker::before_dom_children): Call
16131 gimple_purge_dead_eh_edges. Dump tranformation with details
16133 (strlen_dom_walker::before_dom_children): Update call by adding
16134 new argument cleanup_eh.
16135 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16137 2018-01-03 Martin Liska <mliska@suse.cz>
16140 * cif-code.def (VARIADIC_THUNK): New enum value.
16141 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16144 2018-01-03 Jan Beulich <jbeulich@suse.com>
16146 * sse.md (mov<mode>_internal): Tighten condition for when to use
16147 vmovdqu<ssescalarsize> for TI and OI modes.
16149 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16151 Update copyright years.
16153 2018-01-03 Martin Liska <mliska@suse.cz>
16156 * ipa-visibility.c (function_and_variable_visibility): Skip
16157 functions with noipa attribure.
16159 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16161 * gcc.c (process_command): Update copyright notice dates.
16162 * gcov-dump.c (print_version): Ditto.
16163 * gcov.c (print_version): Ditto.
16164 * gcov-tool.c (print_version): Ditto.
16165 * gengtype.c (create_file): Ditto.
16166 * doc/cpp.texi: Bump @copying's copyright year.
16167 * doc/cppinternals.texi: Ditto.
16168 * doc/gcc.texi: Ditto.
16169 * doc/gccint.texi: Ditto.
16170 * doc/gcov.texi: Ditto.
16171 * doc/install.texi: Ditto.
16172 * doc/invoke.texi: Ditto.
16174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16176 * vector-builder.h (vector_builder::m_full_nelts): Change from
16177 unsigned int to poly_uint64.
16178 (vector_builder::full_nelts): Update prototype accordingly.
16179 (vector_builder::new_vector): Likewise.
16180 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16181 (vector_builder::operator ==): Likewise.
16182 (vector_builder::finalize): Likewise.
16183 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16184 Take the number of elements as a poly_uint64 rather than an
16186 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16187 from unsigned int to poly_uint64.
16188 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16189 (vec_perm_indices::new_vector): Likewise.
16190 (vec_perm_indices::length): Likewise.
16191 (vec_perm_indices::nelts_per_input): Likewise.
16192 (vec_perm_indices::input_nelts): Likewise.
16193 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16194 number of elements per input as a poly_uint64 rather than an
16195 unsigned int. Use the original encoding for variable-length
16196 vectors, rather than clamping each individual element.
16197 For the second and subsequent elements in each pattern,
16198 clamp the step and base before clamping their sum.
16199 (vec_perm_indices::series_p): Handle polynomial element counts.
16200 (vec_perm_indices::all_in_range_p): Likewise.
16201 (vec_perm_indices_to_tree): Likewise.
16202 (vec_perm_indices_to_rtx): Likewise.
16203 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16204 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16205 (tree_vector_builder::new_binary_operation): Handle polynomial
16206 element counts. Return false if we need to know the number
16207 of elements at compile time.
16208 * fold-const.c (fold_vec_perm): Punt if the number of elements
16209 isn't known at compile time.
16211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16213 * vec-perm-indices.h (vec_perm_builder): Change element type
16214 from HOST_WIDE_INT to poly_int64.
16215 (vec_perm_indices::element_type): Update accordingly.
16216 (vec_perm_indices::clamp): Handle polynomial element_types.
16217 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16218 (vec_perm_indices::all_in_range_p): Likewise.
16219 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16221 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16222 polynomial vec_perm_indices element types.
16223 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16224 * fold-const.c (fold_vec_perm): Likewise.
16225 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16226 * tree-vect-generic.c (lower_vec_perm): Likewise.
16227 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16228 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16229 element type to HOST_WIDE_INT.
16231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16232 Alan Hayward <alan.hayward@arm.com>
16233 David Sherwood <david.sherwood@arm.com>
16235 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16236 rather than an int. Use plus_constant.
16237 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16238 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16241 Alan Hayward <alan.hayward@arm.com>
16242 David Sherwood <david.sherwood@arm.com>
16244 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16245 a HOST_WIDE_INT to a poly_int64.
16247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16248 Alan Hayward <alan.hayward@arm.com>
16249 David Sherwood <david.sherwood@arm.com>
16251 * calls.c (load_register_parameters): Cope with polynomial
16252 mode sizes. Require a constant size for BLKmode parameters
16253 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16254 forces a parameter to be padded at the lsb end in order to
16255 fill a complete number of words, require the parameter size
16256 to be ordered wrt UNITS_PER_WORD.
16258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16259 Alan Hayward <alan.hayward@arm.com>
16260 David Sherwood <david.sherwood@arm.com>
16262 * reload1.c (spill_stack_slot_width): Change element type
16263 from unsigned int to poly_uint64_pod.
16264 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
16271 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
16278 operands whose sizes aren't ordered.
16279 (match_reload): Refer to this check here.
16281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16282 Alan Hayward <alan.hayward@arm.com>
16283 David Sherwood <david.sherwood@arm.com>
16285 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16286 that the mode size is in the set {1, 2, 4, 8, 16}.
16288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16289 Alan Hayward <alan.hayward@arm.com>
16290 David Sherwood <david.sherwood@arm.com>
16292 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16293 Use plus_constant instead of gen_rtx_PLUS.
16295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16296 Alan Hayward <alan.hayward@arm.com>
16297 David Sherwood <david.sherwood@arm.com>
16299 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16300 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16301 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16302 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16303 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16304 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16305 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16306 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16307 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16308 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16310 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16311 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16312 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16313 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16314 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16315 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16316 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16317 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16318 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16319 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16321 * expr.c (emit_move_resolve_push): Treat the input and result
16322 of PUSH_ROUNDING as a poly_int64.
16323 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16324 (emit_push_insn): Likewise.
16325 * lra-eliminations.c (mark_not_eliminable): Likewise.
16326 * recog.c (push_operand): Likewise.
16327 * reload1.c (elimination_effects): Likewise.
16328 * rtlanal.c (nonzero_bits1): Likewise.
16329 * calls.c (store_one_arg): Likewise. Require the padding to be
16330 known at compile time.
16332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16333 Alan Hayward <alan.hayward@arm.com>
16334 David Sherwood <david.sherwood@arm.com>
16336 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16337 Use plus_constant instead of gen_rtx_PLUS.
16339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16340 Alan Hayward <alan.hayward@arm.com>
16341 David Sherwood <david.sherwood@arm.com>
16343 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16344 rather than an int.
16346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16347 Alan Hayward <alan.hayward@arm.com>
16348 David Sherwood <david.sherwood@arm.com>
16350 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16351 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16352 via stack temporaries. Treat the mode size as polynomial too.
16354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16355 Alan Hayward <alan.hayward@arm.com>
16356 David Sherwood <david.sherwood@arm.com>
16358 * expr.c (expand_expr_real_2): When handling conversions involving
16359 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16360 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16361 as a poly_uint64 too.
16363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16364 Alan Hayward <alan.hayward@arm.com>
16365 David Sherwood <david.sherwood@arm.com>
16367 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16370 Alan Hayward <alan.hayward@arm.com>
16371 David Sherwood <david.sherwood@arm.com>
16373 * combine.c (can_change_dest_mode): Handle polynomial
16374 REGMODE_NATURAL_SIZE.
16375 * expmed.c (store_bit_field_1): Likewise.
16376 * expr.c (store_constructor): Likewise.
16377 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16378 and polynomial REGMODE_NATURAL_SIZE.
16379 (gen_lowpart_common): Likewise.
16380 * reginfo.c (record_subregs_of_mode): Likewise.
16381 * rtlanal.c (read_modify_subreg_p): Likewise.
16383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16384 Alan Hayward <alan.hayward@arm.com>
16385 David Sherwood <david.sherwood@arm.com>
16387 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16388 numbers of elements.
16390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16391 Alan Hayward <alan.hayward@arm.com>
16392 David Sherwood <david.sherwood@arm.com>
16394 * match.pd: Cope with polynomial numbers of vector elements.
16396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16397 Alan Hayward <alan.hayward@arm.com>
16398 David Sherwood <david.sherwood@arm.com>
16400 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16401 in a POINTER_PLUS_EXPR.
16403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16404 Alan Hayward <alan.hayward@arm.com>
16405 David Sherwood <david.sherwood@arm.com>
16407 * omp-simd-clone.c (simd_clone_subparts): New function.
16408 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16409 (ipa_simd_modify_function_body): Likewise.
16411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16412 Alan Hayward <alan.hayward@arm.com>
16413 David Sherwood <david.sherwood@arm.com>
16415 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16416 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16417 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16418 (expand_vector_condition, vector_element): Likewise.
16419 (subparts_gt): New function.
16420 (get_compute_type): Use subparts_gt.
16421 (count_type_subparts): Delete.
16422 (expand_vector_operations_1): Use subparts_gt instead of
16423 count_type_subparts.
16425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16426 Alan Hayward <alan.hayward@arm.com>
16427 David Sherwood <david.sherwood@arm.com>
16429 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16430 (vect_compile_time_alias): ...this new function. Do the calculation
16431 on poly_ints rather than trees.
16432 (vect_prune_runtime_alias_test_list): Update call accordingly.
16434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16435 Alan Hayward <alan.hayward@arm.com>
16436 David Sherwood <david.sherwood@arm.com>
16438 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16440 (vect_schedule_slp_instance): Likewise.
16442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16443 Alan Hayward <alan.hayward@arm.com>
16444 David Sherwood <david.sherwood@arm.com>
16446 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16447 constant and extern definitions for variable-length vectors.
16448 (vect_get_constant_vectors): Note that the number of units
16449 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
16456 of units as polynomial. Choose between WIDE and NARROW based
16459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16460 Alan Hayward <alan.hayward@arm.com>
16461 David Sherwood <david.sherwood@arm.com>
16463 * tree-vect-stmts.c (simd_clone_subparts): New function.
16464 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16467 Alan Hayward <alan.hayward@arm.com>
16468 David Sherwood <david.sherwood@arm.com>
16470 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16471 vectors as polynomial. Use build_index_vector for
16472 IFN_GOMP_SIMD_LANE.
16474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16475 Alan Hayward <alan.hayward@arm.com>
16476 David Sherwood <david.sherwood@arm.com>
16478 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16479 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16480 for variable-length vectors.
16481 (vectorizable_mask_load_store): Treat the number of units as
16482 polynomial, asserting that it is constant if the condition has
16483 already been enforced.
16484 (vectorizable_store, vectorizable_load): Likewise.
16486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16487 Alan Hayward <alan.hayward@arm.com>
16488 David Sherwood <david.sherwood@arm.com>
16490 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16491 of units as polynomial. Punt if we can't tell at compile time
16492 which vector contains the final result.
16494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16495 Alan Hayward <alan.hayward@arm.com>
16496 David Sherwood <david.sherwood@arm.com>
16498 * tree-vect-loop.c (vectorizable_induction): Treat the number
16499 of units as polynomial. Punt on SLP inductions. Use an integer
16500 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16501 cast of such a series for variable-length floating-point
16504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16505 Alan Hayward <alan.hayward@arm.com>
16506 David Sherwood <david.sherwood@arm.com>
16508 * tree.h (build_index_vector): Declare.
16509 * tree.c (build_index_vector): New function.
16510 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16511 of units as polynomial, forcibly converting it to a constant if
16512 vectorizable_reduction has already enforced the condition.
16513 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16514 to create a {1,2,3,...} vector.
16515 (vectorizable_reduction): Treat the number of units as polynomial.
16516 Choose vectype_in based on the largest scalar element size rather
16517 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
16525 number of units as polynomial.
16527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16528 Alan Hayward <alan.hayward@arm.com>
16529 David Sherwood <david.sherwood@arm.com>
16531 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16532 * target.def (autovectorize_vector_sizes): Return the vector sizes
16533 by pointer, using vector_sizes rather than a bitmask.
16534 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16535 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16536 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16538 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16539 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16540 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16541 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16542 * omp-general.c (omp_max_vf): Likewise.
16543 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16544 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16545 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16546 * tree-vect-slp.c (vect_slp_bb): Likewise.
16547 * doc/tm.texi: Regenerate.
16548 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16550 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16551 the vector size as a poly_uint64 rather than an unsigned int.
16552 (current_vector_size): Change from an unsigned int to a poly_uint64.
16553 (get_vectype_for_scalar_type): Update accordingly.
16554 * tree.h (build_truth_vector_type): Take the size and number of
16555 units as a poly_uint64 rather than an unsigned int.
16556 (build_vector_type): Add a temporary overload that takes
16557 the number of units as a poly_uint64 rather than an unsigned int.
16558 * tree.c (make_vector_type): Likewise.
16559 (build_truth_vector_type): Take the number of units as a poly_uint64
16560 rather than an unsigned int.
16562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16563 Alan Hayward <alan.hayward@arm.com>
16564 David Sherwood <david.sherwood@arm.com>
16566 * target.def (get_mask_mode): Take the number of units and length
16567 as poly_uint64s rather than unsigned ints.
16568 * targhooks.h (default_get_mask_mode): Update accordingly.
16569 * targhooks.c (default_get_mask_mode): Likewise.
16570 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16571 * doc/tm.texi: Regenerate.
16573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16574 Alan Hayward <alan.hayward@arm.com>
16575 David Sherwood <david.sherwood@arm.com>
16577 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16578 * omp-general.c (omp_max_vf): Likewise.
16579 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16580 (expand_omp_simd): Handle polynomial safelen.
16581 * omp-low.c (omplow_simd_context): Add a default constructor.
16582 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16583 (lower_rec_simd_input_clauses): Update accordingly.
16584 (lower_rec_input_clauses): Likewise.
16586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16587 Alan Hayward <alan.hayward@arm.com>
16588 David Sherwood <david.sherwood@arm.com>
16590 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16591 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16592 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16593 (vect_analyze_slp_cost): Likewise.
16594 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16595 (vect_model_load_cost): Likewise.
16597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16598 Alan Hayward <alan.hayward@arm.com>
16599 David Sherwood <david.sherwood@arm.com>
16601 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16602 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16603 from an unsigned int * to a poly_uint64_pod *.
16604 (calculate_unrolling_factor): New function.
16605 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16608 Alan Hayward <alan.hayward@arm.com>
16609 David Sherwood <david.sherwood@arm.com>
16611 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16612 from an unsigned int to a poly_uint64.
16613 (_loop_vec_info::slp_unrolling_factor): Likewise.
16614 (_loop_vec_info::vectorization_factor): Change from an int
16616 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16617 (vect_get_num_vectors): New function.
16618 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16619 (vect_get_num_copies): Use vect_get_num_vectors.
16620 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16621 to an unsigned int *.
16622 (vect_analyze_data_refs): Change min_vf from an int * to a
16624 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16625 than an unsigned HOST_WIDE_INT.
16626 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16627 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16628 to an unsigned int *.
16629 (vect_analyze_data_ref_dependences): Likewise.
16630 (vect_compute_data_ref_alignment): Handle polynomial vf.
16631 (vect_enhance_data_refs_alignment): Likewise.
16632 (vect_prune_runtime_alias_test_list): Likewise.
16633 (vect_shift_permute_load_chain): Likewise.
16634 (vect_supportable_dr_alignment): Likewise.
16635 (dependence_distance_ge_vf): Take the vectorization factor as a
16636 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16637 (vect_analyze_data_refs): Change min_vf from an int * to a
16639 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16640 vfm1 as a poly_uint64 rather than an int. Make the same change
16641 for the returned bound_scalar.
16642 (vect_gen_vector_loop_niters): Handle polynomial vf.
16643 (vect_do_peeling): Likewise. Update call to
16644 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16645 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16647 * tree-vect-loop.c (vect_determine_vectorization_factor)
16648 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16649 (vect_get_known_peeling_cost): Likewise.
16650 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16651 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16652 (vect_transform_loop): Likewise. Use the lowest possible VF when
16653 updating the upper bounds of the loop.
16654 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16655 rather than an int.
16656 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16657 polynomial unroll factors.
16658 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16659 (vect_make_slp_decision): Likewise.
16660 (vect_supported_load_permutation_p): Likewise, and polynomial
16662 (vect_analyze_slp_cost): Handle polynomial vf.
16663 (vect_slp_analyze_node_operations): Likewise.
16664 (vect_slp_analyze_bb_1): Likewise.
16665 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16666 than an unsigned HOST_WIDE_INT.
16667 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16668 (vectorizable_load): Handle polynomial vf.
16669 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16671 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16674 Alan Hayward <alan.hayward@arm.com>
16675 David Sherwood <david.sherwood@arm.com>
16677 * match.pd: Handle bit operations involving three constants
16678 and try to fold one pair.
16680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16682 * tree-vect-loop-manip.c: Include gimple-fold.h.
16683 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16684 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16685 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16686 Add a path that uses a step of VF instead of 1, but disable it
16688 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16689 and niters_no_overflow parameters. Update calls to
16690 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16691 Create a new SSA name if the latter choses to use a ste other
16692 than zero, and return it via niters_vector_mult_vf_var.
16693 * tree-vect-loop.c (vect_transform_loop): Update calls to
16694 vect_do_peeling, vect_gen_vector_loop_niters and
16695 slpeel_make_loop_iterate_ntimes.
16696 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16697 (vect_gen_vector_loop_niters): Update declarations after above changes.
16699 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16701 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16702 128-bit round to integer instructions.
16703 (ceil<mode>2): Likewise.
16704 (btrunc<mode>2): Likewise.
16705 (round<mode>2): Likewise.
16707 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16709 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16710 unaligned VSX load/store on P8/P9.
16711 (expand_block_clear): Allow the use of unaligned VSX
16712 load/store on P8/P9.
16714 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16716 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16718 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16719 swap associated with both a load and a store.
16721 2018-01-02 Andrew Waterman <andrew@sifive.com>
16723 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16724 * config/riscv/riscv.md (clear_cache): Use it.
16726 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16728 * web.c: Remove out-of-date comment.
16730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16732 * expr.c (fixup_args_size_notes): Check that any existing
16733 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16734 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16735 (emit_single_push_insn): ...here.
16737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16739 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16740 (const_vector_encoded_nelts): New function.
16741 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16742 (const_vector_int_elt, const_vector_elt): Declare.
16743 * emit-rtl.c (const_vector_int_elt_1): New function.
16744 (const_vector_elt): Likewise.
16745 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16746 of CONST_VECTOR_ELT.
16748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16750 * expr.c: Include rtx-vector-builder.h.
16751 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16752 directly on the tree encoding.
16753 (const_vector_from_tree): Likewise.
16754 * optabs.c: Include rtx-vector-builder.h.
16755 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16756 sequence of "u" values.
16757 * vec-perm-indices.c: Include rtx-vector-builder.h.
16758 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16759 directly on the vec_perm_indices encoding.
16761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16763 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16764 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16765 * rtx-vector-builder.h: New file.
16766 * rtx-vector-builder.c: Likewise.
16767 * rtl.h (rtx_def::u2): Add a const_vector field.
16768 (CONST_VECTOR_NPATTERNS): New macro.
16769 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16770 (CONST_VECTOR_DUPLICATE_P): Likewise.
16771 (CONST_VECTOR_STEPPED_P): Likewise.
16772 (CONST_VECTOR_ENCODED_ELT): Likewise.
16773 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16774 (unwrap_const_vec_duplicate): Likewise.
16775 (const_vec_series_p): Check for a non-duplicated vector encoding.
16776 Say that the function only returns true for integer vectors.
16777 * emit-rtl.c: Include rtx-vector-builder.h.
16778 (gen_const_vec_duplicate_1): Delete.
16779 (gen_const_vector): Call gen_const_vec_duplicate instead of
16780 gen_const_vec_duplicate_1.
16781 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16782 (gen_const_vec_duplicate): Use rtx_vector_builder.
16783 (gen_const_vec_series): Likewise.
16784 (gen_rtx_CONST_VECTOR): Likewise.
16785 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16786 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16787 Build a new vector rather than modifying a CONST_VECTOR in-place.
16788 (handle_special_swappables): Update call accordingly.
16789 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16790 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16791 Build a new vector rather than modifying a CONST_VECTOR in-place.
16792 (handle_special_swappables): Update call accordingly.
16794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16796 * simplify-rtx.c (simplify_const_binary_operation): Use
16797 CONST_VECTOR_ELT instead of XVECEXP.
16799 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16801 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16802 the selector elements to be different from the data elements
16803 if the selector is a VECTOR_CST.
16804 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16805 ssizetype for the selector.
16807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16809 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16810 before testing each element individually.
16811 * tree-vect-generic.c (lower_vec_perm): Likewise.
16813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16815 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16816 * selftest-run-tests.c (selftest::run_tests): Call it.
16817 * vector-builder.h (vector_builder::operator ==): New function.
16818 (vector_builder::operator !=): Likewise.
16819 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16820 (vec_perm_indices::all_from_input_p): New function.
16821 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16822 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16823 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16824 instead of reading the VECTOR_CST directly. Detect whether both
16825 vector inputs are the same before constructing the vec_perm_indices,
16826 and update the number of inputs argument accordingly. Use the
16827 utility functions added above. Only construct sel2 if we need to.
16829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16831 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16832 the broadcast of the low byte.
16833 (expand_mult_highpart): Use an explicit encoding for the permutes.
16834 * optabs-query.c (can_mult_highpart_p): Likewise.
16835 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16836 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16837 (vectorizable_bswap): Likewise.
16838 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16839 explicit encoding for the power-of-2 permutes.
16840 (vect_permute_store_chain): Likewise.
16841 (vect_grouped_load_supported): Likewise.
16842 (vect_permute_load_chain): Likewise.
16844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16846 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16847 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16848 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16849 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16850 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16851 (vect_gen_perm_mask_any): Likewise.
16853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16855 * int-vector-builder.h: New file.
16856 * vec-perm-indices.h: Include int-vector-builder.h.
16857 (vec_perm_indices): Redefine as an int_vector_builder.
16858 (auto_vec_perm_indices): Delete.
16859 (vec_perm_builder): Redefine as a stand-alone class.
16860 (vec_perm_indices::vec_perm_indices): New function.
16861 (vec_perm_indices::clamp): Likewise.
16862 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16863 (vec_perm_indices::new_vector): New function.
16864 (vec_perm_indices::new_expanded_vector): Update for new
16865 vec_perm_indices class.
16866 (vec_perm_indices::rotate_inputs): New function.
16867 (vec_perm_indices::all_in_range_p): Operate directly on the
16868 encoded form, without computing elided elements.
16869 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16870 encoding. Update for new vec_perm_indices class.
16871 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16872 the given vec_perm_builder.
16873 (expand_vec_perm_var): Update vec_perm_builder constructor.
16874 (expand_mult_highpart): Use vec_perm_builder instead of
16875 auto_vec_perm_indices.
16876 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16877 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16878 or double series encoding as appropriate.
16879 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16880 vec_perm_indices instead of auto_vec_perm_indices.
16881 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16882 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16883 (vect_permute_store_chain): Likewise.
16884 (vect_grouped_load_supported): Likewise.
16885 (vect_permute_load_chain): Likewise.
16886 (vect_shift_permute_load_chain): Likewise.
16887 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16888 (vect_transform_slp_perm_load): Likewise.
16889 (vect_schedule_slp_instance): Likewise.
16890 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16891 (vectorizable_mask_load_store): Likewise.
16892 (vectorizable_bswap): Likewise.
16893 (vectorizable_store): Likewise.
16894 (vectorizable_load): Likewise.
16895 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16896 vec_perm_indices instead of auto_vec_perm_indices. Use
16897 tree_to_vec_perm_builder to read the vector from a tree.
16898 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16899 vec_perm_builder instead of a vec_perm_indices.
16900 (have_whole_vector_shift): Use vec_perm_builder and
16901 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16902 truncation to calc_vec_perm_mask_for_shift.
16903 (vect_create_epilog_for_reduction): Likewise.
16904 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16905 from auto_vec_perm_indices to vec_perm_indices.
16906 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16907 instead of changing individual elements.
16908 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16909 the vector in d.perm.
16910 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16911 from auto_vec_perm_indices to vec_perm_indices.
16912 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16913 instead of changing individual elements.
16914 (arm_vectorize_vec_perm_const): Use new_vector to install
16915 the vector in d.perm.
16916 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16917 Update vec_perm_builder constructor.
16918 (rs6000_expand_interleave): Likewise.
16919 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16920 (rs6000_expand_interleave): Likewise.
16922 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16924 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16925 to qimode could truncate the indices.
16926 * optabs.c (expand_vec_perm_var): Likewise.
16928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16930 * Makefile.in (OBJS): Add vec-perm-indices.o.
16931 * vec-perm-indices.h: New file.
16932 * vec-perm-indices.c: Likewise.
16933 * target.h (vec_perm_indices): Replace with a forward class
16935 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16936 * optabs.h: Include vec-perm-indices.h.
16937 (expand_vec_perm): Delete.
16938 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16939 (expand_vec_perm_const): Declare.
16940 * target.def (vec_perm_const_ok): Replace with...
16941 (vec_perm_const): ...this new hook.
16942 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16943 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16944 * doc/tm.texi: Regenerate.
16945 * optabs.def (vec_perm_const): Delete.
16946 * doc/md.texi (vec_perm_const): Likewise.
16947 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16948 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16949 expand_vec_perm for constant permutation vectors. Assert that
16950 the mode of variable permutation vectors is the integer equivalent
16951 of the mode that is being permuted.
16952 * optabs-query.h (selector_fits_mode_p): Declare.
16953 * optabs-query.c: Include vec-perm-indices.h.
16954 (selector_fits_mode_p): New function.
16955 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16956 is defined, instead of checking whether the vec_perm_const_optab
16957 exists. Use targetm.vectorize.vec_perm_const instead of
16958 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16959 fit in the vector mode before using a variable permute.
16960 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16961 vec_perm_indices instead of an rtx.
16962 (expand_vec_perm): Replace with...
16963 (expand_vec_perm_const): ...this new function. Take the selector
16964 as a vec_perm_indices rather than an rtx. Also take the mode of
16965 the selector. Update call to shift_amt_for_vec_perm_mask.
16966 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16967 Use vec_perm_indices::new_expanded_vector to expand the original
16968 selector into bytes. Check whether the indices fit in the vector
16969 mode before using a variable permute.
16970 (expand_vec_perm_var): Make global.
16971 (expand_mult_highpart): Use expand_vec_perm_const.
16972 * fold-const.c: Includes vec-perm-indices.h.
16973 * tree-ssa-forwprop.c: Likewise.
16974 * tree-vect-data-refs.c: Likewise.
16975 * tree-vect-generic.c: Likewise.
16976 * tree-vect-loop.c: Likewise.
16977 * tree-vect-slp.c: Likewise.
16978 * tree-vect-stmts.c: Likewise.
16979 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16981 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16982 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16983 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16984 (aarch64_vectorize_vec_perm_const): ...this new function.
16985 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16986 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16987 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16988 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16989 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16990 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16991 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16993 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16994 check for NEON modes.
16995 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16996 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16997 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16998 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17000 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17001 the old VEC_PERM_CONST conditions.
17002 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17003 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17004 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17005 (ia64_vectorize_vec_perm_const_ok): Merge into...
17006 (ia64_vectorize_vec_perm_const): ...this new function.
17007 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17008 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17009 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17010 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17011 * config/mips/mips.c (mips_expand_vec_perm_const)
17012 (mips_vectorize_vec_perm_const_ok): Merge into...
17013 (mips_vectorize_vec_perm_const): ...this new function.
17014 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17015 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17016 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17017 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17018 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17019 (rs6000_expand_vec_perm_const): Delete.
17020 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17022 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17023 (altivec_expand_vec_perm_const_le): Take each operand individually.
17024 Operate on constant selectors rather than rtxes.
17025 (altivec_expand_vec_perm_const): Likewise. Update call to
17026 altivec_expand_vec_perm_const_le.
17027 (rs6000_expand_vec_perm_const): Delete.
17028 (rs6000_vectorize_vec_perm_const_ok): Delete.
17029 (rs6000_vectorize_vec_perm_const): New function.
17030 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17031 an element count and rtx array.
17032 (rs6000_expand_extract_even): Update call accordingly.
17033 (rs6000_expand_interleave): Likewise.
17034 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17035 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17036 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17037 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17038 (rs6000_expand_vec_perm_const): Delete.
17039 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17040 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17041 (altivec_expand_vec_perm_const_le): Take each operand individually.
17042 Operate on constant selectors rather than rtxes.
17043 (altivec_expand_vec_perm_const): Likewise. Update call to
17044 altivec_expand_vec_perm_const_le.
17045 (rs6000_expand_vec_perm_const): Delete.
17046 (rs6000_vectorize_vec_perm_const_ok): Delete.
17047 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17048 reference to the SPE evmerge intructions.
17049 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17050 an element count and rtx array.
17051 (rs6000_expand_extract_even): Update call accordingly.
17052 (rs6000_expand_interleave): Likewise.
17053 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17054 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17056 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17060 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17061 vector mode and that that mode matches the mode of the data
17063 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17064 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17065 directly using expand_vec_perm_1 when forcing selectors into
17067 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17071 * optabs-query.h (can_vec_perm_p): Delete.
17072 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17073 * optabs-query.c (can_vec_perm_p): Split into...
17074 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17075 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17076 particular selector is valid.
17077 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17078 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17079 (vect_grouped_load_supported): Likewise.
17080 (vect_shift_permute_load_chain): Likewise.
17081 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17082 (vect_transform_slp_perm_load): Likewise.
17083 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17084 (vectorizable_bswap): Likewise.
17085 (vect_gen_perm_mask_checked): Likewise.
17086 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17087 implementations of variable permutation vectors into account
17088 when deciding which selector to use.
17089 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17090 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17091 with a false third argument.
17092 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17093 to test whether the constant selector is valid and can_vec_perm_var_p
17094 to test whether a variable selector is valid.
17096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17098 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17099 * optabs-query.c (can_vec_perm_p): Likewise.
17100 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17101 instead of vec_perm_indices.
17102 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17103 (vect_gen_perm_mask_checked): Likewise,
17104 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17105 (vect_gen_perm_mask_checked): Likewise,
17107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17109 * optabs-query.h (qimode_for_vec_perm): Declare.
17110 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17111 (qimode_for_vec_perm): ...this new function.
17112 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17114 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17116 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17117 does not have a conditional at the top.
17119 2018-01-02 Richard Biener <rguenther@suse.de>
17121 * ipa-inline.c (big_speedup_p): Fix expression.
17123 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17126 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17129 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17133 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17134 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17135 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17136 cond_taken_branch_cost 3->4.
17138 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17140 PR tree-optimization/83581
17141 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17142 TODO_cleanup_cfg if any changes have been made.
17144 PR middle-end/83608
17145 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17146 convert_modes if target mode has the right side, but different mode
17149 PR middle-end/83609
17150 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17151 last argument when extracting from CONCAT. If either from_real or
17152 from_imag is NULL, use expansion through memory. If result is not
17153 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17154 the parts directly to inner mode, if even that fails, use expansion
17157 PR middle-end/83623
17158 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17159 check for bswap in mode rather than HImode and use that in expand_unop
17162 Copyright (C) 2018 Free Software Foundation, Inc.
17164 Copying and distribution of this file, with or without modification,
17165 are permitted in any medium without royalty provided the copyright
17166 notice and this notice are preserved.