1 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (vect_get_internal_def): New function.
4 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6 (search_type_for_mask_1): Use it.
8 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11 redundant WIDEN_SUM_EXPR handling.
12 (vect_recog_sad_pattern): Likewise.
14 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
17 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
18 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
19 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
20 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
22 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
24 * tree-vect-stmts.c (vectorizable_call): Make sure that we
25 use the stmt_vec_info of the original bb statement for the
26 new zero assignment, even if the call is part of a pattern.
28 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
30 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
31 that the sequence is attached to the original statement rather
32 than the pattern statement.
33 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
34 PATTERN_DEF_SEQ from the original statement rather than
35 the main pattern statement.
36 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
37 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
38 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
40 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
42 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
43 definition statements before the early exit for statements that aren't
45 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
47 (vect_transform_loop): ...here. Process pattern definition
48 statements without first checking whether the main pattern
49 statement is live or relevant.
51 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
53 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
54 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
56 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
58 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
59 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
60 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
61 (expand_block_compare): Change select_block_compare_mode call.
62 (expand_strncmp_align_check): Use new functions, fix comment.
63 (emit_final_str_compare_gpr): New function.
64 (expand_strn_compare): Refactor and clean up code.
65 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
67 2018-06-19 Tony Reix <tony.reix@atos.com>
68 Damien Bergamini <damien.bergamini@atos.com>
69 David Edelsohn <dje.gcc@gmail.com>
71 * collect2.c (static_obj): New variable.
72 (static_libs): New variable.
73 (is_in_list): Uncomment declaration.
74 (main): Track AIX libraries linked statically.
75 (is_in_list): Uncomment definition.
76 (scan_prog_file): Don't add AIX shared libraries initializer
77 to constructor list if linking statically.
79 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
81 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
83 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
85 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
87 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
90 2018-06-19 Martin Liska <mliska@suse.cz>
92 * config/i386/i386.c (ix86_can_inline_p): Do not use
93 ipa_fn_summaries::get_create.
94 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
96 (devirtualization_time_bonus): Likewise.
97 (ipcp_propagate_stage): Likewise.
98 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
99 (edge_set_predicate): Likewise.
100 (evaluate_conditions_for_known_args): Likewise.
101 (evaluate_properties_for_edge): Likewise.
102 (ipa_call_summary::reset): Tranform to ...
103 (ipa_call_summary::~ipa_call_summary): ... this.
104 (ipa_fn_summary::reset): Transform to ...
105 (ipa_fn_summary::~ipa_fn_summary): ... this.
106 (ipa_fn_summary_t::remove): Rename to ...
107 (ipa_fn_summary_t::remove_callees): ... this.
108 (ipa_fn_summary_t::duplicate): Use placement new
109 instead of memory copy.
110 (ipa_call_summary_t::duplicate): Likewise.
111 (ipa_call_summary_t::remove): Remove.
112 (dump_ipa_call_summary): Change get_create to get.
113 (ipa_dump_fn_summary): Dump only when summary exists.
114 (analyze_function_body): Use symbol_summary::get instead
116 (compute_fn_summary): Likewise.
117 (estimate_edge_devirt_benefit): Likewise.
118 (estimate_edge_size_and_time): Likewise.
119 (inline_update_callee_summaries): Likewise.
120 (remap_edge_change_prob): Likewise.
121 (remap_edge_summaries): Likewise.
122 (ipa_merge_fn_summary_after_inlining): Likewise.
123 (write_ipa_call_summary): Likewise.
124 (ipa_fn_summary_write): Likewise.
125 (ipa_free_fn_summary): Likewise.
126 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
127 (struct ipa_call_summary): Likewise.
128 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
130 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
131 (estimate_size_after_inlining): Likewise.
132 (estimate_growth): Likewise.
133 (growth_likely_positive): Likewise.
134 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
135 (inline_call): Likewise.
136 * ipa-inline.c (caller_growth_limits): Likewise.
137 (can_inline_edge_p): Likewise.
138 (can_inline_edge_by_limits_p): Likewise.
139 (compute_uninlined_call_time): Likewise.
140 (compute_inlined_call_time): Likewise.
141 (want_inline_small_function_p): Likewise.
142 (edge_badness): Likewise.
143 (update_caller_keys): Likewise.
144 (update_callee_keys): Likewise.
145 (inline_small_functions): Likewise.
146 (inline_to_all_callers_1): Likewise.
147 (dump_overall_stats): Likewise.
148 (early_inline_small_functions): Likewise.
149 (early_inliner): Likewise.
150 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
151 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
152 * ipa-pure-const.c (malloc_candidate_p): Likewise.
153 * ipa-split.c (execute_split_functions): Likewise.
154 * symbol-summary.h: Likewise.
155 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
157 2018-06-19 Richard Biener <rguenther@suse.de>
159 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
160 (vectorize_loops): ... here. Fix dbgcnt handling.
161 (try_vectorize_loop): Wrap try_vectorize_loop_1.
163 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
166 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
167 ieee128 argument takes up only one (vector) register, not two (floating
170 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
172 * gimplify.c (gimplify_init_constructor): Really never clear for an
173 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
175 2018-06-19 Richard Biener <rguenther@suse.de>
177 PR tree-optimization/86179
178 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
179 after failed recognition.
181 2018-06-18 Martin Sebor <msebor@redhat.com>
184 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
185 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
186 Handle integer subtraction.
187 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
188 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
190 2018-06-18 David Malcolm <dmalcolm@redhat.com>
192 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
193 param from rtx to rtx_insn *.
194 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
196 (frv_ifcvt_modify_insn): Likwise.
197 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
198 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
199 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
200 as_a <rtx_insn *> cast to local "unprotected_region" once
201 it's been established that it's not NULL or pc_rtx.
202 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
203 param "sethi" from rtx to rtx_insn *.
204 (nds32_group_float_insns): Likewise for param "insn".
205 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
207 (vax_output_int_subtract): Likewise.
208 * config/vax/vax.c (vax_output_int_add): Likewise for param
210 (vax_output_int_subtract): Likewise.
211 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
212 (emit_pattern_after): Likewise for param "after".
213 (emit_insn_after): Likewise.
214 (emit_jump_insn_after): Likewise.
215 (emit_call_insn_after): Likewise.
216 (emit_debug_insn_after): Likewise.
217 (emit_pattern_before): Likewise for param "before".
218 (emit_insn_before): Likewise.
219 (emit_jump_insn_before): Likewise.
220 * final.c (get_insn_template): Likewise for param "insn", removing
222 * output.h (get_insn_template): Likewise for 2nd param.
223 * rtl.h (emit_insn_before): Likewise.
224 (emit_jump_insn_before): Likewise.
225 (emit_debug_insn_before_noloc): Likewise.
226 (emit_insn_after): Likewise.
227 (emit_jump_insn_after): Likewise.
228 (emit_call_insn_after): Likewise.
229 (emit_debug_insn_after): Likewise.
230 (set_insn_deleted): Likewise for param.
232 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
235 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
236 floating point modes, so that IFmode is numerically greater than
237 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
238 to declare the ordering. This prevents IFmode from being
239 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
240 machine. Include rs6000-modes.h to share the fractional values
241 between genmodes* and the rest of the compiler.
245 * config/rs6000/rs6000-modes.h: New file.
246 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
247 meaning of rs6000_long_double_size so that 126..128 selects an
248 appropriate 128-bit floating point type.
249 (rs6000_option_override_internal): Likewise.
250 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
251 (TARGET_LONG_DOUBLE_128): Change the meaning of
252 rs6000_long_double_size so that 126..128 selects an appropriate
253 128-bit floating point type.
254 (LONG_DOUBLE_TYPE_SIZE): Update comment.
255 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
256 source and destination to match the standard usage.
257 (truncifkf2): Likewise.
258 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
259 ISA 2.07 to use an explicit clobber, instead of passing in a
261 (copysign<mode>3_soft): Likewise.
263 2018-06-18 David Malcolm <dmalcolm@redhat.com>
265 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
266 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
267 (vect_slp_analyze_instance_dependence): Likewise.
268 (vect_enhance_data_refs_alignment): Likewise.
269 (vect_analyze_data_refs_alignment): Likewise.
270 (vect_slp_analyze_and_verify_instance_alignment
271 (vect_analyze_data_ref_accesses): Likewise.
272 (vect_prune_runtime_alias_test_list): Likewise.
273 (vect_analyze_data_refs): Likewise.
274 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
275 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
276 (vect_analyze_scalar_cycles_1): Likewise.
277 (vect_get_loop_niters): Likewise.
278 (vect_analyze_loop_form_1): Likewise.
279 (vect_update_vf_for_slp): Likewise.
280 (vect_analyze_loop_operations): Likewise.
281 (vect_analyze_loop): Likewise.
282 (vectorizable_induction): Likewise.
283 (vect_transform_loop): Likewise.
284 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
285 * tree-vect-slp.c (vect_analyze_slp): Likewise.
286 (vect_make_slp_decision): Likewise.
287 (vect_detect_hybrid_slp): Likewise.
288 (vect_slp_analyze_operations): Likewise.
289 (vect_slp_bb): Likewise.
290 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
291 (vectorizable_bswap): Likewise.
292 (vectorizable_call): Likewise.
293 (vectorizable_simd_clone_call): Likewise.
294 (vectorizable_conversion): Likewise.
295 (vectorizable_assignment): Likewise.
296 (vectorizable_shift): Likewise.
297 (vectorizable_operation): Likewise.
298 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
300 2018-06-18 Martin Sebor <msebor@redhat.com>
302 PR tree-optimization/81384
303 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
304 * builtins.c (expand_builtin_strnlen): New function.
305 (expand_builtin): Call it.
306 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
307 * builtins.def (BUILT_IN_STRNLEN): New.
308 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
309 Warn for bounds in excess of maximum object size.
310 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
311 single-value ranges. Handle strnlen.
312 (handle_builtin_strlen): Handle strnlen.
313 (strlen_check_and_optimize_stmt): Same.
314 * doc/extend.texi (Other Builtins): Document strnlen.
316 2018-06-18 Maya Rashish <coypu@sdf.org>
318 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
319 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
320 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
322 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
324 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
326 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
328 * tree.c (escaped_string::escape): Replace cast to char * by
329 const_cast<char *> (unescaped).
331 2018-06-18 Nick Clifton <nickc@redhat.com>
334 * tree.c (escaped_string): New class. Converts an unescaped
335 string into its escaped equivalent.
336 (warn_deprecated_use): Use the new class to convert the
337 deprecation message, if present.
338 (test_escaped_strings): New self test.
339 (test_c_tests): Add test_escaped_strings.
340 * doc/extend.texi (deprecated): Add a note that the
341 deprecation message is affected by the -fmessage-length
342 option, and that control characters will be escaped.
343 (#pragma GCC error): Document this pragma.
344 (#pragma GCC warning): Likewise.
345 * doc/invoke.texi (-fmessage-length): Document this option's
346 effect on the #warning and #error preprocessor directives and
347 the deprecated attribute.
349 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
351 * tree.c (decl_value_expr_lookup): Revert latest change.
352 (decl_value_expr_insert): Likewise.
354 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
356 * gimplify.c (nonlocal_vlas): Delete.
357 (nonlocal_vla_vars): Likewise.
358 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
360 (gimplify_body): Do not create and destroy nonlocal_vlas.
361 * tree-nested.c: Include diagnostic.h.
362 (use_pointer_in_frame): Tweak.
363 (lookup_field_for_decl): Add assertion and declare the transformation.
364 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
365 internal error when the reference is in a wrong context. Do not
366 create a debug decl by default.
367 (note_nonlocal_block_vlas): Delete.
368 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
369 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
370 create a debug decl by default.
371 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
372 call is in a wrong context.
373 (fixup_vla_decls): New function.
374 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
375 debug variables were created.
376 * tree.c (decl_value_expr_lookup): Add checking assertion.
377 (decl_value_expr_insert): Likewise.
379 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
382 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
383 * tree-scalar-evolution.c (interpret_expr): Likewise.
384 (expression_expensive_p): Likewise.
385 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
386 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
387 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
388 (ssa_defined_by_minus_one_stmt_p): New.
390 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
393 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
394 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
395 * dojump.c (do_jump): Likewise.
396 * expr.c (expand_expr_real_2): Check operand type's sign.
397 * fold-const.c (const_unop): Handle ABSU_EXPR.
398 (fold_abs_const): Likewise.
399 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
400 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
401 (strip_sign_op_1): Likesise.
402 * match.pd: Add new pattern to generate ABSU_EXPR.
403 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
404 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
405 * tree-eh.c (operation_could_trap_helper_p): Likewise.
406 * tree-inline.c (estimate_operator_cost): Likewise.
407 * tree-pretty-print.c (dump_generic_node): Likewise.
408 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
409 * tree.def (ABSU_EXPR): New.
411 2018-06-16 Jakub Jelinek <jakub@redhat.com>
414 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
415 documented as preserved for backward compatibility only.
416 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
418 PR rtl-optimization/86108
419 * bb-reorder.c (create_forwarder_block): Renamed to ...
420 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
421 jump from new landing pad to the second part.
422 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
425 2018-06-15 Jakub Jelinek <jakub@redhat.com>
428 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
429 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
430 Only call store_expr for halves if the mode is the same.
433 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
434 Fix up comment formatting.
436 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
438 * typed-splay-tree.h (typed_splay_tree::remove): New function.
439 (typed_splay_tree::closure,
440 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
441 (typed_splay_tree::typed_splay_tree,
442 typed_splay_tree::operator =): Declared private.
443 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
444 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
445 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
446 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
447 typed_splay_tree::splay_tree_splay,
448 typed_splay_tree::splay_tree_foreach_helper,
449 typed_splay_tree::splay_tree_insert,
450 typed_splay_tree::splay_tree_remove,
451 typed_splay_tree::splay_tree_lookup,
452 typed_splay_tree::splay_tree_predecessor,
453 typed_splay_tree::splay_tree_successor,
454 typed_splay_tree::splay_tree_min,
455 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
456 (typed_splay_tree::root, typed_splay_tree::comp,
457 typed_splay_tree::delete_key,
458 typed_splay_tree::delete_value): New data members.
459 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
460 typed_splay_tree::remove.
462 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
464 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
465 -mginv and -mno-ginv to the assembler.
466 * config/mips/mips.opt (-mcrc): New option.
468 * doc/invoke.text (-mcrc): Document.
471 2018-06-15 Nick Clifton <nickc@redhat.com>
474 * tree.c (escaped_string): New class. Converts an unescaped
475 string into its escaped equivalent.
476 (warn_deprecated_use): Use the new class to convert the
477 deprecation message, if present.
478 (test_escaped_strings): New self test.
479 (test_c_tests): Add test_escaped_strings.
480 * doc/extend.texi (deprecated): Add a note that the
481 deprecation message is affected by the -fmessage-length
482 option, and that control characters will be escaped.
483 (#pragma GCC error): Document this pragma.
484 (#pragma GCC warning): Likewise.
485 * doc/invoke.texi (-fmessage-length): Document this option's
486 effect on the #warning and #error preprocessor directives and
487 the deprecated attribute.
489 2018-06-15 Richard Biener <rguenther@suse.de>
491 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
492 here, also noting vector size used.
493 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
494 size used in MSG_OPTIMIZED_LOCATIONS dump.
495 (pass_slp_vectorize::execute): Adjust.
497 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
500 * config/arc/arc.c (arc_return_address_register): Fix
503 2018-06-15 Richard Biener <rguenther@suse.de>
506 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
507 leave useless conversion stripping to force_gimple_operand_gsi.
508 (gimplify_build2): Likewise.
509 (gimplify_build1): Likewise.
511 2018-06-15 Richard Biener <rguenther@suse.de>
514 * tree-cfg.c (move_stmt_op): unshare invariant addresses
515 before adjusting their block.
517 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
519 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
520 multilibs for *-*-rtems*.
521 * config/riscv/t-rtems: New file.
523 2018-06-14 Jakub Jelinek <jakub@redhat.com>
526 * match.pd ((A +- CST1) +- CST2): Punt if last resort
527 unsigned_type_for returns NULL.
530 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
531 subregs of multi-word pseudos unless the float mode has word size.
533 2018-06-14 Richard Biener <rguenther@suse.de>
536 * tree-vect-generic.c (build_word_mode_vector_type): Remove
537 duplicate and harmful type_hash_canon.
538 * tree.c (type_hash_canon): Assert we didn't find ourselves.
540 2018-06-14 Richard Biener <rguenther@suse.de>
543 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
546 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
548 * config/rtems.h (STDINT_LONG32): Define.
550 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
551 Prachi Godbole <prachi.godbole@imgtec.com>
553 * config/mips/mips-cpus.def: Define P6600.
554 * config/mips/mips-tables.opt: Regenerate.
555 * config/mips/mips.c (mips_ucbranch_type): New enum.
556 (mips_rtx_cost_data): Add support for P6600.
557 (mips_issue_rate): Likewise.
558 (mips_multipass_dfa_lookahead): Likewise.
559 (mips_avoid_hazard): Likewise.
560 (mips_reorg_process_insns): Likewise.
561 (mips_classify_branch_p6600): New function.
562 * config/mips/mips.h (TUNE_P6600): New define.
563 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
564 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
565 * config/mips/mips.md: Include p6600.md.
566 (processor): Add p6600.
567 * config/mips/p6600.md: New file.
568 * doc/invoke.texi: Add p6600 to supported architectures.
570 2018-06-13 Martin Sebor <msebor@redhat.com>
572 PR tree-optimization/86114
573 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
575 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
577 2018-06-13 Richard Biener <rguenther@suse.de>
579 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
580 Properly set vector type of the intermediate stmt.
581 * tree-vect-stmts.c (vectorizable_operation): The destination
582 var always has vectype_out type.
584 2018-06-13 Jeff Law <law@redhat.com>
586 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
587 integer 0 for argument to print_rtl_with_bb.
588 (rl78_reorg): Likewise.
590 2018-06-13 David Malcolm <dmalcolm@redhat.com>
592 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
593 from rtx to rtx_insn *.
594 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
596 (add_sched_insns_for_speculation): Likewise for local "target",
597 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
598 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
599 from rtx_insn ** to rtx_code_label **.
600 (reorg_emit_nops): Likewise.
601 (c6x_reorg): Likewise for local "call_labels".
602 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
604 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
605 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
606 the loops over LABEL_REFS.
607 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
609 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
610 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
611 (split_branches): Strengthen local "olabel" from rtx to
612 rtx_insn *, adding a safe_as_a cast.
613 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
615 (add_insn_after): Likewise for first two params.
616 (add_insn_before): Likewise.
617 (remove_insn): Likewise for param.
618 (emit_pattern_before_noloc): Likewise for second and third params.
619 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
620 (emit_call_insn_before_noloc): Likewise.
621 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
623 (emit_barrier_before): Likewise.
624 (emit_label_before): Strengthen "label" param from "rtx" to
625 "rtx_code_label *". Strengthen "before" param from "rtx" to
627 (emit_insn_after_1): Strengthen "after" param from "rtx" to
629 (emit_pattern_after_noloc): Likewise.
630 (emit_insn_after_noloc): Likewise.
631 (emit_jump_insn_after_noloc): Likewise.
632 (emit_call_insn_after_noloc): Likewise.
633 (emit_debug_insn_after_noloc): Likewise.
634 (emit_barrier_after): Likewise.
635 (emit_label_after): Likewise for both params.
636 (emit_pattern_after_setloc): Likewise for "after" param. Convert
637 "loc" param from "int" to "location_t".
638 (emit_insn_after_setloc): Likewise.
639 (emit_jump_insn_after_setloc): Likewise.
640 (emit_call_insn_after_setloc): Likewise.
641 (emit_debug_insn_after_setloc): Likewise.
642 (emit_pattern_before_setloc): Likewise for "before" param. Convert
643 "loc" param from "int" to "location_t".
644 (emit_pattern_before): Convert NULL_RTX to NULL.
645 (emit_insn_before_setloc): Convert "loc" param from "int" to
647 (emit_jump_insn_before_setloc): Likewise.
648 (emit_call_insn_before_setloc): Likewise.
649 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
650 rtx_insn *. Convert "loc" param from "int" to "location_t".
651 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
652 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
653 Convert 3rd param from "int" to "location_t".
654 (emit_barrier_before, emit_barrier_after, next_real_insn):
655 Strengthen param from rtx to rtx_insn *.
656 (emit_label_before): Strengthen 1st param from "rtx" to
657 "rtx_code_label *". Strengthen 2nd param from "rtx" to
659 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
660 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
661 Strengthen 2nd param from "rtx" to "rtx_insn *".
662 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
663 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
664 Likewise. Convert 3rd param from "int" to "location_t".
665 (emit_label_after): Strengthen 1st param from "rtx" to
667 (next_real_insn, remove_insn): Strengthen param from "rtx" to
669 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
670 from "rtx" to "rtx_insn *".
672 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
674 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
675 bodies streamed in with -Q.
676 * dumpfile.c (dump_files): Add lto-stream-out dump file.
677 * dumpfile.h (tree_dump_index): Add lto_stream_out.
678 * gimple-streamer-out.c: Include gimple-pretty-print.h
679 (output_bb): Dump stmts streamed.
680 * lto-section-out.c: Include print-tree.h
681 (lto_begin_section): Dump sections created.
682 (lto_output_decl_index): Dump decl encoded.
683 * lto-streamer-out.c: Include print-tree.h
684 (create_output_block): Dump output block created.
685 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
686 (output_function): Dump function output.
687 (output_constructor): Dump constructor streamed.
688 (write_global_stream): Output indexes encoded.
689 (produce_asm_for_decls): Dump streams encoded.
690 * lto-streamer.c (streamer_dump_file): New global var.
691 * lto-streamer.h (streamer_dump_file): Declare.
692 * passes.c (ipa_write_summaries): Initialize streamer dump.
693 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
696 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
699 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
700 offsets for register save directives. Emit a second batch of save
701 directives, if need be, when the function accesses prior frames.
703 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
705 * config/arc/fpu.md (fmasf4): Force operand to register.
708 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
710 * config/arc/arc-protos.h (arc_pad_return): Remove.
711 * config/arc/arc.c (machine_function): Remove force_short_suffix
713 (arc_print_operand): Adjust printing of '&'.
714 (arc_verify_short): Remove conditional printing of short suffix.
715 (arc_final_prescan_insn): Remove reference to size_reason.
716 (pad_return): New function.
717 (arc_reorg): Call pad_return.
718 (arc_pad_return): Remove.
719 (arc_init_machine_status): Remove reference to force_short_suffix.
720 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
721 (attr length): When attribute iscompact is true force to 2
722 regardless; in the case of maybe check if we want to force the
723 instruction to have 4 bytes length.
724 (nopv): Change it to generate 4 byte long nop as well.
725 (blockage): New pattern.
726 (simple_return): Remove call to arc_pad_return.
727 (p_return_i): Likewise.
729 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
731 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
733 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
735 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
738 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
740 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
741 for ARC700 and ARCv2.
743 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
746 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
747 operands[2] instead of operands[1].
750 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
752 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
753 case, check whether the outer register overlaps an unallocatable
754 register, not just whether it fits the required class.
756 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
758 * poly-int.h (can_div_trunc_p): Add new overload in which all values
760 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
761 (memrefs_conflict_p): Likewise.
762 (init_alias_analysis): Likewise.
763 * cfgexpand.c (expand_debug_expr): Likewise.
764 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
765 * cse.c (fold_rtx): Likewise.
766 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
767 * expr.c (emit_block_move_hints): Likewise.
768 (clear_storage_hints, push_block, emit_push_insn): Likewise.
769 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
770 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
771 (emit_group_store): Likewise.
772 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
773 to read the PRE/POST_MODIFY increment.
774 * calls.c (store_one_arg): Use strip_offset.
775 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
777 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
779 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
780 (simplify_binary_operation_1): Extend CONST_INT handling to
782 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
783 than a HOST_WIDE_INT.
784 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
786 (adjust_mems, add_stores): Update accodingly.
787 (vt_canonicalize_addr): Track polynomial offsets.
788 (emit_note_insn_var_location): Likewise.
789 (vt_add_function_parameter): Likewise.
790 (vt_initialize): Likewise.
792 2018-06-12 Jeff Law <law@redhat.com>
794 * config.gcc (alpha*-*-freebsd*): Remove.
795 * config/alpha/freebsd.h: Remove.
797 2018-06-12 David Malcolm <dmalcolm@redhat.com>
800 * spellcheck-tree.c (levenshtein_distance): Rename to...
801 (get_edit_distance): ...this, and update for underlying renaming.
802 * spellcheck-tree.h (levenshtein_distance): Rename to...
803 (get_edit_distance): ...this.
804 * spellcheck.c (levenshtein_distance): Rename to...
805 (get_edit_distance): ...this. Convert from Levenshtein distance
806 to Damerau-Levenshtein distance by supporting transpositions of
807 adjacent characters. Rename "v1" to "v_next" and "v0" to
809 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
810 (selftest::test_edit_distance_unit_test_oneway): ...this, and
811 update for underlying renaming.
812 (selftest::levenshtein_distance_unit_test): Rename to...
813 (selftest::test_get_edit_distance_unit): ...this, and update for
815 (selftest::test_find_closest_string): Add example from PR 69968
816 where transposition helps
817 (selftest::test_metric_conditions): Update for renaming.
818 (selftest::test_metric_conditions): Likewise.
819 (selftest::spellcheck_c_tests): Likewise.
820 * spellcheck.h (levenshtein_distance): Rename both overloads to...
821 (get_edit_distance): ...this.
822 (best_match::consider): Update for renaming.
824 2018-06-12 Martin Sebor <msebor@redhat.com>
826 PR tree-optimization/85259
827 * builtins.c (compute_objsize): Handle constant offsets.
828 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
829 true iff a warning has been issued.
830 * gimple.h (gimple_nonartificial_location): New function.
831 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
832 gimple_nonartificial_location and handle -Wno-system-headers.
833 (handle_builtin_stxncpy): Same.
835 2018-06-12 Martin Sebor <msebor@redhat.com>
838 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
840 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
842 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
843 BUILTIN_VEC_XST entries for pointer to double and long long.
845 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
848 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
850 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
853 2018-06-12 Martin Liska <mliska@suse.cz>
855 * doc/options.texi: Document IntegerRange.
857 2018-06-12 Martin Liska <mliska@suse.cz>
859 * config/i386/i386.opt: Make MPX-related options as Deprecated.
860 * opt-functions.awk: Handle Deprecated flag.
861 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
863 (read_cmdline_option): Report warning for a deprecated option.
864 * opts.h (struct cl_option): Add new field cl_deprecated.
865 (CL_ERR_DEPRECATED): New.
867 2018-06-12 Martin Liska <mliska@suse.cz>
869 * doc/options.texi: Document Deprecated option flag.
871 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
873 * config/arc/arc-arch.h (arc_extras): New enum.
874 (arc_cpu_t):Add field extra.
875 (arc_cpu_types): Consider the extras.
876 * config/arc/arc-cpus.def: Add extras info.
877 * config/arc/arc-opts.h (processor_type): Consider extra field.
878 * config/arc/arc.c (arc_override_options): Handle extra field.
880 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
882 * config/arc/arc-arch.h: Update ARC_OPTX macro.
883 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
885 * config/arc/arc.c (arc_init): Update pic warning.
886 (irq_range): Update irq range parsing warnings.
887 (arc_override_options): Update various warning messages.
888 (arc_handle_aux_attribute): Likewise.
890 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
892 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
894 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
896 * doc/sourcebuild.texi: Document usage of line number 0 in verify
897 compiler messages directives.
899 2018-06-12 Matthew Fortune <mfortune@gmail.com>
901 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
902 * config/mips/mips-tables.opt: Regenerate.
903 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
905 * doc/invoke.texi: Document -march=i6500.
907 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
909 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
910 (i6400_gpmul): Add cpu_unit.
911 (i6400_gpdiv): Likewise.
912 (i6400_msa_add_d): Update reservations.
913 (i6400_msa_int_add) Likewise.
914 (i6400_msa_short_logic3) Likewise.
915 (i6400_msa_short_logic2) Likewise.
916 (i6400_msa_short_logic) Likewise.
917 (i6400_msa_move) Likewise.
918 (i6400_msa_cmp) Likewise.
919 (i6400_msa_short_float2) Likewise.
920 (i6400_msa_div_d) Likewise.
921 (i6400_msa_long_logic1) Likewise.
922 (i6400_msa_long_logic2) Likewise.
923 (i6400_msa_mult) Likewise.
924 (i6400_msa_long_float2) Likewise.
925 (i6400_msa_long_float4) Likewise.
926 (i6400_msa_long_float5) Likewise.
927 (i6400_msa_long_float8) Likewise.
928 (i6400_fpu_fadd): Include frint type.
929 (i6400_fpu_store): New define_insn_reservation.
930 (i6400_fpu_load): Likewise.
931 (i6400_fpu_move): Likewise.
932 (i6400_fpu_fcmp): Likewise.
933 (i6400_fpu_fmadd): Likewise.
934 (i6400_int_mult): Include imul3nc type and update reservation.
935 (i6400_int_div): Include idiv3 type and update reservation.
936 (i6400_int_load): Update to check type not move_type.
937 (i6400_int_store): Likewise.
938 (i6400_int_prefetch): Set zero latency.
940 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
942 * gcc.c: Document new %@{...} sequence.
943 (LINK_COMMAND_SPEC): Use it for the -L switches.
944 (cpp_unique_options): Use it for the -I switches.
945 (at_file_argbuf): New global variable.
946 (in_at_file): Likewise.
947 (alloc_args): Create at_file_argbuf.
948 (clear_args): Truncate at_file_argbuf.
949 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
950 (open_at_file): New function.
951 (close_at_file): Likewise.
952 (create_at_file): Delete.
953 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
956 (validate_switches_from_spec): Deal with %@{...} sequence.
957 (validate_switches): Likewise.
958 (driver::finalize): Call clear_args.
960 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
962 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
964 2018-06-11 Martin Sebor <msebor@redhat.com>
966 * doc/invoke.texi (-Wall): List -Wc++17-compat.
967 (Wno-class-memaccess): Add @opindex.
968 (Wno-templates, Wno-multiple-inheritance): Same.
969 (Wno-virtual-inheritance, Wno-namespaces): Same.
970 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
971 (Wno-format-overflow, Wno-format-truncation): Same.
972 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
973 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
974 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
975 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
976 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
977 (Wno-misspelled-isr): Same.
979 2018-06-11 Martin Sebor <msebor@redhat.com>
981 * PR tree-optimization/86083
982 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
984 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
986 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
988 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
991 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
992 on the correct operand.
993 (*movdi_internal64): Ditto.
995 2018-06-11 Martin Liska <mliska@suse.cz>
997 PR tree-optimization/86089
998 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1000 2018-06-11 Julia Koval <julia.koval@intel.com>
1002 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1003 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1004 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1006 2018-06-11 Olivier Hainque <hainque@adacore.com>
1008 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1009 for Ada with strict dwarf2.
1011 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1014 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1017 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1019 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1021 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1023 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1026 2018-06-08 Martin Liska <mliska@suse.cz>
1028 * tree-cfg.h (debug_function): Fix argument type to match
1031 2018-06-08 Martin Liska <mliska@suse.cz>
1033 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1034 Remove usage of MPX-related (and removed) fields.
1035 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1037 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1039 * cfg.c (debug): Use TDF_NONE rather than 0.
1040 * cfghooks.c (debug): Likewise.
1041 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1042 (struct dump_option_value_info): Convert to...
1043 (struct kv_pair): ...this template type.
1044 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1046 (optinfo_verbosity_options): Likewise.
1047 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1049 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1050 than int for "optgroup_flags" param.
1051 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1053 (dump_dec): Likewise.
1054 (dump_finish): Use TDF_NONE rather than 0.
1055 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1056 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1057 than 0. Update for change to option_ptr.
1058 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1059 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1060 0. Update for changes to optinfo_verbosity_options and
1062 (opt_info_switch_p): Convert optgroup_flags from int to
1064 (dump_basic_block): Use dump_flags_t rather than int
1065 for "dump_kind" param.
1066 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1067 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1068 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1069 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1070 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1071 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1072 TDF_NONE): Convert from macros to...
1073 (enum dump_flag): ...this new enum.
1074 (dump_flags_t): Update to use enum.
1075 (operator|, operator&, operator~, operator|=, operator&=):
1076 Implement for dump_flags_t.
1077 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1078 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1079 Convert from macros to...
1080 (enum optgroup_flag): ...this new enum.
1081 (optgroup_flags_t): New typedef.
1082 (operator|, operator|=): Implement for optgroup_flags_t.
1083 (struct dump_file_info): Convert field "alt_flags" to
1084 dump_flags_t. Convert field "optgroup_flags" to
1086 (dump_basic_block): Use dump_flags_t rather than int for param.
1087 (dump_generic_expr_loc): Likewise.
1088 (dump_dec): Likewise.
1089 (dump_register): Convert param "optgroup_flags" to
1091 (opt_info_enable_passes): Likewise.
1092 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1094 * gimple-pretty-print.c (debug): Likewise.
1095 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1096 (merged_store_group::apply_stores): Likewise.
1097 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1098 * gimple.c (verify_gimple_pp): Likewise.
1099 * graphite-poly.c (print_pbb_body): Likewise.
1100 * passes.c (pass_manager::register_one_dump_file): Convert
1101 local "optgroup_flags" to optgroup_flags_t.
1102 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1104 (debug_body): Likewise.
1105 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1106 to optgroup_flags_t.
1107 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1109 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1110 (convert_mult_to_fma): Likewise.
1111 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1112 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1113 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1114 "dump_kind" to dump_flags_t.
1116 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1118 * config/rs6000/rs6000.c (min, max): Delete.
1120 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1122 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1123 -mabi=spe and -mabi=no-spe.
1125 2018-06-08 Martin Liska <mliska@suse.cz>
1127 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1128 where we expect an existing summary.
1130 2018-06-08 Martin Liska <mliska@suse.cz>
1132 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1133 * ipa-inline.h (estimate_edge_growth): Likewise.
1135 2018-06-08 Martin Liska <mliska@suse.cz>
1137 * cgraph.c (function_version_hasher::hash): Use
1138 cgraph_node::get_uid ().
1139 (function_version_hasher::equal):
1140 * cgraph.h (cgraph_node::get_uid): New method.
1141 * ipa-inline.c (update_caller_keys): Use
1142 cgraph_node::get_uid ().
1143 (update_callee_keys): Likewise.
1144 * ipa-utils.c (searchc): Likewise.
1145 (ipa_reduced_postorder): Likewise.
1146 * lto-cgraph.c (input_node): Likewise.
1147 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1148 * symbol-summary.h (symtab_insertion): Likewise.
1149 (symtab_removal): Likewise.
1150 (symtab_duplication): Likewise.
1151 * tree-pretty-print.c (dump_function_header): Likewise.
1152 * tree-sra.c (convert_callers_for_node): Likewise.
1154 2018-06-08 Martin Liska <mliska@suse.cz>
1156 * cgraph.c (symbol_table::create_edge): Always assign a new
1158 (symbol_table::free_edge): Do not recycle numbers.
1159 * cgraph.h (cgraph_edge::get): New method.
1160 * symbol-summary.h (symtab_removal): Use it.
1161 (symtab_duplication): Likewise.
1162 (call_summary::hashable_uid): Remove.
1164 2018-06-08 Martin Liska <mliska@suse.cz>
1166 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1167 (initialize_growth_caches): Remove.
1168 (free_growth_caches): Likewise.
1169 (do_estimate_edge_time): Use edge_growth_cache.
1170 (do_estimate_edge_size): Likewise.
1171 (do_estimate_edge_hints): Likewise.
1172 * ipa-inline.c (reset_edge_caches): Likewise.
1173 (recursive_inlining): Likewise.
1174 (inline_small_functions): Likewise.
1175 * ipa-inline.h (initialize_growth_caches): Remove.
1176 (estimate_edge_size): Likewise.
1177 (estimate_edge_time): Likewise.
1178 (estimate_edge_hints): Likewise.
1179 (reset_edge_growth_cache): Likewise.
1180 * symbol-summary.h (call_summary::remove): New method.
1182 2018-06-08 Martin Liska <mliska@suse.cz>
1184 * ipa-cp.c (class edge_clone_summary): New summary.
1185 (grow_edge_clone_vectors): Remove.
1186 (ipcp_edge_duplication_hook): Remove.
1187 (class edge_clone_summary_t): New call_summary class.
1188 (ipcp_edge_removal_hook): Remove.
1189 (edge_clone_summary_t::duplicate): New function.
1190 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1191 (create_specialized_node): Likewise.
1192 (ipcp_driver): Initialize edge_clone_summaries and do not
1195 2018-06-08 Martin Liska <mliska@suse.cz>
1197 * symbol-summary.h (get): New function.
1198 (call_summary::m_initialize_when_cloning): New class member.
1200 2018-06-08 Martin Liska <mliska@suse.cz>
1202 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1203 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1204 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1205 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1206 (remove_cgraph_node_from_order): Use the removed_nodes set.
1207 (do_per_function_toporder): Likwise.
1208 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1209 instead of summary_uid.
1210 (symtab_removal): Likewise.
1211 (symtab_duplication): Likewise.
1213 2018-06-08 Martin Liska <mliska@suse.cz>
1215 * ipa-cp.c (ipcp_store_bits_results): Use
1216 ipcp_transformation_sum.
1217 (ipcp_store_vr_results): Likewise.
1218 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1220 (ipcp_transformation_initialize): ... this.
1221 (ipa_set_node_agg_value_chain):
1222 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1223 (write_ipcp_transformation_info): Likewise.
1224 (read_ipcp_transformation_info): Likewise.
1225 (ipcp_update_bits): Likewise.
1226 (ipcp_update_vr): Likewise.
1227 (ipcp_transform_function): Likewise.
1228 * ipa-prop.h: Rename ipcp_transformation_summary to
1229 ipcp_transformation.
1230 (class ipcp_transformation_t): New function summary.
1231 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1232 (ipa_get_agg_replacements_for_node): Likewise.
1234 2018-06-08 Martin Liska <mliska@suse.cz>
1236 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1238 (class funct_state_summary_t): New function_summary class.
1239 (has_function_state): Remove.
1240 (get_function_state): Likewise.
1241 (set_function_state): Likewise.
1242 (add_new_function): Likewise.
1243 (funct_state_summary_t::insert): New function.
1244 (duplicate_node_data): Remove.
1245 (remove_node_data): Remove.
1246 (funct_state_summary_t::duplicate): New function.
1247 (register_hooks): Create new funct_state_summaries.
1248 (pure_const_generate_summary): Use it.
1249 (pure_const_write_summary): Likewise.
1250 (pure_const_read_summary): Likewise.
1251 (propagate_pure_const): Likewise.
1252 (propagate_nothrow): Likewise.
1253 (dump_malloc_lattice): Likewise.
1254 (propagate_malloc): Likewise.
1255 (execute): Do not register hooks, just remove summary
1257 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1260 2018-06-08 Martin Liska <mliska@suse.cz>
1262 * ipa-reference.c (remove_node_data): Remove.
1263 (duplicate_node_data): Likewise.
1264 (class ipa_ref_var_info_summary_t): New class.
1265 (class ipa_ref_opt_summary_t): Likewise.
1266 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1267 (get_reference_optimization_summary): Use
1268 ipa_ref_opt_sum_summaries.
1269 (set_reference_vars_info): Remove.
1270 (set_reference_optimization_summary): Likewise.
1271 (ipa_init): Create summaries.
1272 (init_function_info): Use function summary.
1273 (ipa_ref_opt_summary_t::duplicate): New function.
1274 (ipa_ref_opt_summary_t::remove): New function.
1275 (get_read_write_all_from_node): Fix GNU coding style.
1276 (propagate): Use function summary.
1277 (write_node_summary_p): Fix GNU coding style.
1278 (stream_out_bitmap): Likewise.
1279 (ipa_reference_read_optimization_summary): Use function summary.
1280 (ipa_reference_c_finalize): Do not release hooks.
1282 2018-06-08 Martin Liska <mliska@suse.cz>
1284 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1285 (analyze_function_body): Extract multiple calls of get_create.
1286 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1287 * ipa-inline.c (recursive_inlining): Use ::get method.
1288 * ipa-inline.h (estimate_edge_growth): Likewise.
1290 2018-06-08 Martin Liska <mliska@suse.cz>
1292 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1294 (hsa_function_summary::hsa_function_summary): Use the new enum
1296 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1297 * hsa-gen.c (hsa_get_host_function): Likewise.
1298 (get_brig_function_name): Likewise.
1299 * ipa-hsa.c (process_hsa_functions): Likewise.
1300 (ipa_hsa_write_summary): Likewise.
1301 * symbol-summary.h (symtab_duplication): Use ::get function/
1302 (get): New function.
1304 2018-06-08 Martin Liska <mliska@suse.cz>
1306 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1308 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1309 (hsa_register_kernel): Likewise.
1310 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1311 * hsa-gen.c (hsa_get_host_function): Likewise.
1312 (get_brig_function_name): Likewise.
1313 (generate_hsa): Likewise.
1314 (pass_gen_hsail::execute): Likewise.
1315 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1316 (devirtualization_time_bonus): Likewise.
1317 (ipcp_propagate_stage): Likewise.
1318 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1319 (edge_set_predicate): Likewise.
1320 (evaluate_conditions_for_known_args): Likewise.
1321 (evaluate_properties_for_edge): Likewise.
1322 (ipa_fn_summary::reset): Likewise.
1323 (ipa_fn_summary_t::duplicate): Likewise.
1324 (dump_ipa_call_summary): Likewise.
1325 (ipa_dump_fn_summary): Likewise.
1326 (analyze_function_body): Likewise.
1327 (compute_fn_summary): Likewise.
1328 (estimate_edge_devirt_benefit): Likewise.
1329 (estimate_edge_size_and_time): Likewise.
1330 (estimate_calls_size_and_time): Likewise.
1331 (estimate_node_size_and_time): Likewise.
1332 (inline_update_callee_summaries): Likewise.
1333 (remap_edge_change_prob): Likewise.
1334 (remap_edge_summaries): Likewise.
1335 (ipa_merge_fn_summary_after_inlining): Likewise.
1336 (ipa_update_overall_fn_summary): Likewise.
1337 (read_ipa_call_summary): Likewise.
1338 (inline_read_section): Likewise.
1339 (write_ipa_call_summary): Likewise.
1340 (ipa_fn_summary_write): Likewise.
1341 (ipa_free_fn_summary): Likewise.
1342 * ipa-hsa.c (process_hsa_functions): Likewise.
1343 (ipa_hsa_write_summary): Likewise.
1344 (ipa_hsa_read_section): Likewise.
1345 * ipa-icf.c (sem_function::merge): Likewise.
1346 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1347 (do_estimate_edge_time): Likewise.
1348 (estimate_size_after_inlining): Likewise.
1349 (estimate_growth): Likewise.
1350 (growth_likely_positive): Likewise.
1351 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1352 (inline_call): Likewise.
1353 * ipa-inline.c (caller_growth_limits): Likewise.
1354 (can_inline_edge_p): Likewise.
1355 (can_inline_edge_by_limits_p): Likewise.
1356 (compute_uninlined_call_time): Likewise.
1357 (compute_inlined_call_time): Likewise.
1358 (want_inline_small_function_p): Likewise.
1359 (edge_badness): Likewise.
1360 (update_caller_keys): Likewise.
1361 (update_callee_keys): Likewise.
1362 (recursive_inlining): Likewise.
1363 (inline_small_functions): Likewise.
1364 (inline_to_all_callers_1): Likewise.
1365 (dump_overall_stats): Likewise.
1366 (early_inline_small_functions): Likewise.
1367 (early_inliner): Likewise.
1368 * ipa-inline.h (estimate_edge_growth): Likewise.
1369 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1370 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1371 * ipa-prop.h (IPA_NODE_REF): Likewise.
1372 (IPA_EDGE_REF): Likewise.
1373 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1374 (propagate_malloc): Likewise.
1375 * ipa-split.c (execute_split_functions): Likewise.
1376 * symbol-summary.h: Rename get to get_create.
1378 (get_create): Likewise.
1379 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1381 2018-06-08 Martin Liska <mliska@suse.cz>
1383 * symbol-summary.h (release): Move definition out of class
1385 (symtab_removal): Likewise.
1386 (symtab_duplication): Likewise.
1388 2018-06-08 Martin Liska <mliska@suse.cz>
1390 * symbol-summary.h (function_summary): Move constructor
1391 implementation out of class declaration.
1392 (release): Likewise.
1393 (symtab_insertion): Likewise.
1394 (symtab_removal): Likewise.
1395 (symtab_duplication): Likewise.
1398 2018-06-08 Martin Liska <mliska@suse.cz>
1400 * Makefile.in: Remove support for MPX (macros, related functions,
1401 fields in cgraph_node, ...).
1402 * builtin-types.def (BT_BND): Likewise.
1403 (BT_FN_BND_CONST_PTR): Likewise.
1404 (BT_FN_CONST_PTR_BND): Likewise.
1405 (BT_FN_VOID_PTR_BND): Likewise.
1406 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1407 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1408 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1409 (expand_builtin_mempcpy_with_bounds): Likewise.
1410 (expand_builtin_memset_with_bounds): Likewise.
1411 (expand_builtin_memset_args): Likewise.
1412 (std_expand_builtin_va_start): Likewise.
1413 (expand_builtin): Likewise.
1414 (expand_builtin_with_bounds): Likewise.
1415 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1416 (DEF_LIB_BUILTIN_CHKP): Likewise.
1417 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1418 (DEF_CHKP_BUILTIN): Likewise.
1419 (BUILT_IN_MEMCPY): Likewise.
1420 (BUILT_IN_MEMMOVE): Likewise.
1421 (BUILT_IN_MEMPCPY): Likewise.
1422 (BUILT_IN_MEMSET): Likewise.
1423 (BUILT_IN_STPCPY): Likewise.
1424 (BUILT_IN_STRCAT): Likewise.
1425 (BUILT_IN_STRCHR): Likewise.
1426 (BUILT_IN_STRCPY): Likewise.
1427 (BUILT_IN_STRLEN): Likewise.
1428 (BUILT_IN_MEMCPY_CHK): Likewise.
1429 (BUILT_IN_MEMMOVE_CHK): Likewise.
1430 (BUILT_IN_MEMPCPY_CHK): Likewise.
1431 (BUILT_IN_MEMSET_CHK): Likewise.
1432 (BUILT_IN_STPCPY_CHK): Likewise.
1433 (BUILT_IN_STRCAT_CHK): Likewise.
1434 (BUILT_IN_STRCPY_CHK): Likewise.
1435 * calls.c (store_bounds): Likewise.
1436 (emit_call_1): Likewise.
1437 (special_function_p): Likewise.
1438 (maybe_warn_nonstring_arg): Likewise.
1439 (initialize_argument_information): Likewise.
1440 (finalize_must_preallocate): Likewise.
1441 (compute_argument_addresses): Likewise.
1442 (expand_call): Likewise.
1443 * cfgexpand.c (expand_call_stmt): Likewise.
1444 (expand_return): Likewise.
1445 (expand_gimple_stmt_1): Likewise.
1446 (pass_expand::execute): Likewise.
1447 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1448 (cgraph_node::remove): Likewise.
1449 (cgraph_node::dump): Likewise.
1450 (cgraph_node::verify_node): Likewise.
1451 * cgraph.h (chkp_function_instrumented_p): Likewise.
1452 (symtab_node::get_alias_target): Likewise.
1453 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1454 (cgraph_local_p): Likewise.
1455 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1456 (cgraph_edge::rebuild_references): Likewise.
1457 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1458 (walk_polymorphic_call_targets): Likewise.
1459 (cgraph_node::expand_thunk): Likewise.
1460 (symbol_table::output_weakrefs): Likewise.
1461 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1462 (ix86_handle_option): Likewise.
1463 * config/i386/constraints.md: Likewise.
1464 * config/i386/i386-builtin-types.def (BND): Likewise.
1468 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1469 (BDESC_FIRST): Likewise.
1471 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1472 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1473 * config/i386/i386.c (enum reg_class): Likewise.
1474 (ix86_target_string): Likewise.
1475 (ix86_option_override_internal): Likewise.
1476 (ix86_conditional_register_usage): Likewise.
1477 (ix86_valid_target_attribute_inner_p): Likewise.
1478 (ix86_set_indirect_branch_type): Likewise.
1479 (ix86_set_current_function): Likewise.
1480 (ix86_function_arg_regno_p): Likewise.
1481 (init_cumulative_args): Likewise.
1482 (ix86_function_arg_advance): Likewise.
1483 (ix86_function_arg): Likewise.
1484 (ix86_pass_by_reference): Likewise.
1485 (ix86_function_value_regno_p): Likewise.
1486 (ix86_function_value_1): Likewise.
1487 (ix86_function_value_bounds): Likewise.
1488 (ix86_return_in_memory): Likewise.
1489 (ix86_setup_incoming_vararg_bounds): Likewise.
1490 (ix86_va_start): Likewise.
1491 (indirect_thunk_need_prefix): Likewise.
1492 (print_reg): Likewise.
1493 (ix86_print_operand): Likewise.
1494 (ix86_expand_call): Likewise.
1495 (ix86_output_function_return): Likewise.
1496 (reg_encoded_number): Likewise.
1497 (BDESC_VERIFYS): Likewise.
1498 (ix86_init_mpx_builtins): Likewise.
1499 (ix86_init_builtins): Likewise.
1500 (ix86_emit_cmove): Likewise.
1501 (ix86_emit_move_max): Likewise.
1502 (ix86_expand_builtin): Likewise.
1503 (ix86_builtin_mpx_function): Likewise.
1504 (ix86_get_arg_address_for_bt): Likewise.
1505 (ix86_load_bounds): Likewise.
1506 (ix86_store_bounds): Likewise.
1507 (ix86_load_returned_bounds): Likewise.
1508 (ix86_store_returned_bounds): Likewise.
1509 (ix86_class_likely_spilled_p): Likewise.
1510 (ix86_hard_regno_mode_ok): Likewise.
1511 (x86_order_regs_for_local_alloc): Likewise.
1512 (ix86_mitigate_rop): Likewise.
1513 (ix86_bnd_prefixed_insn_p): Likewise.
1514 (ix86_mpx_bound_mode): Likewise.
1515 (ix86_make_bounds_constant): Likewise.
1516 (ix86_initialize_bounds): Likewise.
1517 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1518 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1519 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1520 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1521 (TARGET_CHKP_BOUND_MODE): Likewise.
1522 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1523 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1524 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1525 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1526 * config/i386/i386.h (TARGET_MPX): Likewise.
1527 (TARGET_MPX_P): Likewise.
1528 (VALID_BND_REG_MODE): Likewise.
1529 (FIRST_BND_REG): Likewise.
1530 (LAST_BND_REG): Likewise.
1531 (enum reg_class): Likewise.
1532 (BND_REG_P): Likewise.
1533 (BND_REGNO_P): Likewise.
1534 (BNDmode): Likewise.
1535 (ADJUST_INSN_LENGTH): Likewise.
1536 * config/i386/i386.md: Likewise.
1537 * config/i386/i386.opt: Likewise.
1538 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1539 (defined): Likewise.
1540 (LINK_MPX): Likewise.
1541 (MPX_SPEC): Likewise.
1542 (LIBMPX_SPEC): Likewise.
1543 (LIBMPXWRAPPERS_SPEC): Likewise.
1544 (CHKP_SPEC): Likewise.
1545 * config/i386/predicates.md: Likewise.
1546 * dbxout.c (dbxout_type): Likewise.
1547 * doc/extend.texi: Likewise.
1548 * doc/invoke.texi: Likewise.
1549 * doc/md.texi: Likewise.
1550 * doc/tm.texi: Likewise.
1551 * doc/tm.texi.in: Likewise.
1552 * dwarf2out.c (is_base_type): Likewise.
1553 (gen_formal_types_die): Likewise.
1554 (gen_subprogram_die): Likewise.
1555 (gen_type_die_with_usage): Likewise.
1556 (gen_decl_die): Likewise.
1557 (dwarf2out_late_global_decl): Likewise.
1558 * expr.c (expand_assignment): Likewise.
1559 (emit_storent_insn): Likewise.
1560 (store_expr_with_bounds): Likewise.
1561 (store_expr): Likewise.
1562 (expand_expr_real_1): Likewise.
1563 * expr.h (store_expr_with_bounds): Likewise.
1564 * function.c (use_register_for_decl): Likewise.
1565 (struct bounds_parm_data): Likewise.
1566 (assign_parms_augmented_arg_list): Likewise.
1567 (assign_parm_find_entry_rtl): Likewise.
1568 (assign_parm_is_stack_parm): Likewise.
1569 (assign_parm_load_bounds): Likewise.
1570 (assign_bounds): Likewise.
1571 (assign_parms): Likewise.
1572 (expand_function_start): Likewise.
1573 * gcc.c (CHKP_SPEC): Likewise.
1574 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1575 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1576 (wrestrict_dom_walker::check_call): Likewise.
1577 * gimple.c (gimple_build_call_from_tree): Likewise.
1578 * gimple.h (enum gf_mask): Likewise.
1579 (gimple_call_with_bounds_p): Likewise.
1580 (gimple_call_set_with_bounds): Likewise.
1581 * gimplify.c (gimplify_init_constructor): Likewise.
1582 * ipa-cp.c (initialize_node_lattices): Likewise.
1583 (propagate_constants_across_call): Likewise.
1584 (find_more_scalar_values_for_callers_subset): Likewise.
1585 * ipa-hsa.c (process_hsa_functions): Likewise.
1586 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1587 * ipa-icf.c (sem_function::merge): Likewise.
1588 * ipa-inline.c (early_inliner): Likewise.
1589 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1590 (warn_function_cold): Likewise.
1591 (propagate_pure_const): Likewise.
1592 * ipa-ref.h (enum GTY): Likewise.
1593 * ipa-split.c (find_retbnd): Likewise.
1594 (consider_split): Likewise.
1595 (split_function): Likewise.
1596 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1597 * ipa.c (walk_polymorphic_call_targets): Likewise.
1598 (symbol_table::remove_unreachable_nodes): Likewise.
1599 (process_references): Likewise.
1600 (cgraph_build_static_cdtor_1): Likewise.
1601 * lto-cgraph.c (lto_output_node): Likewise.
1602 (output_refs): Likewise.
1603 (compute_ltrans_boundary): Likewise.
1604 (input_overwrite_node): Likewise.
1605 (input_node): Likewise.
1606 (input_cgraph_1): Likewise.
1607 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1608 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1609 (class pass_chkp_instrumentation_passes): Likewise.
1610 (make_pass_chkp_instrumentation_passes): Likewise.
1611 * passes.def: Likewise.
1612 * rtl.h (struct GTY): Likewise.
1613 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1614 * stor-layout.c (layout_type): Likewise.
1615 * symtab.c: Likewise.
1616 * target.def: Likewise.
1617 * targhooks.c (default_chkp_bound_type): Likewise.
1618 (default_chkp_bound_mode): Likewise.
1619 (default_builtin_chkp_function): Likewise.
1620 (default_chkp_function_value_bounds): Likewise.
1621 (default_chkp_make_bounds_constant): Likewise.
1622 (default_chkp_initialize_bounds): Likewise.
1623 * targhooks.h (default_chkp_bound_type): Likewise.
1624 (default_chkp_bound_mode): Likewise.
1625 (default_builtin_chkp_function): Likewise.
1626 (default_chkp_function_value_bounds): Likewise.
1627 (default_chkp_make_bounds_constant): Likewise.
1628 (default_chkp_initialize_bounds): Likewise.
1629 * toplev.c (compile_file): Likewise.
1630 (process_options): Likewise.
1631 * tree-core.h (DEF_BUILTIN): Likewise.
1632 (DEF_BUILTIN_CHKP): Likewise.
1633 * tree-inline.c (declare_return_variable): Likewise.
1634 (remap_gimple_stmt): Likewise.
1635 (copy_bb): Likewise.
1636 (initialize_inlined_parameters): Likewise.
1637 (expand_call_inline): Likewise.
1638 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1639 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1640 (make_pass_ipa_chkp_produce_thunks): Likewise.
1641 (make_pass_chkp): Likewise.
1642 (make_pass_chkp_opt): Likewise.
1643 (make_pass_chkp_instrumentation_passes): Likewise.
1644 * tree-pretty-print.c (dump_generic_node): Likewise.
1645 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1646 * tree-ssa-dce.c (propagate_necessity): Likewise.
1647 (eliminate_unnecessary_stmts): Likewise.
1648 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1649 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1650 * tree-ssa-sccvn.h: Likewise.
1651 * tree-ssa-strlen.c (get_string_length): Likewise.
1652 (valid_builtin_call): Likewise.
1653 (adjust_last_stmt): Likewise.
1654 (handle_builtin_strchr): Likewise.
1655 (handle_builtin_strcpy): Likewise.
1656 (handle_builtin_stxncpy): Likewise.
1657 (handle_builtin_memcpy): Likewise.
1658 (handle_builtin_strcat): Likewise.
1659 (strlen_check_and_optimize_stmt): Likewise.
1660 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1661 * tree-streamer-in.c: Likewise.
1662 * tree-streamer.c (record_common_node): Likewise.
1663 * tree.c (tree_code_size): Likewise.
1664 (wide_int_to_tree_1): Likewise.
1665 (type_contains_placeholder_1): Likewise.
1666 (build_common_tree_nodes): Likewise.
1667 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1668 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1669 (POINTER_BOUNDS_P): Likewise.
1670 (BOUNDED_TYPE_P): Likewise.
1671 (BOUNDED_P): Likewise.
1672 (CALL_WITH_BOUNDS_P): Likewise.
1673 (pointer_bounds_type_node): Likewise.
1674 * value-prof.c (gimple_ic): Likewise.
1675 * var-tracking.c (vt_add_function_parameters): Likewise.
1676 * varasm.c (make_decl_rtl): Likewise.
1677 (assemble_start_function): Likewise.
1678 (output_constant): Likewise.
1679 (maybe_assemble_visibility): Likewise.
1680 * varpool.c (ctor_for_folding): Likewise.
1681 * chkp-builtins.def: Remove.
1682 * ipa-chkp.c: Remove.
1683 * ipa-chkp.h: Remove.
1684 * rtl-chkp.c: Remove.
1685 * rtl-chkp.h: Remove.
1686 * tree-chkp-opt.c: Remove.
1687 * tree-chkp.c: Remove.
1688 * tree-chkp.h: Remove.
1690 2018-06-07 Carl Love <cel@us.ibm.com>
1692 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1693 vextract_fp_from_shortl): Add BE support.
1695 2018-06-07 Paul Koning <ni1d@arrl.net>
1697 * compare-elim.c (try_merge_compare): Don't merge compare if
1698 address contains a side effect.
1699 (try_eliminate_compare): Likewise.
1701 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1703 * config.gcc: Support "tremont".
1704 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1705 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1707 * config/i386/i386.c (m_TREMONT): Define.
1708 (processor_target_table): Add "tremont".
1709 (PTA_TREMONT): Define.
1710 (ix86_lea_outperforms): Add TARGET_TREMONT.
1711 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1712 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1713 and M_INTEL_GOLDMONT_PLUS.
1714 (fold_builtin_cpu): Add "tremont".
1715 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1716 (ix86_option_override_internal): Add "tremont".
1717 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1718 (processor_type): Add PROCESSOR_TREMONT.
1719 * config/i386/x86-tune.def: Add m_TREMONT.
1720 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1722 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1724 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1725 symbol defined for msp430i* devices to be lower case.
1727 gcc/testsuite/gcc.target/msp430/
1728 * msp430i-device-symbol.c: New test.
1729 * msp430f-device-symbol.c: New test.
1730 * msp430.h: New test header file.
1732 2018-06-07 Richard Biener <rguenther@suse.de>
1734 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1735 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1736 Properly wrap signed arithmetic if overflow wraps.
1738 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1740 PR tree-optimization/69615
1741 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1742 of a cast from a same precision integral SSA_NAME in a bb dominated
1743 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1744 cast to utype if rhs2 has already a compatible type.
1746 2018-06-07 Richard Biener <rguenther@suse.de>
1748 PR tree-optimization/85935
1749 * graphite-scop-detection.c (find_params_in_bb): Analyze
1750 condition operands with respect to the correct loop. Assert
1751 the analysis doesn't fail.
1753 2018-06-04 Carl Love <cel@us.ibm.com>
1755 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1756 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1757 as it is slightly cheaper.
1758 (first_match_or_eos_index_<mode>):
1759 Calculate index using natural element order.
1760 (first_match_index_<mode>):
1761 Calculate index using natural element order.
1762 (first_match_or_eos_index_<mode>):
1763 Calculate index using natural order.
1764 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1765 for BE and LE modes.
1766 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1767 P9V_BUILTIN_VCLZLSBB_V16QI.
1768 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1771 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1773 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1774 indentation and line wrap for many prototypes. Add missing
1775 @smallexample directives around block of prototypes for vec_xl and
1778 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1780 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1781 track if we pass or return IEEE 128-bit floating point.
1782 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1783 C++ mangling that is compatible with GCC 8.1.
1784 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1785 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1786 floating point types.
1787 (rs6000_function_arg_advance_1): Likewise.
1788 (rs6000_mangle_type): Optionally generate mangled names that match
1789 what GCC 8.1 generated for IEEE 128-bit floating point types.
1790 (rs6000_globalize_decl_name): If we have an external function that
1791 passes or returns IEEE 128-bit types, generate a weak reference
1792 from the mangled name used in GCC 8.1 to the current mangled
1794 (rs6000_init_builtins): Make __ibm128 use the long double type if
1795 long double is IBM extended double. Make __float128 use the long
1796 double type if long double is IEEE 128-bit.
1799 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1800 macro for __ibm128 built-in functions.
1801 (PACK_IF): Add __ibm128 pack/unpack functions.
1802 (UNPACK_IF): Likewise.
1803 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1804 enable long double built-in functions if long double is IEEE
1805 128-bit floating point.
1806 (rs6000_invalid_builtin): Update long double built-in function
1808 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1809 functions, adjust the built-in function to use the long double
1810 built-in function if __ibm128 and long double are the same type.
1811 * doc/extend.texi (PowerPC builtins): Update documention for
1812 __builtin_{,un}pack_longdouble. Add documentation for
1813 __builtin_{,un}pack_ibm128.
1815 2018-06-06 Jim Wilson <jimw@sifive.com>
1817 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1818 (struct machine_function): New field interrupt_mode.
1819 (riscv_handle_type_attribute): New function. Add forward declaration.
1820 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1821 (riscv_expand_epilogue): Check interrupt_mode field.
1822 (riscv_set_current_function): Check interrupt attribute args and
1823 set interrupt_mode field.
1824 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1825 (riscv_sret, riscv_uret): New.
1826 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1827 new arguments to interrupt attribute.
1829 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1832 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1833 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1835 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1837 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1838 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1839 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1840 several redundant entries.
1842 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1844 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1845 type from "rtx" to "rtx_insn *".
1846 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1847 for local "call_insn", removing cast.
1848 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1850 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1852 PR tree-optimization/86066
1853 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1854 for BIT_INSERT_EXPR stores.
1856 2018-06-06 Richard Biener <rguenther@suse.de>
1858 PR tree-optimization/86062
1859 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1860 component refs ontop
1861 of to be offsetted base.
1863 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1865 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1866 to be static and remove check on interrupt attribute name.
1868 gcc/testsuite/gcc.target/msp430/
1869 * function-attributes-4.c: New test.
1870 * static-interrupts.c: New test.
1872 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1874 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1875 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1877 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1880 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1882 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1883 Remove second argument, change how error is called.
1884 (aarch64_layout_arg): Remove second argument from
1885 aarch64_err_no_fpadvsimd call.
1886 (aarch64_init_cumulative_args): Ditto.
1887 (aarch64_gimplify_va_arg_expr): Ditto.
1888 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1890 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1892 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1893 (*simple_return_indirect_internal<mode>): Rename from
1894 simple_return_indirect_internal. Use W mode iterator.
1895 (rstorssp): New expander.
1896 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1897 (clrssbsy): New expander.
1898 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1900 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1902 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1904 (cmse_check_pointed_object): Likewise.
1906 2018-06-05 Martin Liska <mliska@suse.cz>
1908 PR gcov-profile/47618
1909 * doc/invoke.texi: Document how -fprofile-dir format
1912 2018-06-05 Richard Biener <rguenther@suse.de>
1914 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1915 removal pretend DOM info isn't available so we do not update
1916 it and only remove edges, not dominated blocks. Actually free
1917 DOM info in case we removed something. Remove unreachable blocks.
1918 (mfb_keep_latches): Work with either DOM info or marked backedges.
1919 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1920 first. Mark backedges if DOM info isn't available.
1921 (Re-)compute DOM info after cleanup_control_flow_pre.
1923 2018-06-05 Richard Biener <rguenther@suse.de>
1925 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1926 (locus_discrim_hasher::hash): Adjust.
1927 (locus_discrim_hasher::equal): Likewise.
1928 (next_discriminator_for_locus): Work on line directly.
1929 (same_line_p): Pass in expanded locus1 as well.
1930 (assign_discriminators): Avoid redundant location expansions.
1932 2018-06-05 Richard Biener <rguenther@suse.de>
1934 PR tree-optimization/86046
1935 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1936 if required after clearing TREE_ADDRESSABLE.
1938 2018-06-05 Richard Biener <rguenther@suse.de>
1940 PR tree-optimization/86047
1941 * tree-ssa-loop.c (for_each_index): Glob handling of all
1942 decls and constants and really handle all of them.
1944 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1947 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1948 qualifier_void_pointer and qualifier_const_void_pointer.
1949 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1950 (arm_init_builtins): Handle the above.
1951 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1952 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1953 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1956 2018-06-05 Martin Liska <mliska@suse.cz>
1958 * auto-profile.c (read_autofdo_file): Do not use
1959 gcov_ctr_summary struct.
1960 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1961 * coverage.c (struct counts_entry): Likewise.
1962 (read_counts_file): Read just single summary entry.
1963 (get_coverage_counts): Use gcov_summary struct.
1964 * coverage.h (get_coverage_counts): Likewise.
1965 * gcov-dump.c (dump_working_sets): Likewise.
1966 (tag_summary): Dump just single summary.
1967 * gcov-io.c (gcov_write_summary): Write just histogram
1969 (gcov_read_summary): Read just single summary.
1970 (compute_working_sets): Use gcov_summary struct.
1971 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1972 of GCOV_COUNTERS_SUMMABLE.
1973 (GCOV_COUNTERS_SUMMABLE): Remove.
1974 (GCOV_FIRST_VALUE_COUNTER): Replace with
1975 GCOV_COUNTER_V_INTERVAL.
1976 (struct gcov_ctr_summary): Remove.
1977 (struct gcov_summary): Directly use fields of former
1979 (compute_working_sets): Use gcov_summary struct.
1980 * gcov.c (read_count_file): Do not use ctrs fields.
1981 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1983 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1985 * profile.c: Likewise.
1986 * profile.h: Likewise.
1988 2018-06-05 Martin Liska <mliska@suse.cz>
1990 PR gcov-profile/84846
1991 * gcov.c (output_lines): Print working directory only
1992 in intermediate format.
1994 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1996 * config/s390/s390-builtin-types.def: Add void function type.
1997 * config/s390/s390-builtins.def: Use the function type for the
2000 2018-06-04 Jim Wilson <jimw@sifive.com>
2002 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2004 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2005 and maybe_eh_return. Change regno to unsigned int. Use new args to
2006 handle EH_RETURN_DATA_REGNO registers properly.
2007 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2008 (riscv_expand_epilogue): Update comment. Change argument name and
2009 type. Update code to use new name and type. Pass new args to
2010 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2012 * config/riscv/riscv.md (NORMAL_RETURN): New.
2013 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2014 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2015 (eh_return): Call gen_eh_return_internal and emit barrier.
2016 (eh_return_internal): Call riscv_expand_epilogue.
2018 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2020 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2021 bit_insertion field and declare can_be_merged_into method.
2022 (merged_store_group::can_be_merged_into): New method.
2023 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2024 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2025 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2027 2018-06-04 Richard Biener <rguenther@suse.de>
2029 PR tree-optimization/85955
2030 * builtins.c (fold_builtin_sincos): Convert pointers to
2031 destination to appropriate type before dereferencing.
2033 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2035 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2037 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2039 * expr.c (expand_expr_real_1): Force the operand into memory if
2040 its TYPE_MODE is BLKmode and if there is no integer mode for
2041 the number of bits being extracted.
2043 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2047 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2048 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2050 2018-06-04 Richard Biener <rguenther@suse.de>
2052 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2053 (cleanup_tree_cfg_noloop): ... single caller. Do
2054 start_recording_case_labels later.
2056 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2058 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2059 to _IMMINTRIN_H_INCLUDED.
2060 * config/i386/pconfigintrin.h: Ditto.
2061 * config/i386/waitpkgintrin.h: Ditto.
2062 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2063 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2064 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2065 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2066 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2067 waitpkgintrin.h and cldemoteintrin.h.
2069 2018-06-04 Richard Biener <rguenther@suse.de>
2071 PR tree-optimization/86038
2072 * tracer.c (find_best_successor): Check probability for
2073 being initialized, bail out if not.
2075 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2078 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2079 of bits to ignore when comparing architectures.
2081 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2083 PR tree-optimization/69615
2084 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2085 maximum or minimum of the type, try to merge it also as if
2086 range1 is + [-, x - 1] or + [x + 1, -].
2089 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2091 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2093 PR tree-optimization/86034
2094 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2095 the unsigned bitfield type in a bit insertion sequence if it does not
2096 have a larger precision than the bitfield size.
2097 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2099 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2101 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2103 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2105 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2106 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2107 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2108 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2110 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2112 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2113 Disable -fdelete-null-pointer-checks for ELF toolchain.
2115 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2116 Kito Cheng <kito.cheng@gmail.com>
2118 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2119 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2120 (nds32*-*-*): Add glibc and uclibc conditions.
2121 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2122 (TARGET_EXCEPT_UNWIND_INFO): Define.
2123 * config/nds32/elf.h: New file.
2124 * config/nds32/linux.h: New file.
2125 * config/nds32/nds32-elf.opt: New file.
2126 * config/nds32/nds32-linux.opt: New file.
2127 * config/nds32/nds32-fp-as-gp.c
2128 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2129 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2131 (nds32_asm_file_end): Ditto.
2132 (nds32_print_operand): Ditto.
2133 (nds32_insert_attributes): Ditto.
2134 (nds32_init_libfuncs): New function.
2135 (TARGET_HAVE_TLS): Define.
2136 (TARGET_INIT_LIBFUNCS): Define.
2137 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2139 (TARGET_ELF): Apply different mcmodel setting.
2140 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2141 been migrated into elf.h and linux.h files.
2142 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2143 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2144 (mcmodel): The content has been migrated into nds32-elf.opt and
2145 nds32-linux.opt files.
2146 * config/nds32/t-elf: New file.
2147 * config/nds32/t-linux: New file.
2149 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2150 Shiva Chen <shiva0217@gmail.com>
2152 * config/nds32/constants.md (unspec_volatile_element): Add
2153 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2154 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2156 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2157 (make_pass_nds32_fp_as_gp): Declare.
2158 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2160 (nds32_asm_function_end_prologue): Remove unused asm output.
2161 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2162 (nds32_asm_file_start): Output necessary fp_as_gp information.
2163 (nds32_option_override): Adjust register usage.
2164 (nds32_expand_prologue): Consider fp_as_gp situation.
2165 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2166 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2169 (simple_return): Ditto.
2170 (omit_fp_begin): Output special directive for fp_as_gp.
2171 (omit_fp_end): Output special directive for fp_as_gp.
2172 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2173 mforbid-fp-as-gp): New options.
2175 2018-06-01 Mark Wielaard <mark@klomp.org>
2177 * dwarf2out.c (dwarf2out_finish): Remove generation of
2178 DW_AT_loclists_base.
2180 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2182 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2183 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2184 (struct merged_store_group): Add bit_insertion field.
2185 (dump_char_array): Use standard hexadecimal format.
2186 (merged_store_group::merged_store_group): Set bit_insertion to false.
2187 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2188 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2189 also print the mask in the dump file.
2190 (pass_store_merging::gate): Minor tweak.
2191 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2192 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2193 stores with INTEGER_CST stores.
2194 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2195 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2196 and use it throughout. Generate bit insertion sequences if need be.
2197 (pass_store_merging::process_store): Remove redundant condition.
2198 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2200 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2202 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2203 the 128-bit floating point types. Fix function comment.
2205 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2207 * config/aarch64/aarch64-simd.md
2208 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2210 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2213 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2215 PR tree-optimization/85989
2216 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2218 (backprop::intersect_uses): Check it when deciding whether this
2219 is a backedge reference.
2220 (backprop::process_block): Add each phi to m_visited_phis
2221 after visiting it, then clear it at the end.
2223 2018-06-01 Richard Biener <rguenther@suse.de>
2225 * tree-vectorizer.h (vect_dr_stmt): New function.
2226 (vect_get_load_cost): Adjust.
2227 (vect_get_store_cost): Likewise.
2228 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2229 Use vect_dr_stmt instead of DR_SMTT.
2230 (vect_record_base_alignments): Likewise.
2231 (vect_calculate_target_alignment): Likewise.
2232 (vect_compute_data_ref_alignment): Likewise and make static.
2233 (vect_update_misalignment_for_peel): Likewise.
2234 (vect_verify_datarefs_alignment): Likewise.
2235 (vector_alignment_reachable_p): Likewise.
2236 (vect_get_data_access_cost): Likewise. Pass down
2237 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2238 (vect_get_peeling_costs_all_drs): Likewise.
2239 (vect_peeling_hash_get_lowest_cost): Likewise.
2240 (vect_enhance_data_refs_alignment): Likewise.
2241 (vect_find_same_alignment_drs): Likewise.
2242 (vect_analyze_data_refs_alignment): Likewise.
2243 (vect_analyze_group_access_1): Likewise.
2244 (vect_analyze_group_access): Likewise.
2245 (vect_analyze_data_ref_access): Likewise.
2246 (vect_analyze_data_ref_accesses): Likewise.
2247 (vect_vfa_segment_size): Likewise.
2248 (vect_small_gap_p): Likewise.
2249 (vectorizable_with_step_bound_p): Likewise.
2250 (vect_prune_runtime_alias_test_list): Likewise.
2251 (vect_analyze_data_refs): Likewise.
2252 (vect_supportable_dr_alignment): Likewise.
2253 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2254 (vect_gen_prolog_loop_niters): Likewise.
2255 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2256 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2258 (vect_recog_mask_conversion_pattern): Likewise.
2259 (vect_try_gather_scatter_pattern): Likewise.
2260 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2261 to vect_get_store_cost.
2262 (vect_get_store_cost): Get stmt_info instead of DR.
2263 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2264 (vect_get_load_cost): Get stmt_info instead of DR.
2266 2018-06-01 Richard Biener <rguenther@suse.de>
2269 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2270 vars wrapped in MEM_REFs.
2272 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2274 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2275 Fix subreg tests so that we only return a choice between
2276 GENERAL_REGS and FP_REGS if the original classes included both.
2278 2018-06-01 Richard Biener <rguenther@suse.de>
2281 * tree-ssa-structalias.c (get_function_part_constraint):
2282 Handle NULL fi->decl.
2283 (find_func_aliases_for_call): Properly handle indirect
2284 fi from direct call.
2285 (find_func_clobbers): Likewise.
2286 (ipa_pta_execute): Likewise.
2287 (create_variable_info_for): For functions that are ifunc_resolver
2288 resolve to a varinfo that contains the result of the resolver call.
2289 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2292 2018-05-31 Michael Collison <michael.collison@arm.com>
2294 * config/aarch64/aarch64.md:
2295 (*fix_to_zero_extenddfdi2): New pattern.
2296 * gcc.target/aarch64/fix_extend1.c: New testcase.
2298 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2302 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2303 and BUILT_IN_STRNCMP_EQ.
2304 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2305 BUILT_IN_STRNCMP_EQ.
2306 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2307 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2308 (gimple_fold_builtin): Likewise.
2309 * tree-ssa-strlen.c (compute_string_length): New function.
2310 (determine_min_obsize): New function.
2311 (handle_builtin_string_cmp): New function to handle calls to
2312 string compare functions.
2313 (strlen_optimize_stmt): Add handling to builtin string compare
2315 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2316 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2317 * tree.c (build_common_builtin_nodes): Add new defines of
2318 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2320 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2323 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2324 functions with naked attribute.
2326 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2328 * config/i386/sse.md (avx_vec_concat<mode>):
2329 Substitute concat_tg_mode mode attribute with xtg_mode.
2330 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2331 (concat_tg_mode): Remove mode attribute.
2333 2018-05-31 Martin Sebor <msebor@redhat.com>
2336 * calls.c (alloc_max_size): Correct a logic error/typo.
2337 Treat excessive arguments as infinite. Warn for invalid arguments.
2338 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2340 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2343 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2344 and movx for Haswell.
2346 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2347 Cesar Philippidis <cesar@codesourcery.com>
2350 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2351 when emitting error on private/firstprivate reductions.
2352 * omp-low.c (lower_omp_target): Avoid reference-type processing
2353 on pointers for firstprivate clause.
2355 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2357 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2360 * config/aarch64/aarch64-simd.md
2361 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2362 (aarch64_ld1_x3_<mode>): Likewise
2363 (aarch64_st1x2<VALLDIF:mode>): Likewise
2364 (aarch64_st1_x2_<mode>): Likewise
2365 (aarch64_st1x3<VALLDIF:mode>): Likewise
2366 (aarch64_st1_x3_<mode>): Likewise
2367 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2368 (vld1_s8_x3): Likewise.
2369 (vld1_u16_x3): Likewise.
2370 (vld1_s16_x3): Likewise.
2371 (vld1_u32_x3): Likewise.
2372 (vld1_s32_x3): Likewise.
2373 (vld1_u64_x3): Likewise.
2374 (vld1_s64_x3): Likewise.
2375 (vld1_f16_x3): Likewise.
2376 (vld1_f32_x3): Likewise.
2377 (vld1_f64_x3): Likewise.
2378 (vld1_p8_x3): Likewise.
2379 (vld1_p16_x3): Likewise.
2380 (vld1_p64_x3): Likewise.
2381 (vld1q_u8_x3): Likewise.
2382 (vld1q_s8_x3): Likewise.
2383 (vld1q_u16_x3): Likewise.
2384 (vld1q_s16_x3): Likewise.
2385 (vld1q_u32_x3): Likewise.
2386 (vld1q_s32_x3): Likewise.
2387 (vld1q_u64_x3): Likewise.
2388 (vld1q_s64_x3): Likewise.
2389 (vld1q_f16_x3): Likewise.
2390 (vld1q_f32_x3): Likewise.
2391 (vld1q_f64_x3): Likewise.
2392 (vld1q_p8_x3): Likewise.
2393 (vld1q_p16_x3): Likewise.
2394 (vld1q_p64_x3): Likewise.
2395 (vst1_s64_x2): Likewise.
2396 (vst1_u64_x2): Likewise.
2397 (vst1_f64_x2): Likewise.
2398 (vst1_s8_x2): Likewise.
2399 (vst1_p8_x2): Likewise.
2400 (vst1_s16_x2): Likewise.
2401 (vst1_p16_x2): Likewise.
2402 (vst1_s32_x2): Likewise.
2403 (vst1_u8_x2): Likewise.
2404 (vst1_u16_x2): Likewise.
2405 (vst1_u32_x2): Likewise.
2406 (vst1_f16_x2): Likewise.
2407 (vst1_f32_x2): Likewise.
2408 (vst1_p64_x2): Likewise.
2409 (vst1q_s8_x2): Likewise.
2410 (vst1q_p8_x2): Likewise.
2411 (vst1q_s16_x2): Likewise.
2412 (vst1q_p16_x2): Likewise.
2413 (vst1q_s32_x2): Likewise.
2414 (vst1q_s64_x2): Likewise.
2415 (vst1q_u8_x2): Likewise.
2416 (vst1q_u16_x2): Likewise.
2417 (vst1q_u32_x2): Likewise.
2418 (vst1q_u64_x2): Likewise.
2419 (vst1q_f16_x2): Likewise.
2420 (vst1q_f32_x2): Likewise.
2421 (vst1q_f64_x2): Likewise.
2422 (vst1q_p64_x2): Likewise.
2423 (vst1_s64_x3): Likewise.
2424 (vst1_u64_x3): Likewise.
2425 (vst1_f64_x3): Likewise.
2426 (vst1_s8_x3): Likewise.
2427 (vst1_p8_x3): Likewise.
2428 (vst1_s16_x3): Likewise.
2429 (vst1_p16_x3): Likewise.
2430 (vst1_s32_x3): Likewise.
2431 (vst1_u8_x3): Likewise.
2432 (vst1_u16_x3): Likewise.
2433 (vst1_u32_x3): Likewise.
2434 (vst1_f16_x3): Likewise.
2435 (vst1_f32_x3): Likewise.
2436 (vst1_p64_x3): Likewise.
2437 (vst1q_s8_x3): Likewise.
2438 (vst1q_p8_x3): Likewise.
2439 (vst1q_s16_x3): Likewise.
2440 (vst1q_p16_x3): Likewise.
2441 (vst1q_s32_x3): Likewise.
2442 (vst1q_s64_x3): Likewise.
2443 (vst1q_u8_x3): Likewise.
2444 (vst1q_u16_x3): Likewise.
2445 (vst1q_u32_x3): Likewise.
2446 (vst1q_u64_x3): Likewise.
2447 (vst1q_f16_x3): Likewise.
2448 (vst1q_f32_x3): Likewise.
2449 (vst1q_f64_x3): Likewise.
2450 (vst1q_p64_x3): Likewise.
2452 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2454 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2455 user_label_prefix to name.
2457 * tree-core.h: Update comment about the format of NAME string
2458 passed to handler in attribute_spec.
2460 * config/msp430/msp430.md: Remove erroneous subreg expression from
2461 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2462 zero_extend{q,h}isi2.
2464 2018-05-30 Borislav Petkov <bp@suse.de>
2466 * doc/extend.texi: Document some architecture specific
2467 constraints and sort entries.
2469 2018-05-30 Martin Sebor <msebor@redhat.com>
2472 * builtins.c (expand_builtin_stpcpy_1): New function.
2473 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2474 only if the former succeeds.
2476 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2478 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2481 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2483 * doc/invoke.texi (-flinker-output): Document
2485 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2487 * passes.c (ipa_write_summaries): Only modify statements if body
2489 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2490 incrementally linking.
2491 (ipa_passes): Likewise.
2492 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2493 pass down resolution info.
2494 * common.opt (flag_incremental_link): Update info.
2495 * gcc.c (plugin specs): Turn flinker-output=* to
2496 -plugin-opt=-linker-output-known
2497 * toplev.c (compile_file): Also cut compilation when doing incremental
2499 * flag-types. (enum lto_partition_model): Add
2500 LTO_LINKER_OUTPUT_NOLTOREL.
2501 (invoke.texi): Add -flinker-output docs.
2502 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2503 link same way as WPA; do not stream in dead initializers.
2505 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2506 darf2out_register_external_decl): Support incremental link.
2508 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2510 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2512 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2514 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2515 it down to simple_object_copy_lto_debug_sections.
2516 (run_gcc): Determine incremental LTO link time and configure
2517 lto1 into non-wpa mode, disable renaming of debug sections.
2519 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2521 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2522 descriptions of various incorrectly documented functions.
2524 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2527 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2528 address check not strict.
2530 2018-05-30 Richard Biener <rguenther@suse.de>
2532 PR tree-optimization/85964
2533 * tracer.c (better_p): Drop initialized count check, we only
2534 call the function with initialized counts now.
2535 (find_best_successor): Do find a best edge if one
2536 has uninitialized count.
2537 (find_best_predecessor): Likewise. Do BB frequency check only
2538 if count is initialized.
2540 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2542 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2543 (aarch64_ldrstr_offset_compare): New.
2544 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2545 load/store orderings.
2546 (aarch64_gen_adjusted_ldpstp): Likewise.
2548 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2550 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2551 Check for subset of GENERAL_REGS and FP_REGS.
2552 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2555 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2557 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2558 and wi::to_poly_offset. Add the current offset and then check
2559 whether the sum fits, rather than using an unchecked addition of
2560 a checked term. Check for a shwi rather than a uhwi.
2561 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2562 (store_constructor): Use poly_int_tree_p.
2563 (expand_expr_real_1): Likewise.
2564 * function.c (assign_temp): Likewise.
2565 * fold-const.c (const_binop): Use poly_int_tree_p and
2567 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2569 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2570 to_poly_offset for MEM offsets.
2571 * ipa-icf.c (sem_variable::equals): Likewise.
2572 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2573 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2574 wi::to_poly_offset for BIT_FIELD_REF offsets.
2575 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2577 * var-tracking.c (emit_note_insn_var_location): Use
2578 tree_to_poly_uint64.
2580 2018-05-29 Jim Wilson <jimw@sifive.com>
2582 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2584 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2587 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2588 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2590 (sse4_1_round<mode>2): Use nonimmediate_operand
2591 for operand 1 predicate.
2593 2018-05-29 Martin Sebor <msebor@redhat.com>
2594 Richard Biener <rguenther@suse.de>
2597 * calls.c (get_size_range): Call determine_value_range instead
2598 of get_value_range..
2599 * tree-vrp.h (determine_value_range): Declared new function.
2600 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2602 2018-05-29 Richard Biener <rguenther@suse.de>
2604 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2605 sure to use non-pattern stmts for get_earlier_stmt arguments.
2606 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2607 called on pattern stmts.
2608 (get_later_stmt): Likewise.
2610 2018-05-29 Martin Liska <mliska@suse.cz>
2612 PR gcov-profile/85759
2613 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2616 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2618 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2620 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2624 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2625 VEC_PACK_FLOAT_EXPR): New tree codes.
2626 * tree-pretty-print.c (op_code_prio): Handle
2627 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2628 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2629 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2630 * tree-inline.c (estimate_operator_cost): Likewise.
2631 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2632 * fold-const.c (const_binop): Likewise.
2633 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2634 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2635 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2636 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2637 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2638 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2639 * expr.c (expand_expr_real_2): Likewise.
2640 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2641 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2642 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2644 * optabs.c (expand_widen_pattern_expr): For
2645 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2646 sign from result type rather than operand's type.
2647 (expand_binop_directly): For vec_packu_float_optab and
2648 vec_packs_float_optab allow result type to be different from operand's
2650 * optabs-tree.c (optab_for_tree_code): Handle
2651 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2652 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2653 * tree-vect-generic.c (expand_vector_operations_1): Handle
2654 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2655 VEC_PACK_FLOAT_EXPR.
2656 * tree-vect-stmts.c (supportable_widening_operation): Handle
2658 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2659 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2660 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2661 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2662 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2664 (vec_pack<floatprefix>_float_<mode>): New expander.
2665 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2667 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2668 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2669 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2670 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2671 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2673 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2674 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2675 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2676 VEC_PACK_FLOAT_EXPR): Document.
2678 2018-05-29 Richard Biener <rguenther@suse.de>
2680 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2682 (stmt_vec_info_vec): Make pointer.
2683 (init_stmt_vec_info_vec): Remove.
2684 (free_stmt_vec_info_vec): Likewise.
2685 (set_stmt_vec_info_vec): New function.
2686 (free_stmt_vec_infos): Likewise.
2687 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2688 (set_vinfo_for_stmt): Likewise.
2689 (get_earlier_stmt): Likewise.
2690 (get_later_stmt): Likewise.
2691 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2692 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2693 (vec_info::~vec_info): Free stmt_vec_infos.
2694 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2695 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2696 (pass_slp_vectorize::execute): Likewise.
2697 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2698 (free_stmt_vec_info_vec): Likewise.
2699 (set_stmt_vec_info_vec): New function.
2700 (free_stmt_vec_infos): Likewise.
2701 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2702 the global stmt_vec_info_vec.
2703 * tree-parloops.c (gather_scalar_reductions): Use
2704 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2707 2018-05-29 Richard Biener <rguenther@suse.de>
2709 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2711 2018-05-29 Martin Liska <mliska@suse.cz>
2712 David Malcolm <dmalcolm@redhat.com>
2714 * vec.c (test_reverse): New.
2715 (vec_c_tests): Add new test.
2716 * vec.h (vl_ptr>::reverse): New function.
2718 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2720 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2722 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2725 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2727 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2729 2018-05-28 Richard Biener <rguenther@suse.de>
2731 PR tree-optimization/85933
2732 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2733 look at stmts marked as vectorizable.
2735 2018-05-28 Richard Biener <rguenther@suse.de>
2737 PR tree-optimization/85934
2738 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2739 vector boolean check before scalar optimization.
2741 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2743 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2746 2018-05-28 Mark Wielaard <mark@klomp.org>
2748 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2749 if it is an expression containing a minus sign.
2751 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2753 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2755 2018-05-27 Paul Koning <ni1d@arrl.net>
2757 * config/pdp11/pdp11.md (truncsihi2): Remove.
2759 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2760 Chung-Ju Wu <jasonwucj@gmail.com>
2762 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2764 (unaligned_store_dw): Ditto.
2765 * config/nds32/nds32-memory-manipulation.c
2766 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2767 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2768 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2769 (emit_setmem_word_loop): Rename to ...
2770 (emit_setmem_doubleword_loop): ... this.
2771 (nds32_gen_dup_4_byte_to_word_value): New function.
2772 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2773 (nds32_expand_setmem_loop): Refine implementation.
2774 (nds32_expand_setmem_loop_v3m): Ditto.
2775 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2778 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2780 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2782 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2784 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2785 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2786 machine->attr_no_prologue_p.
2787 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2788 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2789 (nds32_expand_epilogue): Consider attr_naked_p.
2790 (nds32_expand_epilogue_v3pop): Likewise.
2791 (nds32_can_use_return_insn): Likewise.
2792 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2793 attr_no_prologue_p fields.
2794 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2796 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2799 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2801 * config/i386/sse.md
2802 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2804 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2806 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2808 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2810 (*<floatsuffix>floatv2div2sf2): Rename to ...
2811 (*float<floatunssuffix>v2div2sf2): ... this.
2812 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2813 (float<floatunssuffix>v2div2sf2_mask): ... this.
2814 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2815 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2816 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2818 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2820 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2822 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2824 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2826 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2828 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2829 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2830 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2831 gen_ufix_truncv8dfv8si2.
2832 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2833 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2834 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2835 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2836 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2837 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2838 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2839 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2841 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2845 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2847 2018-05-25 Jim Wilson <jimw@sifive.com>
2849 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2850 * config/riscv/riscv.c (struct machine_function): Add
2851 interrupt_handler_p and attribute_checked_p fields.
2852 (riscv_attribute_table): Add interrupt.
2853 (riscv_interrupt_type_p): New.
2854 (riscv_save_reg_p): Save extra regs for interrupt handler.
2855 (riscv_use_save_libcall): Return false for interrupt handler.
2856 (riscv_first_stack_step): Add forward declaration.
2857 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2858 for interrupt handler with large frame. Use it for saved reg list.
2859 (riscv_expand_prologue): Move flag_stack_usage_info support to
2860 eliminate duplication.
2861 (riscv_expand_epilogue): Generate mret for interrupt handler.
2862 (riscv_epilogue_uses): New.
2863 (riscv_can_use_return_insn): Return false for interrupt handler.
2864 (riscv_function_ok_for_sibcall): Likewise.
2865 (riscv_set_current_function): Add interrupt handler support.
2866 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2867 * config/riscv/riscv.md (UNSPECV_MRET): New.
2869 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2871 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2873 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2875 PR tree-optimization/85712
2876 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2877 this candidate has already been replaced in-situ by a copy.
2879 2018-05-25 Jason Merrill <jason@redhat.com>
2881 PR c++/80485 - inline function non-zero address.
2882 * symtab.c (nonzero_address): Check DECL_COMDAT.
2884 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2887 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2888 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2889 extension of SImode operation. Use const123_operand predicate.
2890 (*saddsi_1): Remove.
2891 (*saddl_se_1): Ditto.
2893 (*ssubl_se_1): Ditto.
2894 * config/alpha/predicates.md (const123_operand): New predicate.
2895 * config/alpha/constraints.md (P): Use IN_RANGE.
2897 2018-05-25 Richard Biener <rguenther@suse.de>
2899 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2901 (ref_maybe_used_by_stmt_p): Likewise.
2902 (stmt_may_clobber_ref_p): Likewise.
2903 (stmt_may_clobber_ref_p_1): Likewise.
2904 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2906 (ref_maybe_used_by_stmt_p): Likewise.
2907 (stmt_may_clobber_ref_p): Likewise.
2908 (stmt_may_clobber_ref_p_1): Likewise.
2909 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2910 the alias oracle to disambiguate DRs with stmts DR analysis
2912 (vect_analyze_data_refs): Do not give up on not analyzable
2913 DRs for BB vectorization. Remove code truncating the dataref
2916 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2919 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2920 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2921 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2923 2018-05-25 Richard Biener <rguenther@suse.de>
2925 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2926 function, combining stmt data ref gathering and fatal analysis
2928 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2929 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2930 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2931 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2932 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2934 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2936 PR tree-optimization/85720
2937 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2938 SCC if all partitions are builtins.
2939 (version_loop_by_alias_check): New parameter. Generate cancelable
2940 runtime alias check if all partitions are builtins.
2941 (distribute_loop): Update call to above function.
2943 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2945 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2946 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2947 (parm_default_def_partition_arg): Ditto.
2948 (set_parm_default_def_partition): Ditto.
2949 (get_parm_default_def_partitions): Ditto and make it static.
2950 (get_undefined_value_partitions): Ditto and make it static.
2951 (remove_ssa_form): Refactor call to init_var_map here.
2952 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2953 computation for loop region.
2954 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2955 (register_default_def): Delete.
2956 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2957 (parm_default_def_partition_arg): Ditto.
2958 (set_parm_default_def_partition): Ditto.
2959 (get_parm_default_def_partitions): Ditto and make it static.
2960 (get_undefined_value_partitions): Ditto and make it static.
2961 (coalesce_with_default, coalesce_with_default): Update comment.
2962 (create_coalesce_list_for_region): New func factored out from
2963 create_outofssa_var_map.
2964 (populate_coalesce_list_for_outofssa): New func factored out from
2965 create_outofssa_var_map and coalesce_ssa_name.
2966 (create_outofssa_var_map): Delete.
2967 (coalesce_ssa_name): Refactor to support live range computation.
2968 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2969 (get_parm_default_def_partitions): Delete.
2970 (get_undefined_value_partitions): Ditto.
2971 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2972 computation for loop region.
2973 (new_tree_live_info, loe_visit_block): Ditto.
2974 (live_worklist, set_var_live_on_entry): Ditto.
2975 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2976 * tree-ssa-live.h (struct _var_map): New fields.
2977 (init_var_map): Change decl.
2978 (region_contains_p): New.
2980 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2982 * tree-ssa-live.h (live_merge_and_clear): Delete.
2984 2018-05-25 Richard Biener <rguenther@suse.de>
2987 * tree-dump.c (dequeue_and_dump): Remove access to removed
2988 operand 2 of a SWITCH_EXPR.
2990 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2992 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2993 multiplication and division.
2994 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2995 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2996 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2997 (cond_udiv_optab, cond_umod_optab): New optabs.
2998 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2999 (IFN_COND_RDIV): New internal functions.
3000 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3001 TRUNC_MOD_EXPR and RDIV_EXPR.
3002 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3003 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3005 (SVE_INT_BINARY): Include mult.
3006 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3007 (optab, sve_int_op): Handle mult.
3008 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3010 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3011 for SVE_INT_BINARY_SD.
3013 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3015 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3016 (optab, sve_int_op): Handle div and udiv.
3017 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3018 for SVE_INT_BINARY_SD.
3019 (*<optab><mode>3): New insn for the same.
3021 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3023 * tree-vect-patterns.c: Include predict.h.
3024 (vect_recog_divmod_pattern): Restrict check for division support
3025 to when optimizing for size.
3027 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3029 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3030 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3031 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3032 (gimple_match_op::set_op): Likewise.
3033 (gimple_resimplify4): Declare.
3034 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3035 (expr::gen_transform): Likewise.
3036 (decision_tree::gen): Generate a simplification routine for 4 operands.
3037 * gimple-match-head.c (gimple_simplify): Add an overload for
3038 4 operands. In the top-level function, handle up to 4 call
3039 arguments and call gimple_resimplify4.
3040 (gimple_resimplify4): New function.
3041 (build_call_internal): Pass a fourth operand.
3042 (maybe_push_to_seq): Likewise.
3043 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3044 Fold VEC_COND_EXPRs of an operation and a default value into
3045 an IFN_COND_* function if possible.
3046 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3048 (SVE_COND_FP_BINARY): Include them.
3049 (optab, sve_fp_op): Handle them.
3050 (SVE_INT_BINARY_REV): New code iterator.
3051 (SVE_COND_FP_BINARY_REV): New int iterator.
3052 (commutative): New int attribute.
3053 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3055 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3057 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3058 (*cond_<optab><mode>): New patterns for reversed operands.
3060 2018-05-25 Richard Biener <rguenther@suse.de>
3062 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3063 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3064 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3065 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3066 * tree-vect-data-refs.c (everywhere): Adjust users.
3067 * tree-vect-loop.c (everywhere): Likewise.
3068 * tree-vect-slp.c (everywhere): Likewise.
3069 * tree-vect-stmts.c (everywhere): Likewise.
3070 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3072 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3074 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3076 (gcc_cv_as_section_exclude): ... this.
3077 Try Solaris as #exclude syntax.
3078 * configure: Regenerate.
3079 * config.in: Regenerate.
3080 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3082 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3083 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3085 * varasm.c (default_elf_asm_named_section): Don't check if
3086 HAVE_GAS_SECTION_EXCLUDE is defined.
3088 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3090 * doc/md.texi: Update the documentation of the cond_* optabs
3091 to mention the new final operand. Fix GET_MODE_NUNITS call.
3092 Describe the scalar case too.
3093 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3094 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3096 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3097 (get_conditional_internal_fn): Update comment.
3098 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3099 accumulator value as a final argument to conditional functions.
3100 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3101 a define_expand and add an "else" operand. Assert for now that
3102 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3103 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3104 (*cond_<optab><mode>): New patterns.
3105 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3106 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3107 (UNSPEC_COND_EOR): Delete.
3108 (optab): Remove associated mappings.
3109 (SVE_INT_BINARY): New code iterator.
3110 (sve_int_op): Remove int attribute and add "minus" to the code
3112 (SVE_COND_INT_OP): Delete.
3113 (SVE_COND_FP_OP): Rename to...
3114 (SVE_COND_FP_BINARY): ...this.
3116 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3118 * optabs.c (can_reuse_operands_p): New function.
3119 (maybe_legitimize_operands): Try to reuse the results for
3122 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3124 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3125 Add {q} suffix to insn mnemonic.
3127 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3129 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3130 (msp430_warn_func_return): New.
3132 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3134 * fold-const.c (tree_nonzero_bits): New function.
3135 * fold-const.h (tree_nonzero_bits): Likewise.
3136 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3137 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3139 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3143 * varasm.c (assemble_alias): Check ifunc_resolver only on
3146 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3149 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3150 when memory input operand is handled.
3152 2018-05-24 Luis Machado <luis.machado@linaro.org>
3154 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3156 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3158 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3160 * match.pd: Delay FMA folds until after vectorization.
3162 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3165 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3166 address check not strict.
3168 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3170 * gimple-match.h (gimple_match_op): New class.
3171 (mprts_hook): Replace parameters with a gimple_match_op *.
3172 (maybe_build_generic_op): Likewise.
3173 (gimple_simplified_result_is_gimple_val): Replace parameters with
3174 a const gimple_match_op *.
3175 (gimple_simplify): Replace code_helper * and tree * parameters with
3176 a gimple_match_op * parameter.
3177 (gimple_resimplify1): Replace code_helper *, tree and tree *
3178 parameters with a gimple_match_op * parameter.
3179 (gimple_resimplify2): Likewise.
3180 (gimple_resimplify3): Likewise.
3181 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3182 parameters with a gimple_match_op * parameter.
3183 * gimple-match-head.c (gimple_simplify): Change prototypes of
3184 auto-generated functions to take a gimple_match_op * instead of
3185 separate code_helper * and tree * parameters. Make the same
3186 change in the top-level overload and update calls to the
3187 gimple_resimplify routines. Update calls to the auto-generated
3188 functions and to maybe_push_res_to_seq in the publicly-facing
3189 operation-specific gimple_simplify overloads.
3190 (gimple_match_op::MAX_NUM_OPS): Define.
3191 (gimple_resimplify1): Replace rcode and ops with a single res_op
3192 parameter. Update call to gimple_simplify.
3193 (gimple_resimplify2): Likewise.
3194 (gimple_resimplify3): Likewise.
3195 (mprts_hook): Replace parameters with a gimple_match_op *.
3196 (maybe_build_generic_op): Likewise.
3197 (build_call_internal): Replace type, nargs and ops with
3198 a gimple_match_op *.
3199 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3200 with a single gimple_match_op *. Update calls to mprts_hook,
3201 build_call_internal and gimple_simplified_result_is_gimple_val.
3202 Factor out code that is common to the tree_code and combined_fn cases.
3203 * genmatch.c (expr::gen_transform): Replace tem_code and
3204 tem_ops with a gimple_match_op called tem_op. Update calls
3205 to the gimple_resimplify functions and maybe_push_res_to_seq.
3206 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3207 res_ops. Update call to the gimple_resimplify functions.
3208 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3209 (decision_tree::gen): Make the functions take a gimple_match_op *
3210 called res_op instead of separate res_code and res_ops parameters.
3211 Update call accordingly.
3212 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3213 and ops with a single res_op parameter. Update calls to
3214 maybe_build_generic_op and maybe_push_res_to_seq.
3215 (fold_stmt_1): Update calls to gimple_simplify and
3216 replace_stmt_with_simplification.
3217 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3218 and gimple_simplified_result_is_gimple_val.
3219 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3221 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3222 with a gimple_match_op *.
3223 (vn_nary_build_or_lookup): Likewise. Update call to
3224 vn_nary_build_or_lookup_1.
3225 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3226 gimple_match_op *. Update calls to the gimple_resimplify routines
3227 and to gimple_simplified_result_is_gimple_val.
3228 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3229 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3230 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3231 (visit_nary_op): Likewise.
3232 (visit_reference_op_load): Likewise.
3234 2018-05-23 Luis Machado <luis.machado@linaro.org>
3236 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3237 modifier for printing the step amount.
3239 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3242 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3245 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3247 * doc/sourcebuild.texi (Endianness): New subsubsection.
3249 2018-05-23 Luis Machado <luis.machado@linaro.org>
3251 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3252 <prefetch_dynamic_strides>: New const bool field.
3253 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3254 prefetch_dynamic_strides.
3255 (exynosm1_prefetch_tune): Likewise.
3256 (thunderxt88_prefetch_tune): Likewise.
3257 (thunderx_prefetch_tune): Likewise.
3258 (thunderx2t99_prefetch_tune): Likewise.
3259 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3261 (aarch64_override_options_internal): Update to set
3262 PARAM_PREFETCH_DYNAMIC_STRIDES.
3263 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3264 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3265 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3266 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3267 prefetch-dynamic-strides setting.
3269 2018-05-23 Luis Machado <luis.machado@linaro.org>
3271 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3272 <minimum_stride>: New const int field.
3273 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3274 minimum_stride field defaulting to -1.
3275 (exynosm1_prefetch_tune): Likewise.
3276 (thunderxt88_prefetch_tune): Likewise.
3277 (thunderx_prefetch_tune): Likewise.
3278 (thunderx2t99_prefetch_tune): Likewise.
3279 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3280 <default_opt_level>: Set to 3.
3281 (aarch64_override_options_internal): Update to set
3282 PARAM_PREFETCH_MINIMUM_STRIDE.
3283 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3284 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3285 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3286 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3287 stride is constant and is below the minimum stride threshold.
3289 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3291 * config/arm/arm-cpus.in (mode26): Delete.
3292 (armv4): Delete mode26 reference.
3293 * config/arm/arm.c (arm_configure_build_target): Delete use of
3296 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3298 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3300 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3301 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3303 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3304 (floatunsdidf2): Ditto.
3306 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3308 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3309 (fixuns_trunc<mode>si2_avx512f): Ditto.
3310 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3311 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3312 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3314 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3316 PR rtl-optimization/79985
3317 * df-scan.c (df_insn_refs_collect): Remove special case for
3318 global registers and asm statements.
3320 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3322 * extend.texi (Global Register Variables): Rewrite the bullet list.
3323 Note that the register is available for allocation. Note that access
3324 via inline asm must use constraints. Add note about async-signal
3325 handlers. Remove paragraph about automagic register selection.
3327 2018-05-23 Richard Biener <rguenther@suse.de>
3329 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3330 of fixed offset from memset VN.
3332 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3334 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3336 (alloc_cand_and_find_basis): Initialize first_interp field.
3337 (slsr_process_mul): Modify first_interp field.
3338 (slsr_process_add): Likewise.
3339 (slsr_process_cast): Modify first_interp field for each new
3341 (slsr_process_copy): Likewise.
3342 (dump_candidate): Dump first_interp field.
3343 (replace_mult_candidate): Process all interpretations, not just
3345 (replace_rhs_if_not_dup): Likewise.
3346 (replace_one_candidate): Likewise.
3348 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3350 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3352 (aarch64_needs_frame_chain): New function.
3353 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3355 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3358 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3359 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3360 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3361 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3362 as true for strict-align.
3363 (aarch64_can_inline_p): Perform checks even when callee has no
3364 attributes to check for strict alignment.
3365 * doc/extend.texi (AArch64 Function Attributes): Document
3367 * doc/invoke.texi: (AArch64 Options): Likewise.
3369 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3371 PR tree-optimization/85853
3372 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3373 the handling of the root of the node to...
3374 (vect_slp_analyze_node_operations_1): ...this new function,
3375 and run the whole thing with the child nodes' def types
3376 set according to their SLP node's def type.
3378 2018-05-23 Richard Biener <rguenther@suse.de>
3381 * tree-data-ref.c (create_runtime_alias_checks): Defer
3382 and ignore overflow warnings.
3384 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3386 PR tree-optimization/85822
3387 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3390 2018-05-23 Richard Biener <rguenther@suse.de>
3392 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3393 memset constants via native_interpret_expr.
3395 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3398 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3400 (cgraph_node::create_alias): Likewise.
3401 (cgraph_node::get_availability): Check ifunc_resolver instead
3402 of looking up ifunc attribute.
3403 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3404 * varasm.c (do_assemble_alias): Likewise.
3405 (assemble_alias): Likewise.
3406 (default_binds_local_p_3): Likewise.
3407 * cgraph.h (cgraph_node): Add ifunc_resolver.
3408 (cgraph_node::only_called_directly_or_aliased_p): Return false
3410 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3412 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3413 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3414 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3415 instead of looking up ifunc attribute.
3417 2018-05-22 Luis Machado <luis.machado@linaro.org>
3419 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3421 2018-05-22 Martin Sebor <msebor@redhat.com>
3424 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3425 only when expasion succeeds.
3426 (expand_builtin_strcmp): Same.
3427 (expand_builtin_strncmp): Same.
3429 2018-05-22 Martin Sebor <msebor@redhat.com>
3431 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3433 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3434 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3436 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3437 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3439 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3440 Add check for legitimate_address.
3441 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3442 (aarch64_swap_ldrstr_operands): New.
3443 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3446 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3447 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3449 * config/aarch64/aarch64.md: New patterns to generate stp
3451 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3452 single words and double words.
3453 (load_pair_sw, load_pair_dw): Likewise.
3454 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3456 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3458 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3459 for different mode ldpstp and add peephole for merged zero stores.
3461 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3463 (aarch64_gen_store_pair): Rename calls to match new patterns.
3464 (aarch64_gen_load_pair): Rename calls to match new patterns.
3465 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3466 (load_pair<DREG:mode><DREG2:mode>): ... This.
3467 (store_pair<mode>): Rename to...
3468 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3469 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3471 (V_INT_EQUIV): Handle SImode.
3472 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3475 2018-05-22 Martin Sebor <msebor@redhat.com>
3478 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3479 or ajust the presumed bound on an operation to avoid unnecessary
3482 2018-05-22 Martin Sebor <msebor@redhat.com>
3484 PR tree-optimization/85826
3485 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3486 assuming that a DECL necesarily has a constant size.
3488 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3491 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3493 2018-05-22 Richard Biener <rguenther@suse.de>
3495 PR tree-optimization/85834
3496 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3497 non-constant and non-zero memset arguments.
3499 2018-05-22 Martin Liska <mliska@suse.cz>
3502 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3504 2018-05-22 Richard Biener <rguenther@suse.de>
3506 PR tree-optimization/85863
3507 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3508 comparisons when vectype is specified.
3509 (vectorizable_condition): Do not specify vectype for
3510 vect_is_simple_cond when SLP vectorizing.
3512 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3515 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3516 define __ibm128 as long double.
3517 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3518 as a distinct type when IEEE 128-bit support is enabled.
3519 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3520 128-bit types to use the correct functions.
3521 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3522 convert between 128-bit floating point types that have different
3523 modes but the same representation, instead of using gen_lowpart to
3525 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3527 (IFKF_reg): New attributes to give the register constraints for
3529 (extend<mode>tf2_internal): New insns to mark an explicit
3530 conversion between 128-bit floating point types that have a
3531 different mode but share the same representation.
3533 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3535 PR tree-optimization/85814
3536 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3537 a null return from get_strinfo when unsharing the next
3538 strinfo in the chain.
3540 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3543 * varasm.c (weak_finish): Clean up weak_decls.
3545 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3547 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3548 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3549 UNSPEC_UADALP values.
3550 * config/aarch64/iterators.md (ABAL): New int iterator.
3553 (sur): Add mappings for the above.
3554 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3556 (aarch64_<sur>abal<mode>_4): Likewise.
3557 (aarch64_<sur>adalp<mode>_3): Likewise.
3558 (<sur>sadv16qi): New define_expand.
3560 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3562 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3563 (*movdf_internal): Ditto.
3564 (*rcpsf2_sse): Ditto.
3565 (*rsqrtsf2_sse): Ditto.
3566 (*sqrt<mode>2_sse): Ditto.
3568 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3570 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3572 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3573 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3574 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3575 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3577 * config/aarch64/arm_neon.h: Likewise.
3578 * config/aarch64/iterators.md (VQ_I): New.
3580 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3582 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3584 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3587 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3589 * config/nds32/constraints.md (S): New constraint.
3590 * config/nds32/nds32.md (call_internal): Use constraint S.
3591 (call_value_internal): Likewise.
3592 (sibcall_internal): Likewise.
3593 (sibcall_value_internal): Likewise.
3595 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3596 Chung-Ju Wu <jasonwucj@gmail.com>
3598 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3601 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3602 Chung-Ju Wu <jasonwucj@gmail.com>
3604 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3605 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3606 (nds32_rtx_costs_impl): Simplify.
3607 (nds32_address_cost_impl): Simplify.
3608 (nds32_init_rtx_costs): New function.
3609 (nds32_rtx_costs_speed_prefer): Likewise.
3610 (nds32_rtx_costs_size_prefer): Likewise.
3611 (nds32_address_cost_speed_prefer): Likewise.
3612 (nds32_address_cost_speed_fwprop): Likewise.
3613 (nds32_address_cost_size_prefer): Likewise.
3614 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3615 * config/nds32/nds32.c (nds32_option_override): Use
3616 nds32_init_rtx_costs function.
3618 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3620 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3621 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3622 (TARGET_PIPELINE_N8): Likewise.
3623 (TARGET_PIPELINE_N10): Likewise.
3624 (TARGET_PIPELINE_N13): Likewise.
3625 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3627 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3629 * config/nds32/nds32-fpu.md: Update copyright year.
3631 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3635 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3637 * config/nds32/nds32.c
3638 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3639 * config/nds32/nds32.opt (minline-asm-r15): New option.
3641 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3643 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3645 * config/nds32/nds32.md (abssi2): New pattern.
3647 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3649 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3650 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3651 Merge insn pattern from sse_cvtsi2ss<round_name> and
3652 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3653 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3654 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3655 using SWI48 mode iterator.
3656 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3657 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3658 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3659 pattern from sse_cvttss2si<round_saeonly_name>
3660 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3661 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3662 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3663 using SWI48 mode iterator.
3664 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3665 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3666 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3667 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3668 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3669 using SWI48 mode iterator.
3670 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3671 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3672 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3673 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3674 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3675 SWI48 mode iterator.
3676 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3677 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3678 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3679 pattern from sse_cvttsd2si<round_saeonly_name>
3680 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3682 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3684 * config/nds32/nds32-md-auxiliary.c
3685 (nds32_valid_smw_lwm_base_p): Refine.
3686 (nds32_output_smw_single_word): Refine.
3687 (nds32_output_smw_double_word): New.
3688 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3690 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3693 (nds32_output_stack_pop): Refine.
3694 (nds32_expand_unaligned_load): Refine.
3695 (nds32_expand_unaligned_store): Refine.
3697 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3698 Chung-Ju Wu <jasonwucj@gmail.com>
3700 * config/nds32/constants.md: Add TP_REGNUM constant.
3701 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3702 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3704 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3705 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3706 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3707 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3708 and PIC code generation.
3709 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3711 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3713 * config/nds32/nds32.md: Support TLS and PIC.
3714 * config/nds32/nds32.c: Support TLS and PIC.
3715 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3716 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3719 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3721 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3722 mode with E_ prefix.
3724 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3725 Chung-Ju Wu <jasonwucj@gmail.com>
3727 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3728 * config/nds32/nds32-md-auxiliary.c
3729 (symbolic_reference_mentioned_p): New.
3730 (nds32_legitimize_ict_address): New.
3731 (nds32_expand_ict_move): New.
3732 (nds32_indirect_call_referenced_p): New.
3733 (nds32_symbol_binds_local_p): Delete.
3734 (nds32_long_call_p): Modify.
3735 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3736 * config/nds32/nds32-protos.h
3737 (symbolic_reference_mentioned_p): Declare.
3738 (nds32_legitimize_ict_address): Declare.
3739 (nds32_expand_ict_move): Declare.
3740 (nds32_indirect_call_referenced_p): Declare.
3741 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3742 (nds32_relax_group): Use nds32_ict_const_p as condition.
3743 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3744 (nds32_asm_file_start): Output ict_model directive in asm code.
3745 (nds32_legitimate_address_p): Consider indirect call.
3746 (nds32_print_operand): Consider indirect call.
3747 (nds32_print_operand_address): Consider indirect call.
3748 (nds32_insert_attributes): Handle "indirect_call" attribute.
3749 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3750 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3751 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3752 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3753 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3754 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3755 (TARGET_ICT_MODEL_SMALL): Define.
3756 (TARGET_ICT_MODEL_LARGE): Define.
3757 * config/nds32/nds32.md (movsi): Consider ict model.
3758 (call, call_value): Consider ict model.
3759 (sibcall, sibcall_value): Consider ict model.
3760 * config/nds32/nds32.opt (mict-model): New option.
3761 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3764 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3765 Monk Chiang <sh.chiang04@gmail.com>
3766 Jim Wilson <jimw@sifive.com>
3768 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3769 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3770 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3771 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3772 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3773 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3774 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3775 compute save_libcall_adjustment properly.
3776 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3777 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3778 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3779 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3780 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3781 (ABI_SPEC): Handle mabi=ilp32e.
3782 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3783 (RVE): Add RVE mask.
3784 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3785 <-march>: Add rv32e as an example.
3787 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3790 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3791 (intra_create_variable_infos): Handle C++ constructors.
3793 2018-05-18 Martin Liska <mliska@suse.cz>
3795 * passes.def: Remove a redundant pass.
3797 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3800 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3802 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3804 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3806 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3808 (armv2, armv2a, armv3, armv3m): Delete architectures.
3809 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3810 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3811 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3813 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3814 (*mulsidi3adddi): Likewise.
3815 (mulsidi3): Likewise.
3816 (*mulsidi3_nov6): Likewise.
3817 (umulsidi3): Likewise.
3818 (umulsidi3_nov6): Likewise.
3819 (umaddsidi4): Likewise.
3820 (*umulsidi3adddi): Likewise.
3821 (smulsi3_highpart): Likewise.
3822 (*smulsi3_highpart_nov6): Likewise.
3823 (umulsi3_highpart): Likewise.
3824 (*umulsi3_highpart_nov6): Likewise.
3825 * config/arm/arm.h (arm_arch3m): Delete.
3826 * config/arm/arm.c (arm_arch3m): Delete.
3827 (arm_option_override_internal): Update armv3-related comment.
3828 (arm_configure_build_target): Delete use of isa_bit_mode32.
3829 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3830 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3831 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3833 (mulusa3): Likewise.
3834 * config/arm/arm-protos.h (arm_arch3m): Delete.
3835 * config/arm/arm-tables.opt: Regenerate.
3836 * config/arm/arm-tune.md: Likewise.
3837 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3838 deleted architectures.
3840 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3842 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3843 (armv5t, armv5te): New features.
3844 (ARMv5, ARMv5e): Delete fgroups.
3845 (ARMv5t, ARMv5te): Adjust for above changes.
3847 (armv5, armv5e): Delete arches.
3848 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3850 (*call_reg_arm): Likewise.
3851 (*call_value_reg_armv5): Likewise.
3852 (*call_value_reg_arm): Likewise.
3853 (*call_symbol): Likewise.
3854 (*call_value_symbol): Likewise.
3855 (*sibcall_insn): Likewise.
3856 (*sibcall_value_insn): Likewise.
3858 (prefetch): Likewise.
3859 (define_split and define_peephole2 dependent on arm_arch5):
3861 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3863 (TARGET_ARM_QBIT): Likewise.
3864 (TARGET_DSP_MULTIPLY): Likewise.
3865 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3866 (arm_arch5, arm_arch5e): Delete.
3867 (arm_arch5t, arm_arch5te): Declare.
3868 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3869 (arm_arch5t): Declare.
3870 (arm_option_reconfigure_globals): Update for the above.
3871 (arm_options_perform_arch_sanity_checks): Update comment, replace
3872 use of arm_arch5 with arm_arch5t.
3873 (use_return_insn): Likewise.
3874 (arm_emit_call_insn): Likewise.
3875 (output_return_instruction): Likewise.
3876 (arm_final_prescan_insn): Likewise.
3877 (arm_coproc_builtin_available): Likewise.
3878 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3879 arm_arch5e with arm_arch5t and arm_arch5te.
3880 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3881 (arm_arch5t, arm_arch5te): Declare.
3882 * config/arm/arm-tables.opt: Regenerate.
3883 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3884 * config/arm/t-multilib: Likewise.
3885 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3886 instead of arm_arch5.
3887 (*call_reg_thumb1): Likewise.
3888 (*call_value_reg_thumb1_v5): Likewise.
3889 (*call_value_reg_thumb1): Likewise.
3890 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3892 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3894 2018-05-18 Martin Liska <mliska@suse.cz>
3896 PR gcov-profile/84846
3897 * doc/gcov.texi: Document -t option of gcov tool.
3899 2018-05-18 Martin Liska <mliska@suse.cz>
3901 PR gcov-profile/84846
3902 * gcov.c (print_usage): Add new -t option.
3903 (process_args): Handle the option.
3904 (generate_results): Use stdout as output when requested by
3907 2018-05-18 Martin Liska <mliska@suse.cz>
3909 PR gcov-profile/84846
3910 * coverage.c (coverage_init): Write PWD to .gcno file.
3911 * doc/gcov.texi: Document how working directory is printed.
3912 * gcov-dump.c (dump_gcov_file): Print PWD.
3913 * gcov.c (output_intermediate_file): Likewise.
3914 (read_graph_file): Read PWD string.
3915 (output_lines): Print PWD.
3917 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3920 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3921 for retval and return false if all args to phi are zero.
3923 2018-05-18 Richard Biener <rguenther@suse.de>
3925 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3927 (evrp_dom_walker::before_dom_children): Call it.
3929 2018-05-18 Richard Biener <rguenther@suse.de>
3931 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3932 results when processing array refs with variable index.
3934 2018-05-18 Toon Moene <toon@moene.org>
3936 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3937 directly after that of -floop-interchange. Indicate that both
3938 options are enabled by default when specifying -O3.
3940 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3942 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3943 iterator. Delete separate integer-mode vec_set<mode> expander.
3944 (aarch64_simd_vec_setv2di): Delete.
3945 (vec_setv2di): Delete.
3946 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3947 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3948 the "w, r" alternative.
3950 2018-05-18 Martin Liska <mliska@suse.cz>
3952 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3953 * tree-pass.h (make_pass_lower_switch_O0): New function.
3954 * tree-switch-conversion.c (node_has_low_bound): Remove.
3955 (node_has_high_bound): Likewise.
3956 (node_is_bounded): Likewise.
3957 (class pass_lower_switch): Make it a template type and create
3959 (pass_lower_switch::execute): Add template argument.
3960 (make_pass_lower_switch): New function.
3961 (make_pass_lower_switch_O0): New function.
3962 (do_jump_if_equal): Remove.
3963 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3964 all the hard work to tree optimization passes.
3966 2018-05-18 Martin Liska <mliska@suse.cz>
3968 * dbgcnt.c (limit_low): Renamed from limit.
3969 (limit_high): New variable.
3970 (dbg_cnt_is_enabled): Check for upper limit.
3971 (dbg_cnt): Adjust dumping.
3972 (dbg_cnt_set_limit_by_index): Add new argument for high
3974 (dbg_cnt_set_limit_by_name): Likewise.
3975 (dbg_cnt_process_single_pair): Parse new format.
3976 (dbg_cnt_process_opt): Use strtok.
3977 (dbg_cnt_list_all_counters): Remove 'value' and add
3979 * doc/invoke.texi: Document changes.
3981 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3983 * doc/sourcebuild.texi (scalar_all_fma): Document.
3984 * tree.def (FMA_EXPR): Delete.
3985 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3986 * internal-fn.c (ternary_direct): New macro.
3987 (expand_ternary_optab_fn): Likewise.
3988 (direct_ternary_optab_supported_p): Likewise.
3989 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3990 * builtins.c (fold_builtin_fma): Delete.
3991 (fold_builtin_3): Don't call it.
3992 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3993 * expr.c (expand_expr_real_2): Likewise.
3994 * fold-const.c (operand_equal_p): Likewise.
3995 (fold_ternary_loc): Likewise.
3996 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3997 * gimple.c (DEFTREECODE): Likewise.
3998 * gimplify.c (gimplify_expr): Likewise.
3999 * optabs-tree.c (optab_for_tree_code): Likewise.
4000 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4001 * tree-eh.c (operation_could_trap_p): Likewise.
4002 (stmt_could_throw_1_p): Likewise.
4003 * tree-inline.c (estimate_operator_cost): Likewise.
4004 * tree-pretty-print.c (dump_generic_node): Likewise.
4005 (op_code_prio): Likewise.
4006 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4007 * tree-ssa-operands.c (get_expr_operands): Likewise.
4008 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4009 * fold-const-call.h (fold_fma): Delete.
4010 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4011 CFN_FNMA and CFN_FNMS.
4013 * genmatch.c (combined_fn): New enum.
4014 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4015 (commutative_op): New function.
4016 (commutate): Use it. Handle more than 2 operands.
4017 (dt_operand::gen_gimple_expr): Use commutative_op.
4018 (parser::parse_expr): Allow :c to be used with non-binary
4019 operators if the commutative operand is known.
4020 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4021 CFN_FMS, CFN_FNMA and CFN_FNMS.
4022 (backprop::process_assign_use): Remove FMA_EXPR handling.
4023 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4024 (gen_hsa_fma): New function.
4025 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4026 IFN_FNMA and IFN_FNMS.
4027 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4028 * gimple-fold.h (follow_all_ssa_edges): Declare.
4029 * gimple-fold.c (follow_all_ssa_edges): New function.
4030 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4031 gimple_build interface and use follow_all_ssa_edges to fold the result.
4032 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4033 instead of checking for optabs directly.
4034 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4035 rather than FMA_EXPRs.
4036 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4037 call to IFN_FMA instead of an FMA_EXPR.
4039 2018-05-17 Jim Wilson <jimw@sifive.com>
4041 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4042 sign extended promoted subreg, and the range does not have the sign bit
4043 set, then do a sign extend.
4045 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4046 test, check for sign extended subreg and/or constant operands, and
4047 do a sign extend in that case.
4049 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4051 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4052 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4054 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4055 Change logics_shift_reg to logics_shift_imm.
4056 (thunderx2t99_fp_loadpair_basic): Delete.
4057 (thunderx2t99_fp_storepair_basic): Delete.
4058 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4059 (thunderx2t99_asimd_polynomial): Delete.
4060 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4061 and neon_fp_mul_d_scalar_q.
4062 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4063 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4064 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4065 (thunderx2t99_asimd_lut): Add missing tbl types.
4066 (thunderx2t99_asimd_ext): Delete.
4067 (thunderx2t99_asimd_load1_1_mult): Delete.
4068 (thunderx2t99_asimd_load1_2_mult): Delete.
4069 (thunderx2t99_asimd_load1_ldp): New.
4070 (thunderx2t99_asimd_load1): New.
4071 (thunderx2t99_asimd_load2): Add missing *load2* types.
4072 (thunderx2t99_asimd_load3): New.
4073 (thunderx2t99_asimd_load4): New.
4074 (thunderx2t99_asimd_store1_1_mult): Delete.
4075 (thunderx2t99_asimd_store1_2_mult): Delete.
4076 (thunderx2t99_asimd_store2_mult): Delete.
4077 (thunderx2t99_asimd_store2_onelane): Delete.
4078 (thunderx2t99_asimd_store_stp): New.
4079 (thunderx2t99_asimd_store1): New.
4080 (thunderx2t99_asimd_store2): New.
4081 (thunderx2t99_asimd_store3): New.
4082 (thunderx2t99_asimd_store4): New.
4084 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4086 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4087 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4089 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4090 Segher Boessenkool <segher@kernel.crashing.org>
4093 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4096 2018-05-17 Richard Biener <rguenther@suse.de>
4098 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4099 for pruning loop and prune defs feeding only already visited PHIs.
4101 2018-05-17 Richard Biener <rguenther@suse.de>
4103 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4105 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4106 Richard Biener <rguenther@suse.de>
4108 PR tree-optimization/85793
4109 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4110 for VMAT_ELEMENTWISE.
4112 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4114 * internal-fn.h (lookup_internal_fn): Declare
4115 * internal-fn.c (lookup_internal_fn): New function.
4116 * gimple.c (gimple_build_call_from_tree): Handle calls to
4118 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4119 internal function names.
4120 * tree-pretty-print.c (dump_generic_node): Likewise.
4121 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4123 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4125 * gimple-fold.h (gimple_build): Make the function forms take
4126 combined_fn rather than built_in_function.
4127 (gimple_simplify): Likewise.
4128 * gimple-match-head.c (gimple_simplify): Likewise.
4129 * gimple-fold.c (gimple_build): Likewise.
4130 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4131 rather than gimple_build_call_internal.
4132 (get_initial_defs_for_reduction): Likewise.
4133 (vect_create_epilog_for_reduction): Likewise.
4134 (vectorizable_live_operation): Likewise.
4136 2018-05-17 Martin Liska <mliska@suse.cz>
4138 * gimple-ssa-sprintf.c (format_directive): Do not use
4139 space in between 'G_' and '('.
4141 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4144 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4145 even if the mask is not all ones.
4148 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4150 (ix86_gimple_fold_builtin): Likewise.
4153 * config/i386/i386.c: Include tree-vector-builder.h.
4154 (ix86_vector_shift_count): New function.
4155 (ix86_fold_builtin): Fold shift builtins by scalar count.
4156 (ix86_gimple_fold_builtin): Likewise.
4158 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4159 _mm512_setzero): New intrinsics.
4161 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4162 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4164 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4165 code generation for cases where splatting a value is not useful.
4166 * simplify-rtx.c (simplify_ternary_operation): Simplify
4167 vec_merge across a vec_duplicate and a paradoxical subreg forming
4168 a vector mode to a vec_concat.
4170 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4172 * config.gcc: Support "goldmont-plus".
4173 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4175 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4176 PROCESSOR_GOLDMONT_PLUS.
4177 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4178 (processor_target_table): Add "goldmont-plus".
4179 (PTA_GOLDMONT_PLUS): Define.
4180 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4181 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4182 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4183 (fold_builtin_cpu): Add "goldmont-plus".
4184 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4185 (ix86_option_override_internal): Add "goldmont-plus".
4186 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4187 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4188 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4189 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4191 2018-05-17 Richard Biener <rguenther@suse.de>
4193 PR tree-optimization/85757
4194 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4195 remove defs that only feed that PHI from further processing.
4197 2018-05-16 Jim Wilson <jimw@sifive.com>
4199 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4201 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4203 2018-05-16 Mark Wielaard <mark@klomp.org>
4205 * dwarf2out.c (count_index_strings): New function.
4206 (output_indirect_strings): Call count_index_strings and generate
4207 header for dwarf_version >= 5.
4209 2018-05-16 Mark Wielaard <mark@klomp.org>
4211 * dwarf2out.c (dwarf_FORM): New function.
4212 (set_indirect_string): Use dwarf_FORM.
4213 (reset_indirect_string): Likewise.
4214 (size_of_die): Likewise.
4215 (value_format): Likewise.
4216 (output_die): Likewise.
4217 (add_skeleton_AT_string): Likewise.
4218 (output_macinfo_op): Likewise.
4219 (index_string): Likewise.
4220 (output_index_string_offset): Likewise.
4221 (output_index_string): Likewise.
4222 (count_index_strings): Likewise.
4224 2018-05-16 Carl Love <cel@us.ibm.com>
4226 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4227 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4229 2018-05-16 Martin Jambor <mjambor@suse.cz>
4231 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4232 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4234 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4236 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4237 (fnma<mode>4): Likewise.
4238 (fms<mode>4): Likewise.
4239 (fnms<mode>4): Likewise.
4240 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4241 (aarch64_fnma<mode>4): Likewise.
4242 (aarch64_fms<mode>4): Likewise.
4243 (aarch64_fnms<mode>4): Likewise.
4244 (aarch64_fnmadd<mode>4): Likewise.
4246 2018-05-16 Jason Merrill <jason@redhat.com>
4248 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4250 2018-05-16 Richard Biener <rguenther@suse.de>
4252 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4253 (dump_stmt_cost): Declare.
4254 (add_stmt_cost): Dump cost we add.
4255 (add_stmt_costs): New function.
4256 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4258 (vect_analyze_stmt): Adjust prototype.
4259 (vectorizable_condition): Likewise.
4260 (vectorizable_live_operation): Likewise.
4261 (vectorizable_reduction): Likewise.
4262 (vectorizable_induction): Likewise.
4263 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4264 cost vector to pass to vectorizable_ and record afterwards.
4265 (vect_model_reduction_cost): Take cost vector argument and adjust.
4266 (vect_model_induction_cost): Likewise.
4267 (vectorizable_reduction): Likewise.
4268 (vectorizable_induction): Likewise.
4269 (vectorizable_live_operation): Likewise.
4270 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4271 SLP_TREE_NUMBER_OF_VEC_STMTS.
4272 (vect_analyze_slp_cost_1): Remove.
4273 (vect_analyze_slp_cost): Likewise.
4274 (vect_slp_analyze_node_operations): Take visited args and
4275 a target cost vector. Avoid processing already visited stmt sets.
4276 (vect_slp_analyze_operations): Use a local cost vector to gather
4277 costs and register those of non-discarded instances.
4278 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4279 (vect_schedule_slp_instance): Remove copying of
4280 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4282 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4283 adding cost. Record cost entry location.
4284 (vect_prologue_cost_for_slp_op): Function to compute cost of
4285 a constant or invariant generated for SLP vect in the prologue,
4286 split out from vect_analyze_slp_cost_1.
4287 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4288 (vect_model_promotion_demotion_cost): Likewise.
4289 (vect_model_store_cost): Likewise, make static.
4290 (vect_model_load_cost): Likewise.
4291 (vectorizable_bswap): Add cost vector arg and adjust.
4292 (vectorizable_call): Likewise.
4293 (vectorizable_simd_clone_call): Likewise.
4294 (vectorizable_conversion): Likewise.
4295 (vectorizable_assignment): Likewise.
4296 (vectorizable_shift): Likewise.
4297 (vectorizable_operation): Likewise.
4298 (vectorizable_store): Likewise.
4299 (vectorizable_load): Likewise.
4300 (vectorizable_condition): Likewise.
4301 (vectorizable_comparison): Likewise.
4302 (can_vectorize_live_stmts): Likewise.
4303 (vect_analyze_stmt): Likewise.
4304 (vect_transform_stmt): Adjust calls to vectorizable_*.
4305 * tree-vectorizer.c: Include gimple-pretty-print.h.
4306 (dump_stmt_cost): New function.
4308 2018-05-16 Richard Biener <rguenther@suse.de>
4310 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4311 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4312 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4313 (check_name): New callback.
4314 (dse_classify_store): Track cycles via a visited bitmap of PHI
4315 defs and simplify handling of in-loop and across loop dead stores
4316 and properly fail for loop-variant refs. Handle byte-tracking with
4317 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4320 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4322 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4323 (vect_get_mask_type_for_stmt): Likewise.
4324 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4326 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4327 to determine the statement's vector type and the vector type that
4328 should be used for calculating nunits. Deal with cases in which
4329 the type has to be deferred.
4330 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4331 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4332 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4333 (vect_determine_vf_for_stmt): New functions, split out from...
4334 (vect_determine_vectorization_factor): ...here.
4335 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4336 (vect_get_mask_type_for_stmt): New functions, split out from
4337 vect_determine_vectorization_factor.
4339 2018-05-16 Richard Biener <rguenther@suse.de>
4341 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4342 verify the [VEC_]COND_EXPR embedded comparison.
4344 2018-05-15 Martin Sebor <msebor@redhat.com>
4346 PR tree-optimization/85753
4347 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4348 RECORD_TYPE in addition to ARRAY_TYPE.
4350 2018-05-15 Martin Sebor <msebor@redhat.com>
4353 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4355 2018-05-15 Richard Biener <rguenther@suse.de>
4357 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4358 add by_clobber_p one. Change algorithm to collect all defs
4359 representing uses we need to walk and try reducing them to
4360 a single one before failing.
4361 (dse_dom_walker::dse_optimize_stmt): Adjust.
4363 2018-05-13 Mark Wielaard <mark@klomp.org>
4365 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4366 (size_of_loc_descr): Likewise.
4367 (output_loc_operands): Likewise.
4368 (output_loc_operands_raw): Likewise.
4369 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4370 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4371 (hash_loc_operands): Likewise.
4372 (compare_loc_operands): Likewise.
4374 2018-05-14 Mark Wielaard <mark@klomp.org>
4376 * dwarf2out.c (count_index_addrs): New function.
4377 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4379 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4381 PR tree-optimization/83648
4382 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4383 return value as malloc candidate.
4385 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4388 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4389 param as true in call to suggest_attribute.
4391 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4393 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4394 -mreadonly-in-sdata.
4396 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4398 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4400 (aarch64_crypto_aesd_fused): Likewise.
4402 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4404 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4405 (movsi_aarch64): Likewise.
4406 (load_pairsi): Likewise.
4407 (load_pairdi): Likewise.
4408 (store_pairsi): Likewise.
4409 (store_pairdi): Likewise.
4410 (load_pairsf): Likewise.
4411 (load_pairdf): Likewise.
4412 (store_pairsf): Likewise.
4413 (store_pairdf): Likewise.
4414 (zero_extend): Likewise.
4415 (trunc): Swap alternatives.
4416 (fcvt_target): Add '?' to prefer w over r.
4418 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4421 * config/i386/i386.md: Disallow non-commutative arithmetics in
4422 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4423 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4424 in the peephole2 before it.
4426 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4428 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4429 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4430 (ix86_handle_option): Handle -mcldemote.
4431 * config.gcc: New header.
4432 * config/i386/cldemoteintrin.h: New file.
4433 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4434 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4436 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4437 OPTION_MASK_ISA_CLDEMOTE.
4438 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4439 (ix86_valid_target_attribute_inner_p): Ditto.
4440 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4441 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4442 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4443 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4444 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4446 * config/i386/i386.opt: Add -mcldemote.
4447 * config/i386/x86intrin.h: New header.
4448 * doc/invoke.texi: Add -mcldemote.
4450 2018-05-14 Richard Biener <rguenther@suse.de>
4452 * doc/match-and-simplify.texi: Adjust :s documentation.
4454 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4456 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4457 intended memcpy size.
4458 (REORDER_45): Likewise.
4460 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4462 * sort.cc: New file.
4463 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4464 * vec.c (qsort_chk): Use gcc_qsort.
4465 * Makefile.in (OBJS-libcommon): Add sort.o.
4466 (build/sort.o): New target. Use it...
4467 (BUILD_RTL): ... here, and...
4468 (build/gencfn-macros): ... here, and...
4469 (build/genmatch): ... here.
4471 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4472 Chung-Ju Wu <jasonwucj@gmail.com>
4474 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4475 * config/nds32/nds32-graywolf.md: New file.
4476 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4477 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4479 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4480 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4481 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4482 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4483 * config/nds32/pipelines.md: Include n15 settings.
4485 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4486 Chung-Ju Wu <jasonwucj@gmail.com>
4488 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4489 * config/nds32/nds32-n13.md: New file.
4490 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4491 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4493 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4494 * config/nds32/nds32.md (pipeline_model): Add n13.
4495 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4496 * config/nds32/pipelines.md: Include n13 settings.
4498 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4499 Chung-Ju Wu <jasonwucj@gmail.com>
4501 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4502 * config/nds32/nds32-n10.md: New file.
4503 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4504 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4506 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4507 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4508 * config/nds32/nds32.md (pipeline_model): Add n10.
4509 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4510 * config/nds32/pipelines.md: Include n10 settings.
4512 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4513 Kito Cheng <kito.cheng@gmail.com>
4514 Chung-Ju Wu <jasonwucj@gmail.com>
4516 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4517 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4518 Add enum values for DSP extension instructions.
4519 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4521 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4522 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4524 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4525 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4526 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4527 * config/nds32/nds32-intrinsic.md: Likewise.
4528 * config/nds32/nds32_intrinsic.h: Likewise.
4529 * config/nds32/nds32-md-auxiliary.c: Likewise.
4530 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4531 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4532 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4533 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4534 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4535 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4536 TYPE_DMAC in switch statement.
4537 * config/nds32/nds32.c: New checking and implementation for DSP
4538 extension instructions.
4539 * config/nds32/nds32.h: Likewise.
4540 * config/nds32/nds32.md: Likewise.
4541 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4542 * config/nds32/predicates.md: Implement new predicates for DSP
4545 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4547 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4548 Reformat alternatives and attributes so it is easier to identify
4549 which constraints/attributes go with which instruction.
4550 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4551 (mov<mode>_softfloat32, FMOVE64): Likewise.
4552 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4553 (mov<mode>_softfloat64, FMOVE64): Likewise.
4555 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4557 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4559 (Basic PowerPC Built-in Functions): The new name of the
4560 subsection previously known as "PowerPC Built-in Functions".
4561 (Basic PowerPC Built-in Functions Available on all Configurations):
4563 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4564 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4565 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4566 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4568 2018-05-11 Martin Jambor <mjambor@suse.cz>
4571 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4574 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4577 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4579 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4581 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4582 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4583 (ix86_handle_option): Handle -mwaitpkg.
4584 * config.gcc: New header.
4585 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4586 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4587 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4589 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4590 OPTION_MASK_ISA_WAITPKG.
4591 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4592 (ix86_option_override_internal): Add PTA_WAITPKG.
4593 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4594 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4595 IX86_BUILTIN_TPAUSE.
4596 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4597 __builtin_ia32_umwait and __builtin_ia32_tpause.
4598 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4599 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4600 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4601 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4602 UNSPECV_TPAUSE): New.
4603 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4604 * config/i386/i386.opt: Add -mwaitpkg.
4605 * config/i386/waitpkgintrin.h: New file.
4606 * config/i386/x86intrin.h: New header.
4607 * doc/invoke.texi: Add -mwaitpkg.
4609 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4612 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4614 (cortex-m0): Use armv6s-m isa.
4615 (cortex-m0plus): Likewise.
4616 (cortex-m1): Likewise.
4617 (cortex-m0.small-multiply): Likewise.
4618 (cortex-m0plus.small-multiply): Likewise.
4619 (cortex-m1.small-multiply): Likewise.
4621 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4622 Jakub Jelinek <jakub@redhat.com>
4624 PR tree-optimization/85692
4625 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4626 source permute as well.
4628 2018-05-11 Martin Liska <mliska@suse.cz>
4631 * doc/extend.texi: Document LLVM style format for no_sanitize
4634 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4636 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4637 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4638 (mode_supports_vsx_dform_quad): Likewise.
4639 (mode_supports_vmx_dform): Move these functions to be next to the
4640 other mode_supports functions.
4641 (mode_supports_dq_form): Likewise.
4642 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4643 mode_supports_dq_form.
4644 (reg_offset_addressing_ok_p): Likewise.
4645 (offsettable_ok_by_alignment): Likewise.
4646 (rs6000_legitimate_offset_address_p): Likewise.
4647 (legitimate_lo_sum_address_p): Likewise.
4648 (rs6000_legitimize_address): Likewise.
4649 (rs6000_legitimize_reload_address): Likewise.
4650 (rs6000_secondary_reload_inner): Likewise.
4651 (rs6000_preferred_reload_class): Likewise.
4652 (rs6000_output_move_128bit): Likewise.
4654 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4656 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4657 Generate SImode target register for null target.
4658 <case IX86_BUILTIN_XGETBV>: Ditto.
4659 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4660 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4662 2018-05-10 Carl Love <cel@us.ibm.com>
4664 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4665 dcbtt and dcbtstt if operands[2] is 0.
4667 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4670 * config/i386/sse.md (usadv64qi): New expander.
4672 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4674 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4675 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4676 -maltivec=be support.
4677 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4678 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4679 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4680 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4681 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4682 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4683 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4684 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4685 altivec_vsumsws): Adjust.
4686 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4687 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4688 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4690 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4691 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4692 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4693 (altivec_lve<VI_char>x): Delete expand.
4694 (*altivec_lve<VI_char>x_internal): Rename to...
4695 (altivec_lve<VI_char>x): ... this.
4696 (altivec_lvxl_<mode>): Delete expand.
4697 (*altivec_lvxl_<mode>_internal): Rename to ...
4698 (altivec_lvxl_<mode>): ... this.
4699 (altivec_stvxl_<mode>): Delete expand.
4700 (*altivec_stvxl_<mode>_internal): Rename to ...
4701 (altivec_stvxl_<mode>): ... this.
4702 (altivec_stve<VI_char>x): Delete expand.
4703 (*altivec_stve<VI_char>x_internal): Rename to ...
4704 (altivec_stve<VI_char>x): ... this.
4705 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4706 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4707 reduc_plus_scal_<mode>): Adjust.
4708 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4710 (rs6000_cpu_cpp_builtins): Adjust.
4711 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4712 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4713 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4714 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4715 -maltivec=be support.
4716 (rs6000_split_vec_extract_var): Adjust.
4717 (rs6000_split_v4si_init): Adjust.
4718 (swap_selector_for_mode): Delete.
4719 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4720 altivec_expand_stvex_be): Delete.
4721 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4722 -maltivec=be support.
4723 (rs6000_gimple_fold_builtin): Ditto.
4724 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4726 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4727 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4728 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4729 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4730 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4731 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4732 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4733 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4734 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4735 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4736 anonymous split): Adjust.
4737 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4738 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4740 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4742 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4743 when --with-gxx-include-dir is also specified.
4744 * configure: Regenerate.
4746 2018-05-09 Jim Wilson <jimw@sifive.com>
4749 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4750 * config/riscv/t-withmultilib: New.
4751 * config/riscv/withmultilib.h: New.
4752 * doc/install.texi: Document RISC-V --with-multilib-list support.
4754 2018-05-09 Richard Biener <rguenther@suse.de>
4756 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4758 (vect_bb_vectorization_profitable_p): Adjust. Compute
4759 actual scalar cost using the cost vector and the add_stmt_cost
4762 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4764 PR rtl-optimization/85645
4765 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4766 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4768 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4770 PR rtl-optimization/85645
4771 * shrink-wrap.c (spread_components): Return a boolean saying if
4772 anything was changed.
4773 (try_shrink_wrapping_separate): Iterate spread_components until
4774 nothing changes anymore.
4776 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4778 PR rtl-optimization/85645
4779 * regrename.c (build_def_use): Also kill the chains that include the
4780 destination of a REG_CFA_REGISTER note.
4782 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4784 PR rtl-optimization/85645
4785 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4786 insn that has a REG_CFA_REGISTER note.
4788 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4790 * cfgexpand.c (expand_clobber): New function.
4791 (expand_gimple_stmt_1): Use it.
4792 * tree-vect-stmts.c (vect_clobber_variable): New function,
4794 (vectorizable_simd_clone_call): ...here.
4795 (vectorizable_store): Emit a clobber either side of an
4796 IFN_STORE_LANES sequence.
4797 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4799 2018-05-09 Tom de Vries <tom@codesourcery.com>
4802 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4803 (define_insn "trap_if_false"): Add exit after trap.
4805 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4807 PR rtl-optimization/85638
4808 * bb-reorder.c: Include common/common-target.h.
4809 (create_forwarder_block): New function extracted from...
4810 (fix_up_crossing_landing_pad): ...here. Rename into...
4811 (dw2_fix_up_crossing_landing_pad): ...this.
4812 (sjlj_fix_up_crossing_landing_pad): New function.
4813 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4814 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4815 from both partitions and exit the loop after one iteration.
4817 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4820 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4822 (Basic PowerPC Built-in Functions): The new name of the
4823 subsection previously known as "PowerPC Built-in Functions".
4824 (Basic PowerPC Built-in Functions Available on all Configurations):
4826 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4828 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4830 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4832 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4835 2018-05-08 Jim Wilson <jimw@sifive.com>
4837 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4838 (LD_EMUL_SUFFIX): New.
4839 (LINK_SPEC): Use it.
4841 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4843 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4845 (Basic PowerPC Built-in Functions): The new name of the
4846 subsection previously known as "PowerPC Built-in Functions".
4847 (Basic PowerPC Built-in Functions Available on all Configurations):
4849 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4851 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4853 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4855 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4858 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4861 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4862 after cmpelim optimization.
4864 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4866 * config.gcc: Support "goldmont".
4867 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4868 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4870 * config/i386/i386.c (m_GOLDMONT): Define.
4871 (processor_target_table): Add "goldmont".
4872 (PTA_GOLDMONT): Define.
4873 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4874 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4875 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4876 (fold_builtin_cpu): Add "goldmont".
4877 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4878 (ix86_option_override_internal): Add "goldmont".
4879 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4880 (processor_type): Add PROCESSOR_GOLDMONT.
4881 * config/i386/i386.md: Add CPU "glm".
4882 * config/i386/glm.md: New file.
4883 * config/i386/x86-tune.def: Add m_GOLDMONT.
4884 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4886 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4889 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4891 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4892 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4893 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4896 * config/i386/i386.c (ix86_fold_builtin): Handle
4897 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4900 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4901 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4903 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4906 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4907 (check_arch): Likewise.
4908 (check_fpu): Return the result rather than printing it.
4909 (end arch): Fix operator precedence.
4910 (end cpu): Likewise.
4911 (END): Print the result from check_fpu.
4913 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4914 Alan Hayward <alan.hayward@arm.com>
4915 David Sherwood <david.sherwood@arm.com>
4917 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4918 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4919 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4920 (*fcmuo<mode>_and): New patterns.
4922 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4924 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4925 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4926 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4927 (cmp_op, sve_imm_con): New code attributes.
4928 (SVE_COND_INT_CMP, imm_con): Delete.
4929 (cmp_op): Remove above unspecs from int attribute.
4930 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4932 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4933 comparison-specific unspecs.
4934 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4935 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4936 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4937 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4938 (*vec_fcm<cmp_op><mode>): Rename to...
4939 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4940 (*vec_fcmuo<mode>): Rename to...
4941 (*fcmuo<mode>): ...this and adjust likewise.
4942 (*pred_fcm<cmp_op><mode>): New pattern.
4943 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4944 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4946 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4948 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4949 (aarch64_emit_sve_predicated_cond): New function.
4950 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4951 (aarch64_emit_unspec_cond_or): Replace with...
4952 (aarch64_emit_sve_or_conds): ...this new function. Use
4953 aarch64_emit_sve_ptrue_op for the individual comparisons and
4954 aarch64_emit_binop to OR them together.
4955 (aarch64_emit_inverted_unspec_cond): Replace with...
4956 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4957 aarch64_emit_sve_ptrue_op for the comparison and
4958 aarch64_emit_unop to invert the result.
4959 (aarch64_expand_sve_vec_cmp_float): Update after the above
4960 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4962 2018-05-07 Nathan Sidwell <nathan@acm.org>
4964 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4965 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4966 (Backwards Compatibility): Likewise.
4968 2018-05-07 Luis Machado <luis.machado@linaro.org>
4972 2018-05-07 Luis Machado <luis.machado@linaro.org>
4974 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4975 <prefetch_dynamic_strides>: New const bool field.
4976 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4977 prefetch_dynamic_strides.
4978 (exynosm1_prefetch_tune): Likewise.
4979 (thunderxt88_prefetch_tune): Likewise.
4980 (thunderx_prefetch_tune): Likewise.
4981 (thunderx2t99_prefetch_tune): Likewise.
4982 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4984 (aarch64_override_options_internal): Update to set
4985 PARAM_PREFETCH_DYNAMIC_STRIDES.
4986 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4987 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4988 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4989 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4990 prefetch-dynamic-strides setting.
4992 2018-05-07 Luis Machado <luis.machado@linaro.org>
4994 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4995 <minimum_stride>: New const int field.
4996 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4997 minimum_stride field.
4998 (exynosm1_prefetch_tune): Likewise.
4999 (thunderxt88_prefetch_tune): Likewise.
5000 (thunderx_prefetch_tune): Likewise.
5001 (thunderx2t99_prefetch_tune): Likewise.
5002 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5003 (aarch64_override_options_internal): Update to set
5004 PARAM_PREFETCH_MINIMUM_STRIDE.
5005 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5006 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5007 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5008 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5009 stride is constant and is below the minimum stride threshold.
5011 2018-05-07 Luis Machado <luis.machado@linaro.org>
5013 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5016 2018-05-07 Luis Machado <luis.machado@linaro.org>
5018 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5019 <prefetch_dynamic_strides>: New const bool field.
5020 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5021 prefetch_dynamic_strides.
5022 (exynosm1_prefetch_tune): Likewise.
5023 (thunderxt88_prefetch_tune): Likewise.
5024 (thunderx_prefetch_tune): Likewise.
5025 (thunderx2t99_prefetch_tune): Likewise.
5026 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5028 (aarch64_override_options_internal): Update to set
5029 PARAM_PREFETCH_DYNAMIC_STRIDES.
5030 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5031 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5032 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5033 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5034 prefetch-dynamic-strides setting.
5036 2018-05-07 Luis Machado <luis.machado@linaro.org>
5038 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5039 <minimum_stride>: New const int field.
5040 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5041 minimum_stride field.
5042 (exynosm1_prefetch_tune): Likewise.
5043 (thunderxt88_prefetch_tune): Likewise.
5044 (thunderx_prefetch_tune): Likewise.
5045 (thunderx2t99_prefetch_tune): Likewise.
5046 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5047 (aarch64_override_options_internal): Update to set
5048 PARAM_PREFETCH_MINIMUM_STRIDE.
5049 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5050 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5051 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5052 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5053 stride is constant and is below the minimum stride threshold.
5055 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5058 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5059 the type is addressable. Don't force op into register if it has
5062 2018-05-05 Roland McGrath <mcgrathr@google.com>
5065 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5066 any section for which we don't know a specific type it should have,
5067 regardless of name. Previously this was done only for the exact
5068 names ".init_array", ".fini_array", and ".preinit_array".
5069 (default_elf_asm_named_section): Add comment about
5070 relationship with default_section_type_flags and SECTION_NOTYPE.
5071 (get_section): Don't consider it a type conflict if one side has
5072 SECTION_NOTYPE and the other doesn't, as long as neither has the
5073 SECTION_BSS et al used in the default_section_type_flags logic.
5075 2018-05-05 Tom de Vries <tom@codesourcery.com>
5078 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5079 (workaround_barsyncs): New function.
5080 (nvptx_reorg): Use workaround_barsyncs.
5081 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5082 (define_expand "nvptx_membar_cta"): New define_expand.
5083 (define_insn "*nvptx_membar_cta"): New insn.
5085 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5087 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5088 To improve optimization opportunities.
5089 * builtin-types.def: The new needed builtin types for the above.
5091 2018-05-04 Richard Biener <rguenther@suse.de>
5093 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5094 * gimple-ssa-store-merging.c
5095 (imm_store_chain_info::output_merged_store): Remove redundant create,
5096 release split_store vector contents on failure.
5097 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5098 scalar stmt vector on cache hit.
5100 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5102 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5104 * config.gcc (powerpc-xilinx-eabi*): Remove.
5105 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5107 (fusion_addis_mem_combo_load): Ditto.
5108 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5110 (rs6000_cpu_cpp_builtins): Ditto.
5111 * config/rs6000/rs6000-linux.c
5112 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5113 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5114 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5116 (rs6000_setup_reg_addr_masks): Ditto.
5117 (rs6000_init_hard_regno_mode_ok): Ditto.
5118 (rs6000_option_override_internal): Ditto.
5119 (legitimate_lo_sum_address_p): Ditto.
5120 (rs6000_legitimize_address): Ditto.
5121 (rs6000_legitimize_reload_address): Ditto.
5122 (rs6000_legitimate_address_p): Ditto.
5123 (abi_v4_pass_in_fpr): Ditto.
5124 (setup_incoming_varargs): Ditto.
5125 (rs6000_gimplify_va_arg): Ditto.
5126 (rs6000_split_multireg_move): Ditto.
5127 (rs6000_savres_strategy): Ditto.
5128 (rs6000_emit_prologue_components): Ditto.
5129 (rs6000_emit_epilogue_components): Ditto.
5130 (rs6000_emit_prologue): Ditto.
5131 (rs6000_emit_epilogue): Ditto.
5132 (rs6000_elf_file_end): Ditto.
5133 (rs6000_function_value): Ditto.
5134 (rs6000_libcall_value): Ditto.
5135 * config/rs6000/rs6000.h: Ditto.
5136 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5137 (TARGET_MINMAX): ... this. New.
5138 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5139 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5140 (*movsi_internal1_single): Delete.
5141 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5142 mfpu=, mxilinx-fpu): Delete.
5143 * config/rs6000/singlefp.h: Delete.
5144 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5145 * config/rs6000/t-rs6000: Ditto.
5146 * config/rs6000/t-xilinx: Delete.
5147 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5148 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5149 (VStype_simple): Delete.
5150 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5151 * config/rs6000/xfpu.h: Delete.
5152 * config/rs6000/xfpu.md: Delete.
5153 * config/rs6000/xilinx.h: Delete.
5154 * config/rs6000/xilinx.opt: Delete.
5155 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5156 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5158 2018-05-04 Tom de Vries <tom@codesourcery.com>
5161 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5164 2018-05-04 Richard Biener <rguenther@suse.de>
5167 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5168 (expand_complex_div_wide): Likewise.
5169 (expand_complex_operations_1): Likewise.
5170 (expand_complex_libcall): Preserve EH info of the original stmt.
5171 (tree_lower_complex): Handle removed blocks.
5172 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5173 on complex multiplication and division libcall builtins.
5175 2018-05-04 Richard Biener <rguenther@suse.de>
5178 * fold-const.c (negate_expr_p): Restrict negation of operand
5179 zero of a division to when we know that can happen without
5181 (fold_negate_expr_1): Likewise.
5183 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5186 * real.h (real_nextafter): Declare.
5187 * real.c (real_nextafter): New function.
5188 * fold-const-call.c (fold_const_nextafter): New function.
5189 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5190 CASE_CFN_NEXTTOWARD.
5191 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5192 even when arg1_mode is different from arg0_mode.
5194 2018-05-03 Nathan Sidwell <nathan@acm.org>
5196 * doc/extend.texi (Deprecated Features): Remove
5198 (Backwards Compatibility): Likewise.
5199 * doc/invoke.texi (C++ Language Options): Likewise.
5200 (C++ Dialect Options): Likewise.
5202 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5205 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5206 _mm512_mask_mullox_epi64): New intrinsics.
5208 2018-05-03 Tom de Vries <tom@codesourcery.com>
5211 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5212 dump files): Add offload-tree.
5214 2018-05-03 Richard Biener <rguenther@suse.de>
5216 PR tree-optimization/85615
5217 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5218 to loops not nested in BBs loop father to avoid creating multi-entry
5221 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5223 PR tree-optimization/70291
5224 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5225 arguments. Change return type to tree. Emit libcall as a new
5226 statement rather than replacing existing one when inplace_p is true.
5227 (expand_complex_multiplication_components): New function.
5228 (expand_complex_multiplication): Expand floating-point complex
5229 multiplication using the above.
5230 (expand_complex_division): Rename inner_type parameter to type.
5231 Update expand_complex_libcall call-site.
5232 (expand_complex_operations_1): Update expand_complex_multiplication
5233 and expand_complex_division call-sites.
5235 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5238 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5239 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5240 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5241 the highest significant bit of the shift count mask is clear. In
5242 check whether and[sq]i3 is needed verify that all significant bits
5243 of the shift count other than the highest are set.
5245 2018-05-02 Tom de Vries <tom@codesourcery.com>
5248 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5249 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5250 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5251 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5252 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5253 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5254 __builtin_goacc_parlevel_size.
5256 2018-05-02 Richard Biener <rguenther@suse.de>
5258 PR tree-optimization/85597
5259 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5260 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5263 2018-05-02 Tom de Vries <tom@codesourcery.com>
5266 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5267 dump files): Add ltrans-tree.
5269 2018-05-02 Tom de Vries <tom@codesourcery.com>
5272 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5273 dump files): Add wpa-ipa.
5275 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5277 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5278 powerpc*-*-linux*paired* target.
5279 * config/rs6000/750cl.h: Delete.
5280 * config/rs6000/paired.h: Delete.
5281 * config/rs6000/paired.md: Delete.
5282 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5284 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5285 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5286 comment. Remove paired float support.
5287 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5288 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5290 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5291 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5293 * config/rs6000/rs6000.c: Remove paired float support.
5294 (paired_expand_vector_init, paired_expand_vector_move,
5295 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5296 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5297 paired_expand_builtin, paired_expand_predicate_builtin,
5298 paired_init_builtins): Delete.
5299 * config/rs6000/rs6000.h: Remove paired float support.
5300 * config/rs6000/rs6000.md: Remove paired float support.
5301 (move_from_CR_ov_bit): Delete.
5302 * config/rs6000/rs6000.opt (mpaired): Delete.
5303 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5304 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5306 2018-05-02 Richard Biener <rguenther@suse.de>
5309 * gimplify.c (gimplify_save_expr): When in SSA form allow
5310 SAVE_EXPRs to compute to SSA vars.
5312 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5315 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5316 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5317 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5318 clobber operands[2], instead use a new pseudo. Formatting fixes.
5320 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5322 PR tree-optimization/85586
5323 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5324 exit early for statements in the same group if the accesses are
5327 2018-05-02 Tom de Vries <tom@codesourcery.com>
5330 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5333 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5335 PR tree-optimization/85143
5336 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5338 2018-05-01 Tom de Vries <tom@codesourcery.com>
5341 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5342 not found" error message.
5344 2018-05-01 Tom de Vries <tom@codesourcery.com>
5347 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5348 * vec.c (test_ordered_remove_if): New function.
5349 (vec_c_tests): Call test_ordered_remove_if.
5350 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5351 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5352 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5353 VEC_ORDERED_REMOVE_IF.
5355 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5357 PR tree-optimization/82665
5358 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5359 pointer subtraction where arguments come from a memchr call.
5361 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5363 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5364 --push-state --as-needed and --pop-state instead of --as-needed and
5365 --no-as-needed if ld supports it.
5366 * configure: Regenerated.
5369 * doc/install.texi2html: Replace _002d with - and _002a with * in
5370 generated html files using sed.
5372 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5375 * gcc-rich-location.c (blank_line_before_p): New function.
5376 (use_new_line): New function.
5377 (gcc_rich_location::add_fixit_insert_formatted): New function.
5378 * gcc-rich-location.h
5379 (gcc_rich_location::add_fixit_insert_formatted): New function.
5381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5383 * selftest.c (assert_streq): Rename "expected" and "actual" to
5384 "val1" and "val2". Extend NULL-handling to cover both inputs
5385 symmetrically, while still requiring both to be non-NULL for a pass.
5386 * selftest.h (assert_streq): Rename "expected" and "actual" to
5388 (ASSERT_EQ): Likewise.
5389 (ASSERT_EQ_AT): Likewise.
5390 (ASSERT_KNOWN_EQ): Likewise.
5391 (ASSERT_KNOWN_EQ_AT): Likewise.
5392 (ASSERT_NE): Likewise.
5393 (ASSERT_MAYBE_NE): Likewise.
5394 (ASSERT_MAYBE_NE_AT): Likewise.
5395 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5396 the assertion to pass.
5397 (ASSERT_STREQ_AT): Likewise.
5399 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5401 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5402 interaction with -pie.
5404 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5406 * selftest.h: Fix alphabetization of per-source-file selftest
5409 2018-04-30 Jason Merrill <jason@redhat.com>
5411 PR c++/61982 - dead stores to destroyed objects.
5412 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5415 2018-04-30 Jason Merrill <jason@redhat.com>
5417 * tree.c (build_clobber): New.
5418 * tree.h: Declare it.
5419 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5421 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5423 * diagnostic-show-locus.c (layout::layout): Update for
5424 location_get_source_line returning a char_span.
5425 (struct char_span): Move to input.h.
5426 (struct correction): Update for fields in char_span becoming
5428 (struct source_line): Update for location_get_source_line
5429 returning a char_span.
5430 (layout::print_line): Likewise.
5431 * edit-context.c (edited_file::print_content): Likewise.
5432 (edited_file::print_diff_hunk): Likewise.
5433 (edited_file::print_run_of_changed_lines): Likewise.
5434 (edited_file::get_num_lines): Likewise.
5435 (edited_line::edited_line): Likewise.
5436 * final.c (asm_show_source): Likewise.
5437 * input.c (location_get_source_line): Convert return type
5438 from const char * to char_span, losing the final "line_len"
5440 (dump_location_info): Update for the above.
5441 (get_substring_ranges_for_loc): Likewise. Use a char_span
5442 when handling the literal within the line.
5443 (test_reading_source_line): Update for location_get_source_line
5444 returning a char_span.
5445 * input.h (class char_span): Move here from
5446 diagnostic-show-locus.c, converting from a struct to a class.
5447 Make data members private.
5448 (char_span::operator bool): New.
5449 (char_span::length): New.
5450 (char_span::get_buffer): New.
5451 (char_span::operator[]): New.
5452 (char_span::subspan): Make const.
5453 (char_span::xstrdup): New.
5454 (location_get_source_line): Convert return type from const char *
5455 to char_span, losing the final "line_size" param.
5457 2018-04-30 Jan Hubicka <jh@suse.cz>
5459 * lto-wrapper.c (ltrans_priorities): New static var.
5460 (cmp_priority): New.
5461 (run_gcc): Read priorities and if doing parallel build order
5462 the Makefile by them.
5464 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5466 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5468 2018-04-30 Richard Biener <rguenther@suse.de>
5470 * tree-cfg.c (verify_address): Remove base argument, add
5471 flag whether to check TREE_ADDRESSABLE and do that.
5472 (verify_expr): Remove.
5473 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5474 (verify_gimple_assign_single): Likewise.
5475 (verify_gimple_switch): Likewise.
5476 (verify_expr_location_1): Dereference tp once. Add (disabled)
5477 piece from verify_expr.
5478 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5480 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5482 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5484 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5486 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5487 (small_data_pattern): Likewise.
5488 (arc_rewrite_small_data): Likewise.
5489 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5490 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5491 (get_symbol_alignment): New function.
5492 (legitimate_small_data_address_p): Likewise.
5493 (legitimate_scaled_address): Update, call
5494 legitimate_small_data_address_p.
5495 (output_sdata): New static variable.
5496 (arc_print_operand): Update how we handle small data operands.
5497 (arc_print_operand_address): Likewise.
5498 (arc_legitimate_address_p): Update, use
5499 legitimate_small_data_address_p.
5500 (arc_rewrite_small_data_p): Remove.
5501 (arc_rewrite_small_data_1): Likewise.
5502 (arc_rewrite_small_data): Likewise.
5503 (small_data_pattern): Likewise.
5504 (compact_sda_memory_operand): Update to use
5505 legitimate_small_data_address_p and get_symbol_alignment.
5506 (prepare_move_operands): Don't rewite sdata pattern.
5507 (prepare_extend_operands): Remove.
5508 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5510 (zero_extendqisi2): Likewise.
5511 (zero_extendhisi2): Likewise.
5512 (extendqihi2): Likewise.
5513 (extendqisi2): Likewise.
5514 (extendhisi2): Likewise.
5518 * config/arc/constraints.md (Usd): Change it to memory constraint.
5520 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5522 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5523 as source of std instructions.
5524 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5525 allow 6-bit constants as source for store instructions.
5526 (movdi_insn): Update instruction pattern to allow 6-bit constants
5527 as source for store instructions.
5529 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5531 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5533 2018-04-30 Nathan Sidwell <nathan@acm.org>
5534 Sandra Loosemore <sandra@codesourcery.com>
5536 * dumpfile.c (dump_open): Allow '-' for stdout.
5537 * doc/invoke.texi (Developer Options): Document dump filename
5538 determination early. Document stdin/stdout selection.
5540 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5542 Microblaze Target: PIC data text relative
5544 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5545 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5547 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5548 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5549 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5550 New addressing mode for data-text relative position indepenedent code.
5551 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5552 'ADDRESS_SYMBOLIC_TXT_REL'.
5553 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5554 (microblaze_legitimate_pic_operand): Exclude function calls from
5555 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5556 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5558 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5559 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5560 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5561 for 'address + offset'.
5562 (microblaze_expand_prologue): Add new function prologue call for
5564 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5565 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5566 table in case of TARGET_PIC_DATA_TEXT_REL.
5567 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5568 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5569 Add new macros 'UNSPEC_TEXT',
5570 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5571 + exclude function calls from 'UNSPEC_PLT' in case of data text
5573 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5574 new target hook for generating address diff vector tables in case of
5576 * doc/tm.texi : Regenerate.
5577 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5578 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5579 of addr diff vector generation.
5580 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5581 target hook definition.
5582 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5583 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5584 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5585 option for data text relative.
5587 2018-04-30 Richard Biener <rguenther@suse.de>
5589 * tree-chrec.h (evolution_function_is_constant_p): Remove
5591 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5593 2018-04-30 Richard Biener <rguenther@suse.de>
5596 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5598 2018-04-30 Richard Biener <rguenther@suse.de>
5600 PR tree-optimization/28364
5601 PR tree-optimization/85275
5602 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5603 copying first exit test.
5605 2018-04-28 Mark Wielaard <mark@klomp.org>
5607 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5609 (dwarf_AT): Handle DW_AT_addr_base.
5610 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5612 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5615 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5616 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5617 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5618 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5620 2018-04-28 Richard Biener <rguenther@suse.de>
5622 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5623 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5624 to reflect use. Only add interesting stmts.
5626 2018-04-27 Martin Jambor <mjambor@suse.cz>
5629 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5630 the jump function allows for passing through aggregate values.
5632 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5634 * input.h (in_system_header_at): Convert from macro to inline
5636 (from_macro_expansion_at): Likewise.
5637 (from_macro_definition_at): Likewise.
5639 2018-04-27 Jeff Law <law@redhat.com>
5641 * config.gcc: Mark tile* targets as deprecated/obsolete.
5643 2018-04-27 Richard Biener <rguenther@suse.de>
5645 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5648 2018-04-27 Richard Biener <rguenther@suse.de>
5650 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5652 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5654 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5655 with Yd constraint. Set "preferred_for_speed" attribute from
5656 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5658 (*movdi_internal): Ditto.
5659 (movti_interunit splitters): Remove
5660 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5661 (movdi_interunit splitters): Ditto.
5662 * config/i386/constraints.md (Ye): Remove.
5663 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5665 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5668 * config/aarch64/constraints.md (Usg): Limit to 31.
5671 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5673 PR tree-optimization/85529
5674 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5675 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5676 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5677 zero extension or masking of the MSB bit.
5678 (optimize_range_tests): Add FIRST_BB argument, pass it through
5679 to optimize_range_tests_var_bound.
5680 (maybe_optimize_range_tests, reassociate_bb): Adjust
5681 optimize_range_tests callers.
5683 2018-04-26 Richard Biener <rguenther@suse.de>
5684 Jakub Jelinek <jakub@redhat.com>
5686 * cgraph.h (symbol_table): Just declare debug method here.
5687 * symtab.c (symbol_table::debug): Define.
5689 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5691 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5693 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5695 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5696 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5697 (*movdi_internal): Substitute Yi and Yj constraint with x
5698 and Ym and Yn constraint with y constraint. Update "isa"
5699 attribute and set "preferred_for_speed" attribute from
5700 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5701 (*movsi_internal): Ditto.
5702 (*movdf_internal): Ditto.
5703 (*movsf_internal): Ditto.
5704 (*zero_extendsidi2): Ditto.
5705 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5706 (sse2_loadld): Ditto.
5707 (*vec_extract<ssevecmodelower>_0): Ditto.
5708 (*vec_extractv4si_0_zext_sse4): Ditto.
5709 (vec_concatv2di): Ditto.
5710 (*vec_dup<mode>): Ditto.
5711 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5712 * config/i386/constraints.md (Yi): Remove.
5717 2018-04-26 Nathan Sidwell <nathan@acm.org>
5719 * dumpfile.c (dump_open): New.
5720 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5721 (dump_finish): Detect stdio/stderr by value not name.
5723 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5725 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5727 2018-04-26 Tom de Vries <tom@codesourcery.com>
5730 * config/nvptx/nvptx.c (verify_neutering_jumps)
5731 (verify_neutering_labels): New function
5732 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5734 2018-04-26 Tom de Vries <tom@codesourcery.com>
5737 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5738 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5741 2018-04-26 Richard Biener <rguenther@suse.de>
5742 Tom de Vries <tom@codesourcery.com>
5745 * lto-streamer-out.c (output_function): Fixup loops if required to match
5746 discovery done in the reader.
5748 2018-04-26 Richard Biener <rguenther@suse.de>
5750 PR tree-optimization/85116
5751 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5752 have a loop exit from the single latch predecessor. Remove
5753 case of header with just condition.
5754 (ch_base::copy_headers): Exclude infinite loops from any
5756 (pass_ch::execute): Record exits.
5758 2018-04-26 Richard Biener <rguenther@suse.de>
5760 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5761 prologue cost vector and pass it to vect_get_load_cost.
5762 (vect_get_peeling_costs_all_drs): Likewise.
5763 (vect_peeling_hash_get_lowest_cost): Likewise.
5764 (vect_enhance_data_refs_alignment): Likewise.
5766 2018-04-26 Richard Biener <rguenther@suse.de>
5769 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5770 checking of integer<->pointer conversions.
5771 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5772 sign-/zero-extending pointer types.
5773 (expand_omp_for_static_chunk): Likewise.
5775 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5776 Jean Lee <xiaoyur347@gmail.com>
5778 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5779 (TARGET_ASAN_SHADOW_OFFSET): Define.
5780 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5781 true for -fsanitize=address.
5783 2018-04-25 Mark Wielaard <mark@klomp.org>
5785 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5788 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5790 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5791 than "alu", remove explicit "memory" and "imm_disp" attributes.
5792 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5795 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5796 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5797 gen_lowpart_no_emit.
5799 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5802 * config/i386/i386.c (ix86_expand_builtin): Change memory
5803 operand to XI, extend p0 to Pmode.
5804 * config/i386/i386.md: Change unspec volatile and operand
5805 1 mode to XI, change operand 0 mode to P.
5807 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5809 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5810 GET_MODE_MASK before any checking.
5811 (nds32_can_use_bset_p): Likewise.
5812 (nds32_can_use_btgl_p): Likewise.
5814 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5816 * config/nds32/nds32-doubleword.md: New define_split pattern for
5817 illegal register number.
5819 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5821 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5823 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5825 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5827 2018-04-25 Richard Biener <rguenther@suse.de>
5829 * lto-streamer.h (LTO_major_version): Bump to 8.
5831 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5833 * BASE-VER: Set to 9.0.0.
5835 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5837 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5838 in __abskf2 and __powikf2.
5840 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5843 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5844 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5845 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5846 Use the above on operand 2. Reindent.
5847 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5849 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5852 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5854 * config/i386/i386.opt (mcet): Removed.
5855 * doc/install.texi: Remove -mcet documentation.
5856 * doc/invoke.texi: Likewise.
5858 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5861 * doc/install.texi: Remove -mcet from bootstrap-cet.
5863 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5866 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5867 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5871 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5872 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5873 containing a CONST_VECTOR.
5875 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5877 * doc/install.texi: Update newlib dependency for nvptx.
5879 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5882 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5883 instead of INTVAL when shifting x left.
5885 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5887 PR tree-optimization/85478
5888 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5889 vect_grouped_store_supported for single element vectors.
5891 2018-04-24 Richard Biener <rguenther@suse.de>
5894 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5895 load cost increase to the case of non-constant step.
5897 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5900 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5901 destination if any_malformed_asm.
5903 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5906 * expr.c (store_field): In the bitfield case, if the value comes from
5907 a function call and is returned in registers by means of a PARALLEL,
5908 do not change the mode of the temporary unless BLKmode and VOIDmode.
5910 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5912 PR rtl-optimization/85423
5913 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5914 dependencies to debug insns when the previous insn is non-debug.
5916 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5918 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5919 enums into a single definition.
5920 (fls): Fix predicates and printing.
5923 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5925 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5926 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5927 and short u6 immediate.
5928 (check_if_valid_sleep_operand): Remove.
5929 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5931 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5933 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5934 flag_always_save_lp condition.
5935 * config/nds32/nds32.opt (malways-save-lp): New option.
5937 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5939 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5940 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5941 * config/nds32/nds32.h
5942 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5943 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5945 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5947 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5948 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5950 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5951 Chung-Ju Wu <jasonwucj@gmail.com>
5953 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5955 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5956 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5958 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5960 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5962 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5964 * config/nds32/nds32-protos.h (nds32_data_alignment,
5965 nds32_local_alignment): Declare.
5966 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5967 nds32_local_alignment): New functions.
5968 (TARGET_CONSTANT_ALIGNMENT): Define.
5969 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5971 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5973 * config/nds32/nds32.c
5974 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5975 (TARGET_MODES_TIEABLE_P): Likewise.
5977 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5979 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5982 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5983 Chung-Ju Wu <jasonwucj@gmail.com>
5985 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5986 for unaligned access.
5987 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5989 * config/nds32/nds32-intrinsic.md: Likewise.
5990 * config/nds32/nds32_intrinsic.h: Likewise.
5991 * config/nds32/nds32.h (nds32_builtins): Likewise.
5992 * config/nds32/nds32.opt (munaligned-access): New option.
5993 * config/nds32/nds32.c (nds32_asm_file_start): Display
5994 flag_unaligned_access status.
5996 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5998 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5999 -mno-relax is present.
6000 * config/riscv/linux.h (LINK_SPEC): Ditto.
6002 2018-04-20 Martin Sebor <msebor@redhat.com>
6005 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6007 (gimple_fold_builtin_stxcpy_chk): Same.
6008 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6010 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6013 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6014 __powikf2 when long double is IEEE 128-bit.
6016 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6018 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6019 step to make sure stack always aligned.
6021 2018-04-20 Carl Love <cel@us.ibm.com>
6024 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6025 size check for arg0.
6027 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6028 Tom de Vries <tom@codesourcery.com>
6031 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6032 Emit insns for calls too.
6033 (nvptx_find_par): Always look for worker-level predecessor insn.
6034 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6036 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6037 (nvptx_process_pars): Propagate frames for calls.
6039 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6042 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6044 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6045 (ix86_handle_option): Don't handle OPT_mibt.
6046 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6048 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6050 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6051 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6052 (ix86_target_macros): Define __CET__ with flag_cf_protection
6053 for -fcf-protection.
6054 * config/i386/i386.c (isa2_opts): Remove -mibt.
6055 * config/i386/i386.h (TARGET_IBT): Removed.
6056 (TARGET_IBT_P): Likewise.
6057 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6058 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6059 * config/i386/i386.opt (mcet): Update help message.
6062 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6063 -mcet as an alias for -mshstk.
6065 2018-04-20 Richard Biener <rguenther@suse.de>
6068 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6069 complexity by forcing a single use of the multiply operand.
6071 2018-04-20 Martin Jambor <mjambor@suse.cz>
6074 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6075 recursion dependency to only apply to non-clones.
6077 2018-04-20 Martin Jambor <mjambor@suse.cz>
6080 * ipa-cp.c (create_specialized_node): Check that clones of
6081 self-recursive edges exist during IPA-CP.
6083 2018-04-19 Toon Moene <toon@moene.org>
6085 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6088 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6090 PR tree-optimization/85467
6091 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6092 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6093 VECTOR_CST element to type.
6095 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6098 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6099 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6100 (builtin_longjmp): Likewise.
6101 (save_stack_nonlocal): New pattern.
6102 (restore_stack_nonlocal): Likewise.
6104 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6107 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6108 Replace ASM_OUTPUT_LABEL with fprintf.
6110 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6113 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6114 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6115 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6116 define __IBT__ and __SHSTK__ for -fcf-protection.
6117 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6119 (ix86_trampoline_init): Likewise.
6120 (x86_output_mi_thunk): Likewise.
6121 (ix86_notrack_prefixed_insn_p): Likewise.
6122 (ix86_option_override_internal): Don't disallow -fcf-protection.
6123 * config/i386/i386.md (rdssp<mode>): Also enable for
6125 (incssp<mode>): Likewise.
6126 (nop_endbr): Likewise.
6127 * config/i386/i386.opt (mcet): Change help message to built-in
6131 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6132 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6133 enable CET built-in functions.
6135 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6137 * common/config/i386/i386-common.c
6138 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6139 OPTION_MASK_ISA_MOVDIRI_UNSET,
6140 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6141 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6142 * config.gcc (movdirintrin.h): New header.
6143 * config/i386/cpuid.h (bit_MOVDIRI,
6144 bit_MOVDIR64B): New bits.
6145 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6147 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6148 (VOID, PVOID, PCVOID)): New function types.
6149 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6150 __builtin_ia32_directstoreu_u64,
6151 __builtin_ia32_movdir64b): New builtins.
6152 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6153 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6155 (ix86_valid_target_attribute_inner_p): Ditto.
6156 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6157 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6158 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6159 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6160 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6161 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6162 (movdiri<mode>, movdir64b_<mode>): New.
6163 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6164 * config/i386/immintrin.h: Include movdirintrin.h.
6165 * config/i386/movdirintrin.h: New file.
6166 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6168 2018-04-19 Richard Biener <rguenther@suse.de>
6171 * cfg.c (clear_bb_flags): When loop state says we have
6172 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6174 2018-04-19 Richard Biener <rguenther@suse.de>
6176 PR tree-optimization/84737
6177 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6178 copying restrict info.
6179 (vect_setup_realignment): Use it.
6180 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6181 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6182 the first DR to all generated stores.
6183 (vectorizable_load): Likewise for loads.
6185 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6187 PR tree-optimization/85446
6188 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6189 the integral and pointer types to have the same precision.
6191 * doc/install.texi: Document --disable-cet being the default and
6194 2018-04-18 Martin Liska <mliska@suse.cz>
6196 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6199 2018-04-18 Martin Liska <mliska@suse.cz>
6202 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6205 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6206 arguments if they are comparable.
6208 2018-04-18 Martin Liska <mliska@suse.cz>
6211 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6214 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6217 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6220 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6221 ENDBR after calling __morestack.
6223 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6226 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6227 by using gcc_base_ver to generate a gcc_driver_version, and use
6228 it when generating GCC_DRIVER_NAME.
6229 * configure: Regenerate.
6231 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6234 * config.gcc: Obsolete powerpc*-*-*spe*.
6236 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6239 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6240 (stabstr_D): Change type of unum from unsigned int to
6241 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6244 2018-04-17 Jim Wilson <jimw@sifive.com>
6247 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6248 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6249 Set arg_pointer_offset after using pretend_args_size.
6251 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6253 PR rtl-optimization/85431
6254 * dse.c (record_store): Ignore zero width stores.
6257 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6258 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6259 __builtin_stack_restore rather than after it.
6260 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6261 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6262 argument instead of virtual_dynamic_stack_rtx.
6264 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6266 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6268 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6269 Add note to error message to explain internal mapping of overloaded
6270 built-in function name to non-overloaded built-in function name.
6271 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6274 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6277 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6278 where the inputs overlap with the output.
6280 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6282 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6283 (=v, v) alternative and explicit "memory" attribute.
6284 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6285 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6287 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6288 "sselog1" type instead of "sselog".
6289 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6290 "sselog". Remove explicit "memory" attribute.
6291 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6292 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6294 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6295 "sselog1" type instead of "sselog". Remove explicit "memory"
6297 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6298 use "sselog1" type instead of "sselog". Remove explicit "memory"
6300 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6301 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6303 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6304 "sselog1" type instead of "sselog". Remove explicit "memory"
6306 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6307 use "sselog1" type instead of "sselog". Remove explicit "memory"
6311 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6314 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6317 2018-04-17 Martin Jambor <mjambor@suse.cz>
6320 * ipa-cp.c (create_specialized_node): Call
6321 expand_all_artificial_thunks if necessary.
6323 2018-04-17 Martin Liska <mliska@suse.cz>
6326 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6327 in message, remote space in between '_G' and '('.
6329 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6332 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6333 avx512f_vmcmp<mode>3<round_saeonly_name>,
6334 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6335 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6336 avx512f_rndscale<mode><round_saeonly_name>,
6337 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6338 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6339 Use %<iptr>2 instead of %2 for -masm=intel.
6340 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6341 avx512f_vcvttss2usi<round_saeonly_name>,
6342 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6344 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6345 avx512f_vcvttsd2usi<round_saeonly_name>,
6346 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6347 Use %q1 instead of %1 for -masm=intel.
6348 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6349 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6350 of %3 for -masm=intel.
6351 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6353 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6355 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6357 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6358 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6360 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6362 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6363 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6364 %g1 and one with %0 and %1.
6365 (avx512er_vmrcp28<mode><round_saeonly_name>,
6366 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6368 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6369 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6370 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6371 of %0 and %{%4%} for -masm=intel.
6372 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6373 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6374 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6375 order of %0 and %{%5%}%{z%} for -masm=intel.
6377 2018-04-17 Jan Hubicka <jh@suse.cz>
6380 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6382 2018-04-17 Martin Liska <mliska@suse.cz>
6385 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6386 for target_clone error message. Make default implementation
6387 clone to be a local declaration.
6388 (separate_attrs): Add new argument and check for an empty
6390 (expand_target_clones): Handle it.
6391 (ipa_target_clone): Make redirection just for target_clones
6394 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6395 Tom de Vries <tom@codesourcery.com>
6398 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6399 tiled basic blocks without omp continue statements.
6401 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6404 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6405 vec_extract expression as having side effects to make sure it gets
6408 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6411 * config/i386/i386.c (get_builtin_code_for_version): Check
6414 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6417 * gcc/config.gcc: Support "skylake".
6418 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6420 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6421 (processor_target_table): Add "skylake".
6422 (ix86_option_override_internal): Add "skylake".
6423 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6424 PROCESSOR_CANNONLAKE.
6425 (get_builtin_code_for_version): Fix priority for
6426 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6427 PROCESSOR_SKYLAKE-AVX512.
6428 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6429 (processor_type): Add PROCESSOR_SKYLAKE.
6431 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6432 Jason Merrill <jason@redhat.com>
6435 * convert.c (convert_to_integer_1): Use direct recursion for
6436 enumeral types and types with a precision less than the number
6437 of bits in their mode.
6439 2018-04-16 Julia Koval <julia.koval@intel.com>
6442 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6443 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6445 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6448 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6449 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6450 and -mno-direct-move.
6452 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6455 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6456 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6457 Ensure negative shifts result in {0}.
6459 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6461 PR rtl-optimization/79916
6462 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6463 regs (if any) to define how to gnerate SD moves when LRA is in
6466 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6468 PR rtl-optimization/85393
6469 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6470 * except.c (expand_dw2_landing_pad_for_region): Make static.
6471 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6472 a label and unconditional jump to old_bb, rather than
6473 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6476 PR rtl-optimization/85376
6477 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6478 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6479 instead of a specific value.
6481 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6482 Bin Cheng <bin.cheng@arm.com>
6484 PR tree-optimization/82965
6485 PR tree-optimization/83991
6486 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6487 by_profile_only parameter.
6488 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6489 information if the loop was predicted to iterate too many times.
6490 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6492 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6495 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6498 2018-04-13 Martin Liska <mliska@suse.cz>
6499 Jakub Jelinek <jakub@redhat.com>
6502 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6503 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6504 * builtins.c (expand_builtin_memory_copy_args): Use
6505 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6506 handle dest_addr == pc_rtx.
6508 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6511 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6512 asked to not generate direct moves.
6513 (fix_trunc<mode>si2_stfiwx): Similar.
6514 (fix_trunc<mode>si2_internal): Similar.
6516 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6519 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6520 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6521 lookup if dest in some wider mode is known to be const0_rtx and
6522 if so, record permanent equivalence for it to be ZERO_EXTEND of
6523 the narrower mode destination.
6525 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6527 * lto-streamer-out.c (output_function): Revert 259346.
6528 * omp-expand.c (expand_oacc_for): Likewise.
6530 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6532 PR rtl-optimization/85354
6533 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6534 * sel-sched.c (sel_global_init): ... here.
6536 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6539 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6540 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6541 mode for PE-COFF targets.
6542 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6543 (i386_pe_asm_lto_end): Likewise.
6544 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6545 (TARGET_ASM_LTO_END): Likewise.
6546 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6547 (i386_pe_asm_lto_start): New function.
6548 (i386_pe_asm_lto_end): Likewise.
6550 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6551 Richard Biener <rguenther@suse.de>
6554 * lto-streamer-out.c (output_function): Fix CFG loop state before
6556 * omp-expand.c (expand_oacc_for): Handle calls to internal
6557 functions like regular functions.
6559 2018-04-12 Richard Biener <rguenther@suse.de>
6562 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6563 for the early LTO debug to properly generate references to it
6564 during DIE emission. Do not re-use that for the skeleton for
6566 (dwarf2out_early_finish): Likewise.
6568 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6571 * config/i386/sse.md
6572 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6573 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6574 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6575 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6576 and output is a reg, avoid creating invalid lowpart subreg, but
6577 instead split into a 512-bit move. Don't split if not AVX512VL,
6578 input is xmm16+ reg and output is a mem.
6579 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6580 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6581 xmm16+ reg and output is a mem.
6583 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6585 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6586 also for flag_dwarf2_cfi_asm.
6588 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6590 PR rtl-optimization/85342
6591 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6592 a bool scalar var inside of the loop instead. Don't try to update
6593 recog_data.operand after failed apply_change_group.
6595 2018-04-12 Tom de Vries <tom@codesourcery.com>
6598 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6599 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6600 array with flexible array member as array without given dimension.
6601 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6602 argument for undefined param to true.
6604 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6607 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6608 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6609 from PowerPC section.
6610 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6611 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6612 help text that is too long.
6613 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6614 help text that is too long.
6615 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6616 help text that is too long.
6618 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6620 * config/alpha/alpha.md (stack_probe_internal): Rename
6621 from "probe_stack". Update all callers.
6623 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6625 PR rtl-optimization/84566
6626 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6627 sched_macro_fuse_insns.
6629 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6632 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6633 (compute_block_dependences): ... from here.
6635 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6637 PR tree-optimization/85331
6638 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6639 from int to HOST_WIDE_INT.
6641 2018-04-11 Martin Jambor <mjambor@suse.cz>
6644 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6645 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6646 not the same as the source val.
6647 (cgraph_edge_brings_value_p): New parameter.
6648 (gather_edges_for_value): Pass destination value to
6649 cgraph_edge_brings_value_p.
6650 (perhaps_add_new_callers): Likewise.
6651 (get_info_about_necessary_edges): Likewise and exclude values brought
6652 only by self-recursive edges.
6653 (create_specialized_node): Redirect only clones of self-calling edges.
6654 (+self_recursive_pass_through_p): New function.
6655 (find_more_scalar_values_for_callers_subset): Use it.
6656 (find_aggregate_values_for_callers_subset): Likewise.
6657 (known_aggs_to_agg_replacement_list): Removed.
6658 (decide_whether_version_node): Re-calculate known constants for all
6659 remaining context clones.
6661 2018-04-11 Richard Biener <rguenther@suse.de>
6664 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6665 from early DWARF output.
6666 (dwarf2out_early_finish): Output line info unconditionally into
6667 early DWARF and add reference to it.
6669 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6672 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6673 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6674 other than V2DFmode using iptr mode attribute.
6675 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6677 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6679 PR rtl-optimization/84659
6680 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6682 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6685 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6687 (output_loc_list): Pass address of a dummy size variable even in the
6688 locview handling loop.
6689 (index_location_lists): Add comment on why skip_loc_list_entry can't
6692 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6695 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6698 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6701 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6702 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6703 and -mstring-compare-inline-limit.
6705 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6708 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6709 for stack clash protection in a register whenever we need it to be in
6712 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6714 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6715 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6717 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6720 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6722 (mlong-double-): Ditto.
6723 * config/rs6000/sysv4.opt (msdata=): Ditto.
6724 (mtls-size=): Ditto.
6726 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6728 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6729 erroneous entries for
6730 "vector int vec_ldl (int, long int *)", and
6731 "vector unsigned int vec_ldl (int, unsigned long int *)".
6732 Add comments and entries for
6733 "vector bool char vec_ldl (int, bool char *)",
6734 "vector bool short vec_ldl (int, bool short *)",
6735 "vector bool int vec_ldl (int, bool int *)",
6736 "vector bool long long vec_ldl (int, bool long long *)",
6737 "vector pixel vec_ldl (int, pixel *)",
6738 "vector long long vec_ldl (int, long long *)",
6739 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6740 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6741 type tree bool_long_long_type_node and correct definition of
6742 bool_V2DI_type_node to make reference to this new type tree.
6743 (rs6000_mangle_type): Replace erroneous reference to
6744 bool_long_type_node with bool_long_long_type_node.
6745 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6746 comments to emphasize sign distinctions for char and int types and
6747 replace RS6000_BTI_bool_long constant with
6748 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6749 use of RS6000_BTI_pixel.
6750 (bool_long_type_node): Remove this macro definition.
6751 (bool_long_long_type_node): New macro definition
6753 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6755 PR rtl-optimization/85300
6756 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6757 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6758 simplify_unary_operation fails.
6760 2018-04-10 Martin Liska <mliska@suse.cz>
6762 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6763 cgraph_edge and ipa_ref.
6765 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6769 * config/i386/sse.md
6770 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6771 computation of the VEC_MERGE selector from mask.
6772 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6773 Fix decoding of the VEC_MERGE selector into mask.
6775 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6777 PR tree-optimization/85286
6778 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6780 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6782 * final.c (final_1): Set insn_last_address as well as
6783 insn_current_address.
6785 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6788 * explow.c (emit_stack_probe): Call validize_mem on memory location
6789 before passing it to gen_probe_stack. Create address operand and
6790 legitimize it for the probe_stack_address case.
6792 2018-04-09 Jan Hubicka <jh@suse.cz>
6795 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6796 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6797 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6799 (free_lang_data): Rebuild type inheritance graph.
6801 2018-04-09 Martin Sebor <msebor@redhat.com>
6803 * invoke.texi (-finline-small-functions): Mention other optimization
6805 (-findirect-inlining, -fpartial-inlining): Same.
6806 (-finline-functions-called-once): Same.
6807 (-freorder-blocks-and-partition): Same.
6809 2018-04-09 Jan Hubicka <jh@suse.cz>
6812 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6813 jumps; choose last target that matches the criteria (i.e.
6814 no partition changes for non-crossing jumps).
6815 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6816 support for redirecting crossing jumps to non-crossing.
6818 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6820 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6821 also for naked functions.
6823 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6825 * config/arc/arc.md (add_shift): New pattern.
6826 (add_shift2): Likewise.
6827 (sub_shift): Likewise.
6828 (sub_shift_cmp0_noout): Likewise.
6829 (compare_si_ashiftsi): Likewise.
6830 (xbfu_cmp0_noout): New combine pattern.
6831 (xbfu_cmp0"): Likewise.
6832 (movsi_set_cc_insn): Place the predicable variant first.
6833 (commutative_binary_cmp0_noout): Remove clobber.
6834 (commutative_binary_cmp0): New pattern.
6835 (noncommutative_binary_cmp0): Likewise.
6836 (noncommutative_binary_cmp0_noout): Likewise.
6837 (noncommutative_binary_comparison_result_used): Removed.
6838 (rsub_cmp0): New pattern.
6839 (rsub_cmp0_noout): Likewise.
6840 (extzvsi): Changed, keep only meaningful variants.
6841 (SQH, SEZ): New iterators.
6842 (SQH_postfix): New mode attribute.
6843 (SEZ_prefix): New code attribute.
6844 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6845 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6846 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6848 (noncommutative_operator): Check the availability of barrel
6851 2018-04-09 Richard Biener <rguenther@suse.de>
6853 PR tree-optimization/85284
6854 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6855 Only use the niter constraining form of simple_iv when the exit
6858 2018-04-09 Tom de Vries <tom@codesourcery.com>
6861 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6862 (define_expand "*memory_barrier"): New define_expand.
6863 (define_insn "memory_barrier"): New insn.
6865 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6867 PR rtl-optimization/80463
6868 PR rtl-optimization/83972
6869 PR rtl-optimization/83480
6871 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6872 correct producer for the insn.
6873 (tidy_control_flow): Fixup seqnos in case of debug insns.
6875 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6877 PR rtl-optimization/83913
6879 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6880 different sched-times when merging exprs.
6882 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6884 PR rtl-optimization/83962
6886 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6887 tidy_fallthru_edge and tidy_control_flow.
6889 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6891 PR rtl-optimization/83530
6893 * sel-sched.c (force_next_insn): New global variable.
6894 (remove_insn_for_debug): When force_next_insn is true, also leave only
6895 next insn in the ready list.
6896 (sel_sched_region): When the region wasn't scheduled, make another pass
6897 over it with force_next_insn set to 1.
6899 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6901 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6903 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6904 for interrupt control.
6905 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6906 functions for interrupt control.
6907 * config/nds32/nds32-intrinsic.md: Likewise.
6908 * config/nds32/nds32_intrinsic.h: Likewise.
6909 * config/nds32/nds32.h (nds32_builtins): Likewise.
6911 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6913 * config/nds32/nds32.c (nds32_init_machine_status,
6914 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6915 strict_aligned_p field.
6916 (nds32_expand_to_rtl_hook): New function.
6917 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6918 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6920 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6921 Chung-Ju Wu <jasonwucj@gmail.com>
6923 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6924 * config/nds32/nds32-n7.md: New file.
6925 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6926 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6928 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6929 * config/nds32/nds32.md (pipeline_model): Add n7.
6930 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6931 * config/nds32/pipelines.md: Include n7 settings.
6933 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6934 Chung-Ju Wu <jasonwucj@gmail.com>
6936 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6937 * config/nds32/nds32-e8.md: New file.
6938 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6939 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6941 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6942 * config/nds32/nds32.md (pipeline_model): Add e8.
6943 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6944 * config/nds32/pipelines.md: Include e8 settings.
6946 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6947 Chung-Ju Wu <jasonwucj@gmail.com>
6949 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6950 * config/nds32/nds32-n8.md: New file.
6951 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6952 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6954 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6955 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6956 * config/nds32/nds32.md (pipeline_model): Add n8.
6957 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6958 * config/nds32/pipelines.md: Include n8 settings.
6960 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6961 Chung-Ju Wu <jasonwucj@gmail.com>
6963 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6964 * config/nds32/nds32-n9-2r1w.md: New file.
6965 * config/nds32/nds32-n9-3r2w.md: New file.
6966 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6967 nds32_register_ports): New or modify for cpu n9.
6968 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6970 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6971 * config/nds32/nds32-utils.c: New file.
6972 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6973 TARGET_MUL_SLOW): Define.
6974 * config/nds32/nds32.md (pipeline_model): New attribute.
6975 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6976 New options that support cpu n9.
6977 * config/nds32/pipelines.md: Include n9 settings.
6978 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6980 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6982 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6983 information if necessary.
6984 (output_cond_branch_compare_zero): Likewise.
6985 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6986 (nds32_target_alignment): Refine for alignment.
6987 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6988 (FUNCTION_BOUNDARY): Modify.
6989 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6991 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6993 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6995 * config/nds32/constants.md (unspec_volatile_element): Add values for
6996 TLB operation and data prefetch.
6997 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6998 functions for TLB operation and data prefetch.
6999 * config/nds32/nds32-intrinsic.md: Likewise.
7000 * config/nds32/nds32_intrinsic.h: Likewise.
7001 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7002 (nds32_print_operand): Likewise.
7003 * config/nds32/nds32.h (nds32_builtins): Likewise.
7005 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7006 Andrew Pinski <pinsika@gcc.gnu.org>
7009 * match.pd: Use constant_boolean_node of correct type instead of
7010 boolean_true_node or boolean_false_node for simplifying
7011 pointer comparisons to zero.
7013 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7015 PR tree-optimization/80021
7016 * tree.c (verify_type_variant): Make error call in verify_variant_match
7017 translatable and remove final full stop.
7019 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7021 * config/nds32/constants.md (unspec_volatile_element): Add
7022 UNSPEC_VOLATILE_EH_RETURN.
7023 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7024 nds32_output_stack_pop): Support dwarf exception handling process.
7025 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7026 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7027 exception handling process.
7028 (nds32_compute_stack_frame): Likewise.
7029 (nds32_return_addr_rtx): Likewise.
7030 (nds32_initial_elimination_offset): Likewise.
7031 (nds32_expand_prologue): Likewise.
7032 (nds32_expand_epilogue): Likewise.
7033 (nds32_dynamic_chain_address): New function.
7034 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7036 (DYNAMIC_CHAIN_ADDRESS): Define.
7037 (EH_RETURN_DATA_REGNO): Define.
7038 (EH_RETURN_STACKADJ_RTX): Define.
7039 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7040 patterns for dwarf exception handling.
7042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7044 * config/nds32/nds32.h: Clean up obsolete macros.
7046 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7048 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7049 Add enum values for particular instructions.
7050 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7051 particular intrinsic functions.
7052 * config/nds32/nds32-intrinsic.md: Likewise.
7053 * config/nds32/nds32_intrinsic.h: Likewise.
7054 * config/nds32/nds32.h (nds32_builtins): Likewise.
7055 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7056 (btst, ave): New patterns for particular instructions.
7058 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7060 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7061 Add enum values for atomic load/store and memory sync.
7062 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7064 * config/nds32/nds32-intrinsic.md: Likewise.
7065 * config/nds32/nds32_intrinsic.h: Likewise.
7066 * config/nds32/nds32.h (nds32_builtins): Likewise.
7068 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7070 PR tree-optimization/85257
7071 * fold-const.c (native_encode_vector): If not all elts could fit
7072 and off is -1, return 0 rather than offset.
7073 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7074 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7075 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7076 adjust buffer in native_interpret_expr call.
7078 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7080 * config/nds32/constants.md (unspec_volatile_element): Add cache
7081 control enum values.
7082 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7083 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7084 * config/nds32/nds32.c (nds32_cctl_names): New.
7085 (nds32_print_operand): Handle cache control register names.
7086 * config/nds32/nds32.h (nds32_builtins): New enum values.
7087 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7089 * config/nds32/nds32.md (type): Add mmu.
7090 * config/nds32/pipelines.md (simple_insn): Add mmu.
7092 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7094 * config/nds32/nds32.md (type): Remove call.
7095 * config/nds32/pipelines.md (simple_insn): Likewise.
7097 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7099 * config/nds32/constants.md (unspec_volatile_element): Add
7100 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7101 UNSPEC_VOLATILE_FMFCFG.
7102 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7103 description for fmfcfg and fmfcsr.
7104 (bdesc_1arg): Add fmtcsr.
7105 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7106 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7107 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7108 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7109 unspec_fmfcfg): New patterns.
7110 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7111 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7112 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7113 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7114 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7115 __nds32__fmfcfg): Define.
7117 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7119 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7120 intrinsic register names.
7121 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7122 intrinsic register enum values and macros.
7124 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7126 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7127 for load/store addressing form.
7128 (nds32_print_operand_address): Likewise.
7130 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7133 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7134 based on LABEL_REF. Remove useless assertion.
7135 (pic_address_needs_scratch): Fix formatting.
7136 (sparc_legitimize_pic_address): Minor tweaks.
7137 (sparc_delegitimize_address): Adjust assertion accordingly.
7138 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7139 into symbolic_operand.
7140 (movsi_high_pic_label_ref): Likewise.
7141 (movsi_lo_sum_pic_label_ref): Likewise.
7142 (movdi_pic_label_ref): Likewise.
7143 (movdi_high_pic_label_ref): Likewise.
7144 (movdi_lo_sum_pic_label_ref): Likewise.
7146 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7148 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7149 custom LIB_SPEC setup.
7151 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7152 Kito Cheng <kito.cheng@gmail.com>
7154 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7155 * config/riscv/freebsd.h: New.
7157 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7159 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7160 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7163 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7164 Kito Cheng <kito.cheng@gmail.com>
7166 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7167 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7168 * config/nds32/nds32-protos.h (nds32_output_call,
7169 nds32_output_return): Declare.
7170 * config/nds32/nds32.md: Refine all the call and return patterns.
7172 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7175 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7176 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7178 PR rtl-optimization/84872
7179 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7180 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7183 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7185 * expr.c (copy_blkmode_to_reg): Revert 254862.
7186 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7188 2018-04-06 Richard Biener <rguenther@suse.de>
7191 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7192 after seeing a component reference with an adjacent field. Treat
7193 refs to arrays at struct end of external decls similar to
7194 refs to unconstrained commons.
7196 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7199 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7200 look through SAVE_EXPRs with non-side-effects argument. Adjust
7202 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7205 2018-04-06 Richard Biener <rguenther@suse.de>
7208 * alias.c (find_base_term): New wrapper around find_base_term
7209 unwinding CSELIB_VAL_PTR changes.
7210 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7213 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7215 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7217 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7218 constant definitions.
7219 ("nop"): lr 0,0 -> nopr r0
7220 ("nop_lr0", "nop_lr1"): New insn definitions.
7222 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7224 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7225 NDS32_V3PUSH_AVAILABLE_P macro.
7227 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7228 Chung-Ju Wu <jasonwucj@gmail.com>
7230 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7231 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7232 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7233 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7234 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7235 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7236 * config/nds32/constraints.md: New constraints and checking for hard
7237 float configuration.
7238 * config/nds32/iterators.md: New mode iterator and attribute for hard
7239 float configuration.
7240 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7242 * config/nds32/nds32-fpu.md: New file.
7243 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7244 deal with hard float code generation.
7245 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7247 (abi_type, float_reg_number): New enum type.
7248 * config/nds32/nds32-predicates.c: New predicates for hard float.
7249 * config/nds32/nds32-protos.h: Declare functions for hard float.
7250 * config/nds32/nds32.c: Implementation for hard float configuration.
7251 * config/nds32/nds32.h: Definitions for hard float configuration.
7252 * config/nds32/nds32.md: Include hard float machine description and
7253 modify patterns for hard float configuration.
7254 * config/nds32/nds32.opt: New options for hard float configuration.
7255 * config/nds32/predicates.md: New predicates for hard float
7258 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7260 * common/config/nds32/nds32-common.c
7261 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7263 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7266 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7267 CONSTRUCTOR_ELT (ctor, ...)->value.
7269 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7272 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7274 2018-04-05 Tom de Vries <tom@codesourcery.com>
7277 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7280 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7281 Kito Cheng <kito.cheng@gmail.com>
7283 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7284 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7285 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7286 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7288 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7289 Kito Cheng <kito.cheng@gmail.com>
7291 * config/nds32/constraints.md (Ufe): New memory constraint.
7292 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7293 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7294 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7296 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7297 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7299 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7301 * config/nds32/nds32.md: Use optimize_size in the condition for
7302 alu-shift instructions.
7304 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7306 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7308 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7310 * config/nds32/nds32.md (negsi2): Refine pattern.
7312 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7313 Chung-Ju Wu <jasonwucj@gmail.com>
7315 * config/nds32/iterators.md (shift_rotate): New code iterator.
7316 (shift): New code attribute.
7317 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7318 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7319 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7320 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7321 bit-wise operations.
7322 (andsi3, *andsi3): Ditto.
7323 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7324 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7325 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7326 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7327 nds32_ior_operand, nds32_xor_operand): New predicates.
7329 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7331 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7332 (addsi3, subsi3): ... this.
7334 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7336 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7338 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7340 * config/nds32/nds32.md: Adjust indention.
7342 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7344 * config/nds32/nds32.md (feature): New attribute.
7346 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7348 * config/nds32/nds32.md (subtype): New attribute.
7350 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7353 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7354 expansion to perform a bitwise AND of the argument followed by a
7355 boolean negation of the result.
7357 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7359 PR rtl-optimization/84878
7360 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7361 the basic block. Assert the use reference is not artificial and that
7362 it has an associated insn.
7364 2018-04-04 Michael Matz <matz@suse.de>
7366 * builtins.c (compute_objsize): Pass correct operand
7367 to array_at_struct_end_p.
7369 2018-04-04 Richard Biener <rguenther@suse.de>
7372 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7373 from contexts for DINFO_LEVEL_TERSE and below.
7375 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7377 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7378 resiter_operand condition.
7379 * config/nds32/nds32.md (*move<mode>): Ditto.
7381 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7382 Monk Chiang <sh.chiang04@gmail.com>
7384 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7386 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7388 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
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_inverse_cond_code,
7394 nds32_cond_code_str, output_cond_branch,
7395 output_cond_branch_compare_zero, nds32_expand_cbranch,
7396 nds32_expand_cstore, nds32_expand_movcc,
7397 nds32_output_cbranchsi4_equality_zero,
7398 nds32_output_cbranchsi4_equality_reg,
7399 nds32_output_cbranchsi4_equality_reg_or_const_int,
7400 nds32_output_cbranchsi4_greater_less_zero: New functions.
7401 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7402 nds32_expand_cstore, nds32_expand_movcc,
7403 nds32_output_cbranchsi4_equality_zero,
7404 nds32_output_cbranchsi4_equality_reg,
7405 nds32_output_cbranchsi4_equality_reg_or_const_int,
7406 nds32_output_cbranchsi4_greater_less_zero): Declare.
7407 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7408 nds32_rimm11s_operand): New predicates.
7409 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7410 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7413 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7415 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7416 * config/nds32/nds32.md: Ditto.
7417 * config/nds32/pipelines.md: Ditto.
7419 2018-04-04 Richard Biener <rguenther@suse.de>
7421 PR tree-optimization/85168
7422 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7423 propagating abnormals.
7425 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7427 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7429 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7430 Kito Cheng <kito.cheng@gmail.com>
7432 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7433 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7434 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7435 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7436 * config/nds32/nds32.md (sibcall_internal): New.
7437 (sibcall_register): Remove.
7438 (sibcall_immediate): Remove.
7439 (sibcall_value_internal): New.
7440 (sibcall_value_register): Remove.
7441 (sibcall_value_immediate): Remove.
7442 * config/nds32/predicates.md (nds32_general_register_operand): New.
7443 (nds32_call_address_operand): New.
7445 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7447 PR rtl-optimization/85167
7448 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7449 bb_defs if *split_p, instead preinitialize it to NULL.
7451 PR tree-optimization/85156
7452 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7453 evaluating the argument multiple times.
7455 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7457 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7459 (_mm_cvtpd_ps): Likewise.
7460 (_mm_cvttpd_epi32): Likewise.
7461 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7462 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7463 vector, pixel, and bool following altivec.h include.
7465 2018-04-03 Martin Sebor <msebor@redhat.com>
7467 * doc/extend.texi (Common Function Attributes): Clarify.
7468 (const attribute): Likewise.
7469 (pure attribute): Likewise.
7471 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7474 * config/i386/i386.c (ix86_expand_vector_set): Use
7475 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7477 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7479 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7480 instructions when changing rounding bits to preserve precision bits
7481 in the x87 control word.
7483 2018-04-03 Martin Liska <mliska@suse.cz>
7485 PR tree-optimization/82491
7486 * rtl.h (strip_offset_and_add): Replace += suboffset with
7487 poly_uint64 () + suboffset.
7489 2018-03-29 Martin Liska <mliska@suse.cz>
7490 Martin Jambor <mjambor@suse.cz>
7493 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7494 param_type is not an integral or pointer type.
7496 2018-04-03 Richard Biener <rguenther@suse.de>
7498 * sese.h (recompute_all_dominators): Remove.
7500 2018-04-02 Martin Sebor <msebor@redhat.com>
7502 * doc/invoke.texi (-Wrestrict): Fix typos.
7504 2018-04-02 Jim Wilson <jimw@sifive.com>
7506 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7507 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7508 (<optab>di3, <optab>si3_extend): Likewise.
7509 (<optab>si3_mask, <optab>si3_mask_1): New.
7510 (<optab>di3_mask, <optab>di3_mask_1): New.
7511 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7512 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7513 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7515 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7517 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7520 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7522 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7523 (nds32_canonicalize_comparison): New function.
7525 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7526 Kito Cheng <kito.cheng@gmail.com>
7527 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7529 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7530 * config/nds32/constants.md (unspec_volatile_element): Add
7531 UNSPEC_VOLATILE_RELAX_GROUP.
7532 * config/nds32/nds32-relax-opt.c: New file.
7533 * config/nds32/nds32-predicates.c
7534 (nds32_symbol_load_store_p): New function.
7535 * config/nds32/nds32-protos.h
7536 (nds32_symbol_load_store_p): Declare function.
7537 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7538 * config/nds32/nds32.c
7539 (nds32_register_pass): New function to register pass.
7540 (nds32_register_passes): New function to register passes.
7541 * config/nds32/nds32.md (relax_group): New pattern.
7542 * config/nds32/nds32.opt (mrelax-hint): New option.
7543 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7545 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7547 * config/nds32/t-nds32: Modify files dependency.
7549 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7551 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7552 (PROFILE_HOOK): Define its implementation.
7554 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7556 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7557 type and 32-bit size.
7559 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7562 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7563 (V_128_256): New mode iterator.
7564 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7565 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7566 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7568 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7569 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7571 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7574 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7575 NaN inputs correctly.
7577 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7580 * config/rs6000/vsx.md (??r): New mode attribute.
7581 (*vsx_mov<mode>_64bit): Use it.
7582 (*vsx_mov<mode>_32bit): Likewise.
7584 2018-03-30 Martin Sebor <msebor@redhat.com>
7586 PR tree-optimization/84818
7587 * builtins.c (check_access): Use warning_n.
7589 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7592 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7594 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7597 2018-03-30 Julia Koval <julia.koval@intel.com>
7600 * x86-tune.def (movx, partial_reg_dependency): Enable for
7603 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7606 * lra-constraints.c (process_alt_operands): Move setting
7607 this_alternative_matches below.
7609 2018-03-29 Martin Liska <mliska@suse.cz>
7612 * doc/invoke.texi: Document how LTO works with debug info.
7613 Describe auto-load support of binutils. Mention 'x86-64'
7614 as valid option value of -march option.
7616 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7618 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7621 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7622 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7623 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7626 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7629 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7630 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7631 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7632 for RS6000_BTM_POWERPC64.
7633 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7634 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7635 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7640 2018-03-28 Carl Love <cel@us.ibm.com>
7643 2017-09-27 Carl Love <cel@us.ibm.com>
7645 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7646 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7647 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7650 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7652 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7653 instead of __vector bool.
7654 (_mm_max_pu8): Likewise.
7655 (_mm_min_pi16): Likewise.
7657 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7660 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7661 (DIVWEUO): Likewise.
7663 (DIVDEUO): Likewise.
7664 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7665 DIVWEUO and DIVDEUO.
7666 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7667 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7668 (div_extend): Likewise.
7669 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7671 (__builtin_divweuo): Likewise.
7672 (__builtin_divdeo): Likewise.
7673 (__builtin_divdeuo): Likewise.
7675 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7678 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7679 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7681 PR tree-optimization/82004
7682 * gimple-match-head.c (optimize_pow_to_exp): New function.
7683 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7684 Don't fold to exp if optimize_pow_to_exp is false.
7686 2018-03-28 Martin Liska <mliska@suse.cz>
7689 * calls.c (initialize_argument_information): Fix trailing space.
7690 * common.opt: Fix typo and provide better explanation for
7691 -fsanitize-coverage option.
7692 * config/i386/i386.opt: Fix typo.
7694 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7695 Martin Liska <mliska@suse.cz>
7698 * gimplify.c (asan_poison_variable): Don't do the check for
7699 gimplify_omp_ctxp here.
7700 (gimplify_decl_expr): Do it here.
7701 (gimplify_target_expr): Likewise.
7703 2018-03-28 Martin Liska <mliska@suse.cz>
7706 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7707 chkp_type_bounds_count if MPX is not enabled.
7709 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7711 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7713 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7716 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7717 function to create the function decl for complex long double
7718 multiply and divide for -mabi=ieeelongdouble.
7719 (init_float128_ieee): Call it.
7721 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7724 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7725 -fcf-protection=branch -mibt.
7726 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7728 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7731 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7732 arm_word_relocations.
7734 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7737 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7738 extern array declarations.
7740 2018-03-27 Richard Biener <rguenther@suse.de>
7743 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7744 explicit single_use checks.
7746 2018-03-27 Richard Biener <rguenther@suse.de>
7748 PR tree-optimization/85082
7749 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7752 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7754 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7755 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7756 Turn on fasynchronous-unwind-tables and funwind-tables.
7758 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7761 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7762 (*bmi_blsr_<mode>_ccz): Ditto.
7764 2018-03-26 Tom de Vries <tom@codesourcery.com>
7766 PR tree-optimization/85063
7767 * omp-general.c (offloading_function_p): New function. Factor out
7769 * omp-offload.c (pass_omp_target_link::gate): ... here.
7770 * omp-general.h (offloading_function_p): Declare.
7771 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7772 with attribute omp declare target for offloading functions.
7774 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7776 PR tree-optimization/84005
7777 * tree-data-ref.h (get_base_for_alignment): Declare.
7778 * tree-data-ref.c (get_base_for_alignment_1): New function.
7779 (get_base_for_alignment): Likewise.
7780 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7781 get_base_for_alignment to find a suitable base object, instead
7782 of always using drb->base_address.
7784 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7787 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7788 known size by default.
7790 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7793 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7794 and non BLKmode operands.
7796 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7799 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7800 Clean up attributes.
7802 2018-03-23 Richard Biener <rguenther@suse.de>
7805 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7806 we are going to emit early debug for LTO.
7808 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7811 * function.c (match_asm_constraints_1): Don't optimize if input
7812 doesn't satisfy general_operand predicate for output's mode.
7815 * alias.c (write_dependence_p): Don't require for x_canonicalized
7816 non-VOIDmode if x has VOIDmode.
7819 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7820 just don't try to optimize it rather than assert it never happens.
7822 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7824 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7825 macro expansions for definition of ST_INTERNAL_<mode> and
7826 LD_INTERNAL_<mode> builtins.
7827 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7829 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7831 (altivec_expand_st_builtin): Likewise.
7832 (altivec_expand_builtin): Remove calls to deleted functions.
7833 (rs6000_address_for_altivec): Delete this function.
7834 * config/rs6000/vector.md: Remove expands for
7835 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7837 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7840 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7841 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7842 re-computing once computed.
7843 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7844 (arm_init_machine_status): Initialize
7845 machine->static_chain_stack_bytes.
7847 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7850 * doc/extend.texi: Add four new prototypes for vec_ld.
7851 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7852 definitions for more logical presentation.
7853 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7854 entries for V1TI variants of __builtin_altivec_ld builtin.
7855 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7856 handling of V1TI variant of LVX icode pattern.
7857 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7858 (rs6000_gimple_fold_builtin): Likewise.
7859 (altivec_init_builtins): Add code to define
7860 __builtin_altivec_lvx_v1ti function.
7862 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7865 * function.c (match_asm_constraints_1): Don't do the optimization
7866 if input isn't a REG, SUBREG, MEM or constant.
7868 2018-03-22 Tom de Vries <tom@codesourcery.com>
7870 PR tree-optimization/84956
7871 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7872 bb_has_abnormal_pred.
7874 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7877 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7878 DECL_INITIAL (decl) to decl at the end.
7879 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7882 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7884 * doc/extend.texi (__builtin_tgmath): Document when complex
7885 integer types are treated as _Complex _Float64.
7887 2018-03-21 Tom de Vries <tom@codesourcery.com>
7889 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7891 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7893 PR tree-optimization/84960
7894 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7895 if it is ENTRY block, move them into single succ of ENTRY in that case.
7897 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7899 PR tree-optimization/84811
7900 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7901 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7902 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7903 (known_subrange_p): Update accordingly. Cast each value involved
7904 in the size comparison, rather than casting the result of the
7907 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7909 PR tree-optimization/84982
7910 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7911 by flipping the least significant bit rather than all bits from
7912 bitpos to bitpos + bitsize - 1.
7914 2018-03-21 Nathan Sidwell <nathan@acm.org>
7916 * doc/extend.texi (Deprecated Features): Remove mention of
7917 long-deleted deprecations.
7919 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7922 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7923 * configure: Regenerate.
7925 2018-03-21 Tom de Vries <tom@codesourcery.com>
7927 PR tree-optimization/83126
7928 * tree-parloops.c (num_phis): New function.
7929 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7931 2018-03-21 Nathan Sidwell <nathan@acm.org>
7933 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7934 mention anon-struct/union members and trailing attributes.
7936 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7938 PR tree-optimization/84969
7939 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7940 builtin memset partitions if they set different rhs values.
7942 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7944 PR rtl-optimization/84989
7945 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7946 VEC_DUPLICATE with scalar result mode.
7948 2018-03-21 Martin Liska <mliska@suse.cz>
7951 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7952 not intended return statement.
7954 2018-03-21 Martin Liska <mliska@suse.cz>
7957 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7958 (chkp_find_bound_slots_1): Limit number of iterations.
7960 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7963 * Minor grammar fixes for x86 options.
7965 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7968 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7969 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7971 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7974 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7975 (altivec_lvx_<mode>_1op): Likewise.
7976 (altivec_stvx_<mode>_2op): Likewise.
7977 (altivec_stvx_<mode>_1op): Likewise.
7978 (altivec_lvx_<VM2:mode>): New define_expand.
7979 (altivec_stvx_<VM2:mode>): Likewise.
7980 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7981 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7982 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7983 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7984 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7985 (rs6000_gen_lvx): Likewise.
7986 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7987 (altivec_expand_stv_builtin): Likewise.
7988 (altivec_expand_builtin): Likewise.
7989 * config/rs6000/vector.md: Likewise.
7991 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7994 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7997 2018-03-20 Richard Biener <rguenther@suse.de>
8000 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8001 sign-conversions as zero, fall back to standard scalar_stmt
8004 2018-03-20 Martin Liska <mliska@suse.cz>
8007 * predict.c (rebuild_frequencies): Handle case when we have
8008 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8010 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8013 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8014 flag_section_anchors.
8015 * varasm.c (use_blocks_for_decl_p): Remove hack for
8016 dw2_force_const_mem.
8019 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8021 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8022 be created, use lowpart_subreg of operands[0] rather than operands[0]
8024 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8025 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8026 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8027 and n constraint instead of aarch64_shift_imm_di and Usd.
8028 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8029 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8031 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8034 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8035 to favor GPR over NEON registers.
8036 (<shift>di3_neon): Likewise.
8038 2018-03-20 Tom de Vries <tom@codesourcery.com>
8041 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8042 (nvptx_process_pars): Emit bar.sync asap and alap.
8044 2018-03-20 Tom de Vries <tom@codesourcery.com>
8047 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8048 seen_label if seen_label is already set.
8050 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8053 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8054 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8055 Use 1U instead of 1. Formatting fixes.
8058 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8059 instead of TREE_TYPE (s1) for the return value.
8061 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8063 PR tree-optimization/84946
8064 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8065 bitsize + bitsize in poly_uint64 rather than poly_int64.
8068 * dwarf2asm.c: Include fold-const.c.
8069 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8070 of decl rather than decl itself.
8072 PR rtl-optimization/84643
8073 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8075 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8078 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8079 calling assemble_variable.
8081 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8084 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8085 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8087 2018-03-19 Jim Wilson <jimw@sifive.com>
8090 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8091 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8092 (riscv_first_stack_step): Likewise.
8093 (riscv_option_override): Use STACK_BOUNDARY instead of
8095 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8097 (BIGGEST_ALIGNMENT): Set to 128.
8098 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8099 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8102 2018-03-19 Richard Biener <rguenther@suse.de>
8104 PR tree-optimization/84933
8105 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8106 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8108 2018-03-19 Richard Biener <rguenther@suse.de>
8110 PR tree-optimization/84859
8111 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8112 (cond_if_else_store_replacement): Perform sinking operation on
8113 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8114 Generalize what a BB with a single eligible store is.
8116 2018-03-19 Richard Biener <rguenther@suse.de>
8118 PR tree-optimization/84929
8119 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8120 chrec_is_positive against non-chrec arg.
8122 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8125 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8127 2018-03-18 Martin Liska <mliska@suse.cz>
8129 PR rtl-optimization/84635
8130 * regrename.c (build_def_use): Use matches_mode only when
8133 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8135 PR tree-optimization/84913
8136 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8137 vectorize chains of COND_EXPRs.
8139 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8141 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8143 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8145 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8147 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8149 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8151 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8152 Kito Cheng <kito.cheng@gmail.com>
8154 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8155 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8156 (nds32_adjust_reg_alloc_order): New function.
8157 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8159 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8161 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8162 nds32_print_operand, nds32_print_operand_address): Use
8163 HOST_WIDE_INT_PRINT_DEC instead.
8165 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8167 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8169 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8172 * config/i386/i386.c (initial_ix86_tune_features,
8173 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8175 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8176 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8177 rather than 1u << ix86_tune. Formatting fix.
8178 (ix86_option_override_internal): Change ix86_arch_mask from
8179 unsigned int to unsigned HOST_WIDE_INT, initialize to
8180 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8181 (ix86_function_specific_restore): Likewise.
8183 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8186 * postreload.c (reload_combine_recognize_pattern): Perform
8187 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8188 truncate_int_for_mode the result for the destination's mode.
8191 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8192 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8194 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8197 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8200 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8201 regno_allocno_class_array and sorted_pseudos.
8202 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8203 insns where regno is used.
8205 2018-03-16 Martin Liska <mliska@suse.cz>
8208 * multiple_target.c (create_dispatcher_calls): Redirect
8209 reference in the symbol table.
8211 2018-03-16 Martin Liska <mliska@suse.cz>
8214 * multiple_target.c (create_dispatcher_calls): Redirect also
8217 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8221 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8222 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8223 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8225 2018-03-16 Julia Koval <julia.koval@intel.com>
8227 * doc/invoke.texi (Skylake Server): Add CLWB.
8228 Cannonlake): Remove CLWB.
8230 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8232 PR tree-optimization/84841
8233 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8235 (FLOAT_ONE_CONST_TYPE): Define.
8236 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8237 (sort_by_operand_rank): Put entries with higher constant_type last
8238 rather than first to match comments.
8240 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8242 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8245 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8248 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8249 check and use address of target always.
8251 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8254 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8255 (indirect_thunk_bnd_needed): Likewise.
8256 (indirect_thunks_used): Likewise.
8257 (indirect_thunks_bnd_used): Likewise.
8258 (indirect_return_needed): New.
8259 (indirect_return_bnd_needed): Likewise.
8260 (output_indirect_thunk_function): Add a bool argument for
8262 (output_indirect_thunk_function): Don't generate alias for
8263 function return thunk.
8264 (ix86_code_end): Call output_indirect_thunk_function to generate
8265 function return thunks.
8266 (ix86_output_function_return): Set indirect_return_bnd_needed
8267 and indirect_return_needed instead of indirect_thunk_bnd_needed
8268 and indirect_thunk_needed.
8270 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8272 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8273 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8274 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8276 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8277 Paul Hua <paul.hua.gm@gmail.com>
8280 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8282 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8284 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8285 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8288 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8291 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8292 instead of GET_MODE_SIZE when comparing Units.
8294 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8297 * varasm.c (hash_section): Return an unchangeble hash value
8298 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8299 Return !aarch64_can_use_per_function_literal_pools_p ().
8301 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8304 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8305 rather than address of cmode as last argument to prepare_cmp_insn.
8307 2018-03-15 Julia Koval <julia.koval@intel.com>
8309 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8310 F_AVX512VNNI, F_AVX512BITALG): New.
8312 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8315 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8316 insn for floating-point loads and stores.
8318 2018-03-14 Carl Love <cel@us.ibm.com>
8320 * config/rs6000/rs6000-c.c: Add macro definitions for
8321 ALTIVEC_BUILTIN_VEC_PERMXOR.
8322 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8323 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8324 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8325 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8327 * config/doc/extend.texi: Add prototypes for vec_permxor.
8329 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8332 * diagnostic-show-locus.c (class layout_point): Convert m_line
8333 from int to linenum_type.
8334 (line_span::comparator): Use linenum "compare" function when
8335 comparing line numbers.
8336 (test_line_span): New function.
8337 (layout_range::contains_point): Convert param "row" from int to
8339 (layout_range::intersects_line_p): Likewise.
8340 (layout::will_show_line_p): Likewise.
8341 (layout::print_source_line): Likewise.
8342 (layout::should_print_annotation_line_p): Likewise.
8343 (layout::print_annotation_line): Likewise.
8344 (layout::print_leading_fixits): Likewise.
8345 (layout::annotation_line_showed_range_p): Likewise.
8346 (struct line_corrections): Likewise for field m_row.
8347 (line_corrections::line_corrections): Likewise for param "row".
8348 (layout::print_trailing_fixits): Likewise.
8349 (layout::get_state_at_point): Likewise.
8350 (layout::get_x_bound_for_row): Likewise.
8351 (layout::print_line): Likewise.
8352 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8353 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8354 * input.c (selftest::test_linenum_comparisons): New function.
8355 (selftest::input_c_tests): Call it.
8356 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8357 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8358 * selftest.h (ASSERT_GT): New macro.
8359 (ASSERT_GT_AT): New macro.
8360 (ASSERT_LT): New macro.
8361 (ASSERT_LT_AT): New macro.
8363 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8365 PR rtl-optimization/84780
8366 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8368 2018-03-14 Martin Liska <mliska@suse.cz>
8370 * tree.c (record_node_allocation_statistics): Use
8371 get_stats_node_kind.
8372 (get_stats_node_kind): New function extracted from
8373 record_node_allocation_statistics.
8374 (free_node): Use get_stats_node_kind.
8376 2018-03-14 Richard Biener <rguenther@suse.de>
8378 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8379 that the value-set of ANTIC_IN doesn't grow.
8382 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8384 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8385 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8387 2018-03-14 Julia Koval <julia.koval@intel.com>
8389 * config.gcc (icelake-client, icelake-server): New.
8391 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8392 (initial_ix86_arch_features): Ditto.
8393 (PTA_SKYLAKE): Add SGX.
8394 (PTA_ICELAKE): Remove.
8395 (PTA_ICELAKE_CLIENT): New.
8396 (PTA_ICELAKE_SERVER): New.
8397 (ix86_option_override_internal): Split up icelake on icelake client and
8399 (get_builtin_code_for_version): Ditto.
8400 (fold_builtin_cpu): Ditto.
8401 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8402 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8403 * config/i386/i386.h (processor_type): Ditto.
8404 * doc/invoke.texi: Ditto.
8406 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8409 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8410 INTEGER_CST offset, add it together with bitpos / 8 and
8411 sign extend based on POINTER_SIZE.
8415 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8418 * config/i386/constraints.md (Yc): New register constraint.
8419 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8420 Use Yc constraint for alternative 2 of operand 0. Remove
8421 preferred_for_speed attribute.
8423 2018-03-14 Richard Biener <rguenther@suse.de>
8425 PR tree-optimization/84830
8426 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8427 with the old one to avoid oscillations.
8429 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8432 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8434 (assign_by_spills): Return a flag of reload assignment failure.
8435 Do not process the reload assignment failures. Do not spill other
8436 reload pseudos if they has the same reg class. Update n if
8438 (lra_assign): Add a return arg. Set up from the result of
8439 assign_by_spills call.
8440 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8441 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8442 usage_insns if it is not NULL.
8443 (spill_hard_reg_in_range): New function.
8444 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8445 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8446 function prototypes.
8447 (lra_assign): Change prototype.
8448 * lra.c (lra): Add code to deal with fails by splitting hard reg
8451 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8453 * config/riscv/riscv.opt (mrelax): New option.
8454 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8455 "norelax" when riscv_mrelax is disabled.
8456 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8458 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8461 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8462 reassociation for int modes.
8464 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8466 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8467 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8469 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8470 * config/aarch64/aarch64-sve.md
8471 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8472 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8473 (*extend<mode><Vwide>2): Rename to...
8474 (aarch64_sve_extend<mode><Vwide>2): ...this.
8475 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8476 renaming the old pattern to...
8477 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8479 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8480 define_expand, renaming the old pattern to...
8481 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8482 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8483 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8484 account when deciding which SVE instruction the optab should use.
8485 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8487 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8489 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8490 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8491 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8492 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8493 (tlsdesc_small_<mode>): Turn a define_expand and use
8494 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8495 (tlsdesc_small_advsimd_<mode>): ...this.
8496 (tlsdesc_small_sve_<mode>): New pattern.
8498 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8500 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8501 (UNSPEC_UMUL_HIGHPART): New constants.
8502 (MUL_HIGHPART): New int iteraor.
8503 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8504 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8506 (*<su>mul<mode>3_highpart): New define_insn.
8508 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8511 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8514 2018-03-13 Martin Liska <mliska@suse.cz>
8517 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8519 (sem_item_optimizer::~sem_item_optimizer): Release it.
8520 (sem_item_optimizer::merge_classes): Register variable aliases.
8521 (sem_item_optimizer::fixup_pt_set): New function.
8522 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8523 * ipa-icf.h: Declare new variables and functions.
8525 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8528 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8529 integer_pow2p@2 and test integer_pow2p in condition.
8530 (A < 0 ? C : 0): Similarly for @1.
8533 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8534 characters starting at p contain '\0' character, don't look beyond
8538 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8539 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8542 * reg-stack.c (change_stack): Change update_end var from int to
8543 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8544 also call set_block_for_insn on the newly added insns and rescan.
8547 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8548 on the last operand.
8551 * tree.c (stabilize_reference_1): Return save_expr (e) for
8552 STATEMENT_LIST even if it doesn't have side-effects.
8554 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8556 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8558 2018-03-12 Renlin Li <renlin.li@arm.com>
8560 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8561 aarch64_output_scalar_simd_mov_immediate.
8563 2018-03-12 Martin Sebor <msebor@redhat.com>
8565 PR tree-optimization/83456
8566 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8567 for perfectly overlapping calls to memcpy.
8568 (gimple_fold_builtin_memory_chk): Same.
8569 (gimple_fold_builtin_strcpy): Handle no-warning.
8570 (gimple_fold_builtin_stxcpy_chk): Same.
8571 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8573 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8575 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8576 parameter. Use it for SFmode.
8577 (rs6000_function_arg_advance_1): Adjust.
8578 (rs6000_function_arg): Adjust.
8579 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8581 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8583 PR rtl-optimization/84169
8584 PR rtl-optimization/84780
8585 * combine.c (can_combine_p): Check for a 2-insn combination whether
8586 the destination register is used between the two insns, too.
8588 2018-03-12 Richard Biener <rguenther@suse.de>
8590 PR tree-optimization/84803
8591 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8592 for refs DR analysis didn't process.
8594 2018-03-12 Richard Biener <rguenther@suse.de>
8596 PR tree-optimization/84777
8597 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8598 force-vectorize loops ignore whether we are optimizing for size.
8600 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8602 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8603 (TARGET_MD_ASM_ADJUST): Define.
8605 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8606 Kito Cheng <kito.cheng@gmail.com>
8607 Chung-Ju Wu <jasonwucj@gmail.com>
8609 * config/nds32/nds32.c (nds32_compute_stack_frame,
8610 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8611 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8612 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8613 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8614 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8615 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8616 * config/nds32/nds32.md (prologue, epilogue): Use macro
8617 NDS32_V3PUSH_AVAILABLE_P to do checking.
8619 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8622 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8623 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8624 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8625 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8626 addition of most attributes on !orig_type_die or the attribute not
8627 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8629 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8630 Chung-Ju Wu <jasonwucj@gmail.com>
8632 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8634 * config/nds32/nds32.opt (mvh): New option.
8636 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8637 Chung-Ju Wu <jasonwucj@gmail.com>
8639 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8641 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8642 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8645 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8646 Chung-Ju Wu <jasonwucj@gmail.com>
8648 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8650 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8651 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8653 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8654 Kito Cheng <kito.cheng@gmail.com>
8655 Chung-Ju Wu <jasonwucj@gmail.com>
8657 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8658 UNSPEC_FFMISM and UNSPEC_FLMISM.
8659 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8660 for ffb, ffmism and flmism.
8661 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8662 (unspec_ffmism): Ditto.
8663 (unspec_flmism): Ditto.
8664 (nds32_expand_builtin_impl): Check if string extension is available.
8665 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8666 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8668 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8671 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8674 * lra-assigns.c (assign_by_spills): Return a flag of reload
8675 assignment failure. Do not process the reload assignment
8676 failures. Do not spill other reload pseudos if they has the same
8678 (lra_assign): Add a return arg. Set up from the result of
8679 assign_by_spills call.
8680 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8681 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8682 usage_insns if it is not NULL.
8683 (spill_hard_reg_in_range): New function.
8684 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8685 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8686 function prototypes.
8687 (lra_assign): Change prototype.
8688 * lra.c (lra): Add code to deal with fails by splitting hard reg
8691 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8694 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8696 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8699 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8700 (dw_val_node): Add val_symbolic_view.
8701 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8702 (symview_upper_bound): New.
8703 (new_line_info_table): Initialize symviews_since_reset.
8704 (dwarf2out_source_line): Count symviews_since_reset and set
8705 symview_upper_bound.
8706 (dw_val_equal_p): Handle symview.
8707 (add_AT_symview): New.
8708 (print_dw_val): Handle symview.
8709 (attr_checksum, attr_checksum_ordered): Likewise.
8710 (same_dw_val_p, size_of_die): Likewise.
8711 (value_format, output_die): Likewise.
8712 (add_high_low_attributes): Use add_AT_symview for entry_view.
8713 (dwarf2out_finish): Reset symview_upper_bound, clear
8716 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8719 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8720 Add strict argument and use it.
8721 (rs6000_split_multireg_move): Update for new strict argument.
8722 (mem_operand_gpr): Disallow all non-offsettable addresses.
8723 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8725 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8728 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8729 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8730 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8733 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8734 decl, use remap_type if we want to use the type.
8736 2018-03-09 Martin Sebor <msebor@redhat.com>
8738 PR tree-optimization/84526
8739 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8741 (builtin_access::generic_overlap): Be prepared to handle non-array
8744 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8746 PR rtl-optimization/84682
8747 * lra-constraints.c (process_address_1): Check is_address flag
8748 for address constraints.
8749 (process_alt_operands): Likewise.
8750 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8751 preprocess_constraints.
8752 * recog.h (preprocess_constraints): Add oploc parameter.
8754 * recog.c (preprocess_constraints): Test address_operand for
8755 CT_ADDRESS constraints.
8757 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8760 * lra-assigns.c (assign_by_spills): Return a flag of reload
8761 assignment failure. Do not process the reload assignment
8762 failures. Do not spill other reload pseudos if they has the same
8764 (lra_assign): Add a return arg. Set up from the result of
8765 assign_by_spills call.
8766 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8767 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8768 usage_insns if it is not NULL.
8769 (spill_hard_reg_in_range): New function.
8770 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8771 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8772 function prototypes.
8773 (lra_assign): Change prototype.
8774 * lra.c (lra): Add code to deal with fails by splitting hard reg
8777 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8780 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8781 Accept complain bool parameter. Only emit errors if it is true.
8782 (arm_parse_cpu_option_name): Likewise.
8783 (arm_target_thumb_only): Adjust callers of the above.
8784 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8785 prototype to take a default true bool parameter.
8786 (arm_parse_arch_option_name): Likewise.
8788 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8789 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8793 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8794 * configure: Regenerate.
8795 * configure.ac ("linker --version-script option"): New.
8796 ("linker soname option"): New.
8798 2018-03-09 Richard Biener <rguenther@suse.de>
8800 PR tree-optimization/84775
8801 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8802 immediate uses of predicate stmts and mark them modified.
8805 PR tree-optimization/84178
8806 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8808 (version_loop_for_if_conversion): Delay update_ssa call.
8809 (tree_if_conversion): Delay update_ssa until after predicate
8812 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8815 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8816 when the function accesses prior frames.
8818 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8821 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8822 gen_llsym, otherwise call maybe_gen_llsym.
8825 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8826 has ',' character inside of it.
8828 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8831 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8832 as clobbering CC_REGNUM.
8834 2018-03-08 Richard Biener <rguenther@suse.de>
8837 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8838 (follow_copies_to_constant): Do not follow SSA names registered
8841 2018-03-08 Richard Biener <rguenther@suse.de>
8843 PR tree-optimization/84178
8844 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8846 (version_loop_for_if_conversion): Delay update_ssa call.
8847 (tree_if_conversion): Delay update_ssa until after predicate
8850 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8852 PR tree-optimization/84178
8853 * tree-if-conv.c (release_bb_predicate): Remove the
8854 the assertion that the stmts have NULL use_ops.
8855 Discard the statements, asserting that they haven't
8856 yet been added to a BB.
8858 2018-03-08 Richard Biener <rguenther@suse.de>
8860 PR tree-optimization/84746
8861 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8862 (phi_translate): Pass in destination ANTIC_OUT set.
8863 (phi_translate_1): Likewise. For a simplified result lookup
8864 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8865 (phi_translate_set): Adjust.
8866 (do_pre_regular_insertion): Likewise.
8867 (do_pre_partial_partial_insertion): Likewise.
8869 2018-03-08 Martin Liska <mliska@suse.cz>
8871 PR gcov-profile/84735
8872 * doc/gcov.texi: Document usage of profile files.
8873 * gcov-io.h: Document changes in the format.
8875 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8879 * dwarf2out.c (struct dw_line_info_table): Update comments for
8881 (FORCE_RESET_NEXT_VIEW): New.
8882 (FORCE_RESETTING_VIEW_P): New.
8883 (RESETTING_VIEW_P): Check for -1 too.
8884 (ZERO_VIEW_P): Likewise.
8885 (new_line_info_table): Force-reset next view.
8886 (dwarf2out_begin_function): Likewise.
8887 (dwarf2out_source_line): Simplify zero_view_p initialization.
8888 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8889 view directly. Omit view when omitting .loc at line 0.
8891 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8893 PR tree-optimization/84740
8894 * tree-switch-conversion.c (process_switch): Call build_constructors
8895 only if info.phi_count is non-zero.
8897 PR tree-optimization/84739
8898 * tree-tailcall.c (find_tail_calls): Check call arguments against
8899 DECL_ARGUMENTS (current_function_decl) rather than
8900 DECL_ARGUMENTS (func) when checking for tail recursion.
8902 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8904 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8905 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8906 Volker Reichelt's entry and add entries for people that perform
8907 GCC fuzzy testing and report numerous bugs.
8909 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8912 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8913 readonly data in sdata, if that is disabled.
8914 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8915 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8916 -mreadonly-in-sdata option.
8918 2018-03-07 Martin Sebor <msebor@redhat.com>
8920 PR tree-optimization/84468
8921 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8922 basic block when looking for nul assignment.
8924 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8927 * except.h (output_function_exception_table): Adjust prototype.
8928 * except.c (output_function_exception_table): Remove FNNAME parameter
8929 and add SECTION parameter. Ouput one part of the table at a time.
8930 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8931 the first part of the exception table and emit unwind directives.
8932 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8933 (i386_pe_seh_cold_init): Likewise.
8934 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8935 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8936 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8937 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8938 * config/i386/winnt.c: Include except.h.
8939 (struct seh_frame_state): Add reg_offset, after_prologue and
8940 in_cold_section fields.
8941 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8942 (i386_pe_seh_cold_init): New function.
8943 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8944 to seh->in_cold_section.
8945 (seh_emit_push): Record the offset of the push.
8946 (seh_emit_save): Record the offet of the save.
8947 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8948 Test seh->after_prologue to disregard the epilogue.
8949 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8950 (i386_pe_end_cold_function): New function.
8952 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8955 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8956 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8959 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8961 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8962 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8965 * multiple_target.c: Include tree-inline.h and intl.h.
8966 (expand_target_clones): Diagnose and fail if node->definition and
8967 !tree_versionable_function_p (node->decl).
8969 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8971 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8973 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8974 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8975 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8977 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8980 * combine.c (try_combine): Use reg_or_subregno instead of handling
8981 just paradoxical SUBREGs and REGs.
8983 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8985 * config/arc/arc.c (arc_finalize_pic): Remove function.
8986 (arc_must_save_register): We use single base PIC register, remove
8987 checks to save/restore the PIC register.
8988 (arc_expand_prologue): Likewise.
8989 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8991 (arc_verify_short): Likewise.
8992 (arc_attr_type): Likewise.
8993 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8994 (walk_stores): Likewise.
8995 (arc_address_cost): Make it static.
8996 (arc_verify_short): Likewise.
8997 (branch_dest): Likewise.
8998 (arc_attr_type): Likewise.
8999 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9000 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9001 (arc_final_prescan_insn): Remove inserting the nops due to
9002 hardware hazards. It is done in reorg step.
9003 (insn_length_variant_t): Remove.
9004 (insn_length_parameters_t): Likewise.
9005 (arc_insn_length_parameters): Likewise.
9006 (arc_get_insn_variants): Likewise.
9007 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9009 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9012 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9015 PR tree-optimization/84687
9016 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9018 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9020 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9022 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9023 Rename to ppc_speculation_barrier.
9024 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9025 __builtin_ppc_speculation_barrier.
9027 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9030 * combine.c (combine_simplify_rtx): Don't try to simplify if
9031 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9034 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9036 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9037 to 32 bytes when compiling for POWER9.
9039 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9042 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9043 regparm >= 3 with no arg reg available also for calls with
9044 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9047 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9049 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9051 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9054 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9056 2018-03-05 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/84486
9059 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9060 When inserting a __builtin_assume_aligned call set the LHS
9061 SSA name alignment info accordingly.
9063 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9065 PR tree-optimization/84114
9066 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9067 Avoid reassociation of FLOAT_MODE addition.
9069 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9071 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9072 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9073 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9074 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9075 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9076 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9077 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9079 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9080 __builtin_ia32_wbinvd): New builtins.
9081 (SPECIAL_ARGS2): New.
9082 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9083 (SPECIAL_ARGS2): New.
9084 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9085 (ix86_valid_target_attribute_inner_p): Ditto.
9086 (ix86_init_mmx_sse_builtins): Add special_args2.
9087 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9088 TARGET_WBNOINVD_P): New.
9089 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9090 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9091 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9092 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9093 * config/i386/pconfigintrin.h: New file.
9094 * config/i386/wbnoinvdintrin.h: Ditto.
9095 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9097 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9099 2018-03-05 Richard Biener <rguenther@suse.de>
9101 PR tree-optimization/84670
9102 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9104 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9105 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9106 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9107 doesn't grow if all successors (recursively) were visited at least
9110 2018-03-05 Richard Biener <rguenther@suse.de>
9112 PR tree-optimization/84650
9113 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9114 if executed in the loop pipeline.
9116 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9118 * doc/configfiles.texi (Configuration Files): Move info about
9119 conditionalizing $target-protos.h to...
9120 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9121 differs from $target-protos.h.
9123 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9124 Chung-Ju Wu <jasonwucj@gmail.com>
9126 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9127 * config/nds32/nds32-multiple.md (setmemsi): Define.
9128 * config/nds32/nds32-memory-manipulation.c
9129 (nds32_gen_dup_4_byte_to_word_value): New.
9130 (emit_setmem_word_loop): New.
9131 (emit_setmem_byte_loop): New.
9132 (nds32_expand_setmem_loop): New.
9133 (nds32_expand_setmem_loop_v3m): New.
9134 (nds32_expand_setmem_unroll): New.
9135 (nds32_expand_setmem): New.
9137 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9138 Chung-Ju Wu <jasonwucj@gmail.com>
9140 * config/nds32/nds32-memory-manipulation.c
9141 (nds32_emit_load_store): New.
9142 (nds32_emit_post_inc_load_store): New.
9143 (nds32_emit_mem_move): New.
9144 (nds32_emit_mem_move_block): New.
9145 (nds32_expand_movmemsi_loop_unknown_size): New.
9146 (nds32_expand_movmemsi_loop_known_size): New.
9147 (nds32_expand_movmemsi_loop): New.
9148 (nds32_expand_movmemsi_unroll): New.
9149 (nds32_expand_movmemqi): Rename ...
9150 (nds32_expand_movmemsi): ... to this.
9151 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9152 (movmemsi): ... to this.
9153 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9154 (nds32_expand_movmemsi): ... to this.
9156 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9157 Monk Chiang <sh.chiang04@gmail.com>
9158 Chung-Ju Wu <jasonwucj@gmail.com>
9160 * config/nds32/nds32-protos.h
9161 (nds32_expand_load_multiple): New arguments.
9162 (nds32_expand_store_multiple): Ditto.
9163 (nds32_valid_multiple_load_store): Rename ...
9164 (nds32_valid_multiple_load_store_p): ... to this.
9165 * config/nds32/nds32-memory-manipulation.c
9166 (nds32_expand_load_multiple): Refine implementation.
9167 (nds32_expand_store_multiple): Ditto.
9168 * config/nds32/nds32-multiple.md
9169 (load_multiple): Update nds32_expand_load_multiple interface.
9170 (store_multiple): Update nds32_expand_store_multiple interface.
9171 * config/nds32/nds32-predicates.c
9172 (nds32_valid_multiple_load_store): Rename ...
9173 (nds32_valid_multiple_load_store_p): ... to this and refine
9175 * config/nds32/predicates.md
9176 (nds32_load_multiple_and_update_address_operation): New predicate.
9177 (nds32_store_multiple_and_update_address_operation): New predicate.
9179 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9180 Chung-Ju Wu <jasonwucj@gmail.com>
9182 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9183 (combo): New attribute.
9184 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9186 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9188 * config/nds32/nds32.opt: Change -mcmodel= default value.
9190 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9191 Monk Chiang <sh.chiang04@gmail.com>
9192 Chung-Ju Wu <jasonwucj@gmail.com>
9194 * config/nds32/constants.md (unspec_element): New enum.
9195 * config/nds32/constraints.md (Umw): New constraint.
9196 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9197 * config/nds32/nds32-intrinsic.md: Likewise.
9198 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9199 (nds32_valid_smw_lwm_base_p): New.
9200 (nds32_output_smw_single_word): New.
9201 (nds32_output_lmw_single_word): New.
9202 (nds32_expand_unaligned_load): New.
9203 (nds32_expand_unaligned_store): New.
9204 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9205 (nds32_output_smw_single_word): Declare.
9206 (nds32_output_lmw_single_word): Declare.
9207 (nds32_expand_unaligned_load): Declare.
9208 (nds32_expand_unaligned_store): Declare.
9209 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9210 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9211 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9212 NDS32_BUILTIN_UASTORE_DW.
9213 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9216 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9217 Kito Cheng <kito.cheng@gmail.com>
9218 Chung-Ju Wu <jasonwucj@gmail.com>
9220 * config/nds32/nds32-intrinsic.c
9221 (nds32_expand_builtin_null_ftype_reg): Delete.
9222 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9223 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9224 (nds32_read_argument): New.
9225 (nds32_legitimize_target): Ditto.
9226 (nds32_legitimize_argument): Ditto.
9227 (nds32_check_constant_argument): Ditto.
9228 (nds32_expand_unop_builtin): Ditto.
9229 (nds32_expand_unopimm_builtin): Ditto.
9230 (nds32_expand_binop_builtin): Ditto.
9231 (nds32_builtin_decl_impl): Ditto.
9232 (builtin_description): Ditto.
9233 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9234 (nds32_init_builtins_impl): Ditto.
9235 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9236 (nds32_builtin_decl): New.
9237 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9238 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9240 2018-03-02 Jeff Law <law@redhat.com>
9242 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9243 (redundant_insn, fill_simple_delay_slots): Likewise.
9244 (fill_slots_from_thread): Likewise.
9245 * resource.c (mark_referenced_resources): Likewise.
9246 (mark_set_resources, find_dead_or_set_registers): Likewise.
9248 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9250 * substring-locations.h (format_warning_va): Formatting fix for
9252 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9254 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9255 * substring-locations.c: Include intl.h.
9256 (format_warning_va): Turned into small wrapper around
9257 format_warning_n_va, renamed to ...
9258 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9259 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9261 (format_warning_at_substring_n): New function.
9262 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9263 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9264 format_warning_at_substring with just a shorter name instead of
9265 const function pointer.
9266 (fmtwarn_n): New function.
9267 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9268 appropriate, get rid of all the fmtstr temporaries, move conditionals
9269 with G_() wrapped string literals directly into fmtwarn arguments,
9270 cast dir.len to (int), formatting fixes.
9272 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9274 * doc/invoke.texi: Remove "Cilk Plus" references.
9276 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9277 Richard Biener <rguenther@suse.de>
9280 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9281 for error or warning attributes if CALL_FROM_THUNK_P is set.
9284 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9287 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9288 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9291 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9292 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9294 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9295 instead of -1U in last predictors element's probability member.
9297 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9300 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9301 arguments if they are comparable.
9303 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9305 PR tree-optimization/84634
9306 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9307 masks and masked_loop_p with a single loop_masks, making sure it's
9308 null for bb vectorization.
9310 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9312 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9313 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9314 loop->force_vectorize to check whether there is no alias.
9316 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9319 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9321 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9323 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9324 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9325 instead of a loop around prev_real_insn.
9326 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9330 * config/i386/i386.c (ix86_print_operand): Use conditional
9331 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9334 2018-03-02 Richard Biener <rguenther@suse.de>
9336 PR tree-optimization/84427
9337 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9338 (bitmap_set_subtract_values): Rewrite to handle multiple
9341 (prune_clobbered_mems): Likewise.
9342 (phi_translate): Take edge instead of pred/phiblock.
9343 (phi_translate_1): Likewise.
9344 (phi_translate_set): Likewise. Insert all translated
9345 exprs for a value into the set, keeping possibly multiple
9346 expressions per value.
9347 (compute_antic_aux): Adjust for phi_translate changes.
9348 When intersecting union the expressions and prune those
9349 not in the final value set, keeping possibly multiple
9350 expressions per value. Do not use value-insertion
9351 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9352 all expressions. Add verification that the value-sets
9353 only shrink during iteration.
9354 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9355 (do_pre_regular_insertion): Likewise.
9356 (do_pre_partial_partial_insertion): Likewise.
9358 2018-03-02 Richard Biener <rguenther@suse.de>
9361 * config/darwin.c (saved_debug_info_level): New static global.
9362 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9363 (darwin_asm_lto_end): Restore debug info generation settings.
9365 2018-03-01 Martin Liska <mliska@suse.cz>
9368 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9371 2018-03-01 Richard Biener <rguenther@suse.de>
9374 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9375 type annotation with LTO when debug was disabled at compile-time.
9377 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9379 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9381 (mips_final_postscan_insn): Likewise.
9383 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9385 PR rtl-optimization/84528
9386 * alias.c (init_alias_target): Add commentary.
9387 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9388 a unique base value if the frame pointer is not eliminated
9389 to the stack pointer.
9391 2018-03-01 Tom de Vries <tom@codesourcery.com>
9393 PR rtl-optimization/83327
9394 * lra-int.h (hard_regs_spilled_into): Declare.
9395 * lra.c (hard_regs_spilled_into): Define.
9396 (init_reg_info): Init hard_regs_spilled_into.
9397 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9398 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9399 (process_bb_lives): Handle hard_regs_spilled_into.
9400 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9401 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9403 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9405 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9406 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9407 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9408 * config/rs6000/aix72.h: New file.
9410 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9412 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9413 instead of warning_at with conditional singular and plural messages
9417 * stor-layout.c (update_alignment_for_field): For
9418 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9419 && !DECL_PACKED (field), do the alignment update, just use
9420 only desired_align instead of MAX (type_align, desired_align)
9422 (place_field): Don't do known_align < desired_align handling
9423 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9424 is non-NULL, instead do it after rli->prev_field handling and
9425 only if not within a bitfield word. For DECL_PACKED (field)
9426 use type_align of BITS_PER_UNIT.
9428 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9430 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9431 superfluous parentheses and trailing spaces.
9433 2018-02-28 Richard Biener <rguenther@suse.de>
9435 PR tree-optimization/84584
9436 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9437 SCoPs with fake exit edge.
9439 2018-02-28 Martin Liska <mliska@suse.cz>
9442 * timevar.c (timer::print): Fix format to properly print 100%
9445 2018-02-28 Richard Biener <rguenther@suse.de>
9448 * genmatch.c (capture_info::walk_match): Do not mark
9449 captured expressions without operands as expr_p given
9450 they act more like predicates and should be subject to
9451 "lost tail" side-effect preserving.
9453 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9455 PR rtl-optimization/81611
9456 * auto-inc-dec.c (attempt_change): Move dead note from
9457 mem_insn if it's the next use of regno
9458 (find_address): Take address use of reg holding
9459 non-incremented value. Add parm to limit search to the named
9461 (merge_in_block): Attempt to use a mem insn that is the next
9462 use of the original regno.
9464 2018-02-27 Martin Sebor <msebor@redhat.com>
9467 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9468 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9470 2018-02-27 Martin Sebor <msebor@redhat.com>
9472 PR translation/84207
9473 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9474 n argument to unsigned HOST_WIDE_INT.
9475 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9476 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9477 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9478 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9480 2018-02-27 Richard Biener <rguenther@suse.de>
9482 PR tree-optimization/84512
9483 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9484 Do not use the estimate returned from record_stmt_cost for
9485 the scalar iteration cost but sum properly using add_stmt_cost.
9487 2018-02-27 Richard Biener <rguenther@suse.de>
9489 PR tree-optimization/84466
9490 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9491 Adjust last change to less strictly validate use operands.
9493 2018-02-27 Martin Liska <mliska@suse.cz>
9495 PR gcov-profile/84548
9496 * gcov.c (process_file): Allow partial overlap and consider it
9497 also as group functions.
9498 (output_lines): Properly calculate range of lines for a group.
9500 2018-02-27 Martin Liska <mliska@suse.cz>
9502 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9503 'ggc' suffixes. Change first column width.
9504 (timer::print): Fix formatting of the column.
9506 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9508 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9509 preserve inline entry blocks for the sake of debug inline
9510 entry point markers alone.
9511 (remove_unused_locals): Suggest in comments a better place to
9512 force the preservation of inline entry blocks that are
9513 otherwise unused, but do not preserve them.
9515 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9517 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9519 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9522 * config/i386/constraints.md (Bs): Replace
9523 ix86_indirect_branch_register with
9524 TARGET_INDIRECT_BRANCH_REGISTER.
9526 * config/i386/i386.md (indirect_jump): Likewise.
9527 (tablejump): Likewise.
9528 (*sibcall_memory): Likewise.
9529 (*sibcall_value_memory): Likewise.
9530 Peepholes of indirect call and jump via memory: Likewise.
9531 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9532 (*sibcall_value_GOT_32): Likewise.
9533 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9534 (GOT_memory_operand): Likewise.
9535 (call_insn_operand): Likewise.
9536 (sibcall_insn_operand): Likewise.
9537 (GOT32_symbol_operand): Likewise.
9538 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9540 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9542 PR rtl-optimization/83496
9543 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9544 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9545 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9546 redundant insn, if any.
9547 (relax_delay_slots): Likewise.
9548 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9550 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9552 PR tree-optimization/83965
9553 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9554 that grouped statements are part of a reduction chain. Return
9555 true if the statement is not marked as a reduction itself but
9557 (vect_recog_dot_prod_pattern): Don't check whether the statement
9558 is part of a group here.
9559 (vect_recog_sad_pattern): Likewise.
9560 (vect_recog_widen_sum_pattern): Likewise.
9562 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9565 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9567 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9570 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9572 (ix86_output_indirect_function_return): New prototype.
9573 (ix86_split_simple_return_pop_internal): Likewise.
9574 * config/i386/i386.c (indirect_return_via_cx): New.
9575 (indirect_return_via_cx_bnd): Likewise.
9576 (indirect_thunk_name): Handle return va CX_REG.
9577 (output_indirect_thunk_function): Create alias for
9578 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9579 (ix86_output_indirect_jmp): Remove the bool argument.
9580 (ix86_output_indirect_function_return): New function.
9581 (ix86_split_simple_return_pop_internal): Likewise.
9582 * config/i386/i386.md (*indirect_jump): Don't pass false
9583 to ix86_output_indirect_jmp.
9584 (*tablejump_1): Likewise.
9585 (simple_return_pop_internal): Change it to define_insn_and_split.
9586 Call ix86_split_simple_return_pop_internal to split it for
9588 (simple_return_indirect_internal): Call
9589 ix86_output_indirect_function_return instead of
9590 ix86_output_indirect_jmp.
9592 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9595 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9596 memset and value initialization afterwards.
9598 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9600 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9602 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9605 * common/config/aarch64/aarch64-common.c
9606 (aarch_option_optimization_table[]): Switch
9607 off fomit-frame-pointer
9609 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9610 Chung-Ju Wu <jasonwucj@gmail.com>
9612 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9614 (store_multiple): Ditto.
9616 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9618 * config.gcc: Add --with-cpu support for nds32 target.
9619 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9620 * config/nds32/nds32.opt: Add -mcpu= option.
9622 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9624 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9625 isel=yes): Warn for these deprecated options.
9627 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9629 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9630 ISA_2_5_MASKS_EMBEDDED.
9632 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9634 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9635 p->max as pointers rather than using iterative_hash_expr.
9637 2018-02-23 Carl Love <cel@us.ibm.com>
9639 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9640 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9642 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9643 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9644 P8V_BUILTIN_VEC_VUNSIGNED2.
9646 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9649 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9650 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9652 * lra-constraints.c (curr_insn_transform): Set up
9653 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9655 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9656 * lra-eliminations.c (spill_pseudos): Ditto.
9657 (process_insn_for_elimination): Ditto.
9658 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9659 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9660 LRA_NON_CLOBBERED_ALT.
9662 2018-02-22 Martin Sebor <msebor@redhat.com>
9664 PR tree-optimization/84480
9665 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9666 to maybe_diag_stxncpy_trunc. Call it.
9667 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9668 from gimple_fold_builtin_strcpy. Print inlining stack.
9669 (handle_builtin_stxncpy): Print inlining stack.
9670 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9672 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9675 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9676 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9677 and -fcheck-pointer-bounds are used together.
9678 (indirect_thunk_prefix): New enum.
9679 (indirect_thunk_need_prefix): New function.
9680 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9681 "_nt" instead of "_bnd" for NOTRACK prefix.
9682 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9683 (output_indirect_thunk_function): Likewise.
9685 (ix86_code_end): Update output_indirect_thunk_function calls.
9686 (ix86_output_indirect_branch_via_reg): Replace
9687 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9688 (ix86_output_indirect_branch_via_push): Likewise.
9689 (ix86_output_function_return): Likewise.
9690 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9691 incompatible with -fcf-protection=branch and
9692 -fcheck-pointer-bounds.
9694 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9697 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9698 Change gcc_assert call to output_operand_lossage.
9700 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9702 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9704 2018-02-22 DJ Delorie <dj@redhat.com>
9705 Sebastian Perta <sebastian.perta@renesas.com>
9706 Oleg Endo <olegendo@gcc.gnu.org>
9708 * config/rx/rx.c (rx_rtx_costs): New function.
9709 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9711 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9713 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9715 2018-02-22 Martin Liska <mliska@suse.cz>
9718 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9719 Add "native" as a possible value.
9721 2018-02-22 Martin Liska <mliska@suse.cz>
9724 * config/i386/i386.c (ix86_option_override_internal):
9725 Add "native" as a possible value for -march and -mtune.
9727 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9730 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9731 to all type variants.
9733 PR tree-optimization/84503
9734 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9735 width as info->bitpos + info->bitsize - start.
9736 (merged_store_group::merge_overlapping): Simplify width computation.
9737 (check_no_overlap): New function.
9738 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9739 start + width and last_order of the group, fail if check_no_overlap
9741 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9742 to group if check_no_overlap fails.
9744 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9746 * config/rs6000/altivec.md: Delete contraint arguments to
9747 define_expand, define_split, and define_peephole2, and in
9748 define_insn_and_split if always unused.
9749 * config/rs6000/darwin.md: Ditto.
9750 * config/rs6000/dfp.md: Ditto.
9751 * config/rs6000/rs6000.md: Ditto.
9752 * config/rs6000/sync.md: Ditto.
9753 * config/rs6000/vector.md: Ditto.
9754 * config/rs6000/vsx.md: Ditto.
9756 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9758 * config/rs6000/altivec.md: Write output control strings as braced
9759 blocks instead of double-quoted strings.
9760 * config/rs6000/darwin.md: Ditto.
9761 * config/rs6000/rs6000.md: Ditto.
9762 * config/rs6000/vector.md: Ditto.
9763 * config/rs6000/vsx.md: Ditto.
9765 2018-02-21 Jason Merrill <jason@redhat.com>
9767 PR c++/84314 - ICE with templates and fastcall attribute.
9768 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9770 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9772 * ipa-cp.c (determine_versionability): Fix comment typos.
9774 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9777 * ipa-cp.c (determine_versionability): Do not version functions caling
9780 2018-02-21 Martin Liska <mliska@suse.cz>
9783 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9784 Add "native" as a possible value.
9785 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9786 the macro when native cpu detection is available.
9788 2018-02-21 Martin Liska <mliska@suse.cz>
9791 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9792 Add "native" as a possible value.
9793 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9794 when native cpu detection is available.
9796 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9797 Martin Sebor <msebor@redhat.com>
9799 PR tree-optimization/84478
9800 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9802 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9803 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9804 support which is conservatively correct, for 2 only stay conservative
9805 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9806 argument to the 2 argument get_range_strlen, adjust 6 arg
9807 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9809 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9810 (gimple_fold_builtin_strlen): Pass true as last argument to
9813 2018-02-20 Martin Sebor <msebor@redhat.com>
9816 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9817 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9818 (builtin_memref::builtin_memref): Factor out parts into
9819 set_base_and_offset and call it.
9821 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9824 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9825 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9826 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9827 search at the associated MODE_INT.
9829 2018-02-20 Jeff Law <law@redhat.com>
9832 PR tree-optimization/81592
9834 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9835 for range data rather than using global data.
9836 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9837 range data rather than using global data.
9838 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9839 pass it to children as needed.
9840 (struct directive::fmtresult): Similarly.
9841 (struct directive::set_width): Similarly.
9842 (struct directive::set_precision): Similarly.
9843 (format_integer, format_directive, parse_directive): Similarly.
9844 (format_none): Accept unnamed vr_values parameter.
9845 (format_percent, format_floating, format_character): Similarly.
9846 (format_string, format_plain): Similarly.
9847 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9848 the EVRP range analyzer for range data rather than using global data.
9849 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9850 gimple-ssa-evrp-analyze.h
9851 (class sprintf_dom_walker): Add after_dom_children member function.
9852 Add evrp_range_analyzer member.
9853 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9854 range analyzer as needed.
9855 (sprintf_dom_walker::after_dom_children): New member function.
9856 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9858 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9859 (evrp_range_analyzer::pop_to_marker): Likewise.
9861 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9863 PR tree-optimization/84419
9864 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9865 with the required type if its current type is compatible but
9868 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9871 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9872 after vectorization.
9874 2018-02-20 Martin Liska <mliska@suse.cz>
9877 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9878 possible values if we don't have a hint.
9880 2018-02-20 Martin Liska <mliska@suse.cz>
9884 * final.c (shorten_branches): Build align_tab array with one
9886 * opts.c (finish_options): Add alignment option limit check.
9887 (MAX_CODE_ALIGN): Likewise.
9888 (MAX_CODE_ALIGN_VALUE): Likewise.
9889 * doc/invoke.texi: Document maximum allowed option value for
9890 all -falign-* options.
9892 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9895 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9896 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9897 * var-tracking.c (emit_note_insn_var_location): Remove all references
9898 to NOTE_INSN_CALL_ARG_LOCATION.
9899 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9900 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9901 Use copy_rtx_if_shared.
9902 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9903 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9904 (dwarf2out_var_location): Remove handling of
9905 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9907 * final.c (final_scan_insn): Remove all references to
9908 NOTE_INSN_CALL_ARG_LOCATION.
9909 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9910 before dumping final insns.
9911 * except.c (emit_note_eh_region_end): Remove all references to
9912 NOTE_INSN_CALL_ARG_LOCATION.
9913 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9914 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9915 * config/arc/arc.c (hwloop_optimize): Likewise.
9916 * config/arm/arm.c (create_fix_barrier): Likewise.
9917 * config/s390/s390.c (s390_chunkify_start): Likewise.
9918 * config/sh/sh.c (find_barrier): Likewise.
9919 * config/i386/i386.c (rest_of_insert_endbranch,
9920 ix86_seh_fixup_eh_fallthru): Likewise.
9921 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9922 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9923 * config/frv/frv.c (frv_function_prologue): Likewise.
9924 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9926 (note_outside_basic_block_p): Remove all references to
9927 NOTE_INSN_CALL_ARG_LOCATION.
9928 * gengtype.c (adjust_field_rtx_def): Likewise.
9929 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9931 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9932 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9935 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9938 PR tree-optimization/84452
9939 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9940 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9943 2018-02-19 Martin Liska <mliska@suse.cz>
9946 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9948 2018-02-19 Martin Liska <mliska@suse.cz>
9949 Richard Sandiford <richard.sandiford@linaro.org>
9951 PR tree-optimization/82491
9952 * gimple-fold.c (get_base_constructor): Make earlier bail out
9955 2018-02-19 Carl Love <cel@us.ibm.com>
9957 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9958 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9960 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9961 P8V_BUILTIN_VEC_NEG.
9963 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9965 * config/rl78/rl78.md (movdf): New define expand.
9967 2018-02-19 Martin Liska <mliska@suse.cz>
9970 * doc/invoke.texi: Fix typo.
9971 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9973 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9976 handle rs6000_single_float and rs6000_double_float specially for
9979 2018-02-16 Jeff Law <law@redhat.com>
9981 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9984 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9987 * ipa-inline.c (inline_small_functions): Fix a typo.
9989 2018-02-16 Nathan Sidwell <nathan@acm.org>
9991 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9993 2018-02-16 Carl Love <cel@us.ibm.com>
9995 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9996 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9997 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9998 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9999 expansion to P8V_BUILTIN_VEC_FLOAT2.
10001 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10003 PR rtl-optimization/70023
10004 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10005 src_regno into account.
10007 2018-02-16 Carl Love <cel@us.ibm.com>
10009 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10010 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10011 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10012 * config/rs6000/rs6000.c: Remove case statements for
10013 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10014 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10015 and P9V_BUILTIN_VEC_VINSERT4B.
10016 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10017 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10018 * config/rs6000/vsx.md:
10019 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10022 2018-02-16 Carl Love <cel@us.ibm.com>
10024 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10026 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10028 * config/rs6000/rs6000-c.c: Add the definitions for
10029 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10030 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10031 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10032 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10033 definition for insert4b and define insn *insert3b_internal.
10034 * doc/extend.texi: Add documentation for vec_extract4b.
10036 2018-02-16 Nathan Sidwell <nathan@acm.org>
10038 * doc/extend.texi (Backwards Compatibility): Mention friend
10039 injection. Note for-scope is deprecated.
10040 * doc/invoke.texi (-ffriend-injection): Deprecate.
10042 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10044 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10045 that moved to I2, also allow destinations that are a paradoxical
10046 subreg (instead of a normal reg).
10048 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10051 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10054 2018-02-16 Richard Biener <rguenther@suse.de>
10056 PR tree-optimization/84037
10057 PR tree-optimization/84016
10059 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10060 Adjust vec_construct for the fact we need additional higher latency
10061 128bit inserts for AVX256 and AVX512 vector builds.
10062 (ix86_add_stmt_cost): Scale vector construction cost for
10065 2018-02-16 Richard Biener <rguenther@suse.de>
10067 PR tree-optimization/84417
10068 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10069 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10070 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10072 2018-02-16 Martin Liska <mliska@suse.cz>
10075 * internal-fn.def (ASAN_CHECK): Set proper flags.
10076 (ASAN_MARK): Likewise.
10078 2018-02-16 Julia Koval <julia.koval@intel.com>
10080 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10081 from PTA_CANNONLAKE.
10083 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10086 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10087 Use ++iter rather than iter++ for std::list iterators.
10088 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10089 defer deleting them until all nodes in the forest are processed. Do
10090 free even leaf nodes. Change to_process into auto_vec.
10093 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10094 * vec.h (vec_default_construct): Use memset instead of placement new
10095 if BROKEN_VALUE_INITIALIZATION is defined.
10096 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10097 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10100 PR rtl-optimization/83723
10101 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10102 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10103 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10105 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10107 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10109 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10111 PR rtl-optimization/81443
10112 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10113 from inner REGs to paradoxical SUBREGs.
10115 2018-02-16 Richard Biener <rguenther@suse.de>
10117 PR tree-optimization/84399
10118 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10119 For operands we can analyze at their definition make sure we can
10120 analyze them at each use as well.
10122 2018-02-16 Richard Biener <rguenther@suse.de>
10124 PR tree-optimization/84190
10125 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10126 volatile accesses if the decl isn't volatile.
10128 2018-02-15 Jason Merrill <jason@redhat.com>
10130 PR c++/84314 - ICE with templates and fastcall attribute.
10131 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10132 TYPE_CANONICAL on an existing type.
10134 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10136 PR tree-optimization/84383
10137 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10138 dstoff nor call operand_equal_p if dstbase is NULL.
10140 PR tree-optimization/84334
10141 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10142 also a CONSTANT_CLASS_P, punt.
10144 2018-02-14 Jim Wilson <jimw@sifive.com>
10146 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10147 first SMALL_OPERAND check. New local min_second_step. Move assert
10148 to where locals are set. Add TARGET_RVC support.
10149 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10151 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10153 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10155 2018-02-14 Martin Sebor <msebor@redhat.com>
10157 PR tree-optimization/83698
10158 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10159 arrays constrain the offset range to their bounds.
10160 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10161 (builtin_access::overlap): Avoid setting the size of overlap if it's
10163 (maybe_diag_overlap): Also consider arrays when deciding what values
10164 of offsets to include in diagnostics.
10166 2018-02-14 Martin Sebor <msebor@redhat.com>
10169 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10170 that correspond to the kind of a declaration.
10172 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10175 * config/pa/pa.md: Load address of PIC label using the linkage table
10176 if the label is nonlocal.
10178 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10180 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10181 warning message if user requests -maltivec=be.
10182 * doc/invoke.texi: Document deprecation of -maltivec=be.
10184 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10187 * config/rs6000/rs6000-c.c: Update definitions for
10188 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10189 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10191 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10194 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10195 add _get_ssp intrinsics. Remove argument from
10196 __builtin_ia32_rdssp[d|q].
10197 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10198 * config/i386/i386-builtin.def: Remove argument from
10199 __builtin_ia32_rdssp[d|q].
10200 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10201 ix86_expand_special_args_builtin for _rdssp[d|q].
10202 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10203 Clear register before usage.
10204 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10205 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10207 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10209 PR tree-optimization/84357
10210 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10211 operand 1 of an ARRAY_REF too.
10213 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10216 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10217 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10219 (set_of_reg): New struct.
10220 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10221 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10222 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10224 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10225 Split into bitclr, bitset, bitinvert patterns if appropriate.
10226 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10227 use rx_fuse_in_memory_bitop.
10228 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10229 to named insn, correct maximum insn length.
10231 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10234 * machmode.def: Define a complex mode for PARTIAL_INT.
10235 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10237 * doc/rtl.texi: Document CSPImode.
10238 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10240 (msp430_hard_regno_nregs_with_padding): Likewise.
10242 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10245 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10247 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10249 PR rtl-optimization/84169
10250 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10251 we generated a parallel as new i3 and we split that to new i2 and i3
10252 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10253 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10254 those to i2, not i1. Partially rewrite this scan code.
10256 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10259 * stor-layout.c (place_field): For variable length fields, adjust
10260 offset_align afterwards not just based on the field's alignment,
10261 but also on the size.
10263 PR middle-end/84309
10264 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10265 of exps and logs in the use_exp2 case.
10267 2018-02-13 Jeff Law <law@redhat.com>
10269 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10270 entry for "vector".
10272 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10275 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10279 * common.opt (gas-loc-support, gas-locview-support): New.
10280 (ginline-points, ginternal-reset-location-views): New.
10281 * doc/invoke.texi: Document them. Use @itemx where intended.
10282 (gvariable-location-views): Adjust.
10283 * target.def (reset_location_view): New.
10284 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10285 (TARGET_RESET_LOCATION_VIEW): New.
10286 * doc/tm.texi: Rebuilt.
10287 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10288 (dwarf2out_default_as_locview_support): New.
10289 (output_asm_line_debug_info): Use option variables.
10290 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10291 (output_loc_list): Likewise.
10292 (add_high_low_attributes): Check option variables.
10293 Don't output entry view attribute in strict mode.
10294 (gen_inlined_subroutine_die): Check option variables.
10295 (dwarf2out_inline_entry): Likewise.
10296 (init_sections_and_labels): Likewise.
10297 (dwarf2out_early_finish): Likewise.
10298 (maybe_reset_location_view): New, from...
10299 (dwarf2out_var_location): ... here. Call it.
10300 * debug.h (dwarf2out_default_as_loc_support): Declare.
10301 (dwarf2out_default_as_locview_support): Declare.
10302 * hooks.c (hook_int_rtx_insn_0): New.
10303 * hooks.h (hook_int_rtx_insn_0): Declare.
10304 * toplev.c (process_options): Take -gas-loc-support and
10305 -gas-locview-support from dwarf2out. Enable
10306 -gvariable-location-views by default only with locview
10307 assembler support. Enable -ginternal-reset-location-views by
10308 default only if the target defines the corresponding hook.
10309 Enable -ginline-points by default if location views are
10310 enabled; force it disabled if statement frontiers are
10312 * tree-inline.c (expand_call_inline): Check option variables.
10313 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10315 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10317 PR tree-optimization/84321
10318 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10319 handling. Also check whether the anti-range contains any values
10320 that satisfy the mask; switch to a VR_RANGE if not.
10322 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10325 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10327 2018-02-13 Martin Jambor <mjambor@suse.cz>
10330 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10331 of call statements, also set location of a load to a temporary.
10333 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10335 * config/rl78/rl78.c (add_vector_labels): New function.
10336 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10337 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10338 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10339 which checks that no arguments are passed.
10340 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10341 * doc/extend.texi: Documentation for the new attribute.
10343 2018-02-13 Andreas Schwab <schwab@suse.de>
10345 * config/riscv/linux.h (CPP_SPEC): Define.
10347 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10350 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10351 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10352 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10353 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10354 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10355 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10356 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10357 temporarily for AES and PCLMUL builtins.
10359 PR tree-optimization/84339
10360 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10361 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10364 PR middle-end/84309
10365 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10366 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10367 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10369 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10371 * omp-simd-clone.h: New file.
10372 * omp-simd-clone.c: Include omp-simd-clone.h.
10373 (expand_simd_clones): No longer static.
10374 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10375 cgraph.h and omp-simd-clone.h.
10376 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10377 (vect_recog_widen_shift_pattern): Formatting fix.
10378 (vect_pattern_recog_1): Don't check optab for calls.
10381 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10382 operands[2] into a REG before using gen_lowpart on it.
10384 2018-02-12 Jeff Law <law@redhat.com>
10387 * config/sh/sh.c (find_barrier): Consider a sibling call
10390 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10391 successfully back substituting a reg.
10393 2018-02-12 Richard Biener <rguenther@suse.de>
10395 PR tree-optimization/84037
10396 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10397 parameter, move visited init to caller.
10398 (vect_slp_analyze_operations): Separate cost from validity
10399 check, initialize visited once for all instances.
10400 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10402 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10404 (vect_model_promotion_demotion_cost): Likewise.
10405 (vectorizable_bswap): Guard cost modeling with !slp_node
10406 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10408 (vectorizable_call): Likewise.
10409 (vectorizable_conversion): Likewise.
10410 (vectorizable_assignment): Likewise.
10411 (vectorizable_shift): Likewise.
10412 (vectorizable_operation): Likewise.
10413 (vectorizable_store): Likewise.
10414 (vectorizable_load): Likewise.
10415 (vectorizable_condition): Likewise.
10416 (vectorizable_comparison): Likewise.
10418 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10421 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10422 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10423 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10425 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10427 PR middle-end/83665
10428 * params.def (inline-min-speedup): Increase from 8 to 15.
10429 (max-inline-insns-auto): Decrease from 40 to 30.
10430 * ipa-split.c (consider_split): Add some buffer for function to
10431 be considered inlining candidate.
10432 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10435 2018-02-12 Richard Biener <rguenther@suse.de>
10437 PR tree-optimization/84037
10438 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10439 matched stmts if we cannot swap the non-matched ones.
10441 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10443 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10444 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10445 _mm_maskz_scalef_round_ss): New intrinsics.
10446 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10447 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10448 __builtin_ia32_scalefss_round): Remove.
10449 (__builtin_ia32_scalefsd_mask_round,
10450 __builtin_ia32_scalefss_mask_round): New intrinsics.
10451 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10452 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10453 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10454 "<round_constraint>")): Changed to ...
10455 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10456 "<round_scalar_constraint>")): ... this.
10457 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10458 %0, %1, %2<round_op3>}"): Changed to ...
10459 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10460 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10461 %2<round_scalar_mask_op3>}"): ... this.
10462 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10464 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10466 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10467 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10468 (_mm_maskz_sqrt_round_ss): New intrinsics.
10469 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10470 (__builtin_ia32_sqrtsd_mask_round)
10471 (__builtin_ia32_sqrtss_mask_round): New builtins.
10472 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10473 (__builtin_ia32_sqrtss_round): Remove.
10474 (__builtin_ia32_sqrtsd_mask_round)
10475 (__builtin_ia32_sqrtss_mask_round): New builtins.
10476 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10477 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10478 ((match_operand:VF_128 1 "vector_operand"
10479 "xBm,<round_constraint>")): Changed to ...
10480 ((match_operand:VF_128 1 "vector_operand"
10481 "xBm,<round_scalar_constraint>")): ... this.
10482 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10483 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10484 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10485 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10486 %<iptr>1<round_scalar_mask_op3>}): ... this.
10487 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10488 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10490 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10493 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10494 Cast vec_cmpeq result to correct type.
10495 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10496 Cast vec_cmpgt result to correct type.
10498 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10500 * final.c (final_scan_insn_1): Renamed from...
10501 (final_scan_insn): ... this. New wrapper, to recover
10502 seen from the outermost call in recursive ones.
10503 * config/sparc/sparc.c (output_return): Drop seen from call.
10504 (output_sibcall): Likewise.
10505 * config/visium/visium.c (output_branch): Likewise.
10507 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10509 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10512 2018-02-10 Alan Modra <amodra@gmail.com>
10515 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10516 Specify LR as an input.
10518 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10521 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10522 remove_member_access_dummy_vars): New functions.
10523 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10524 lower_omp_1, execute_lower_omp): Use them.
10526 PR rtl-optimization/84308
10527 * shrink-wrap.c (spread_components): Release todo vector.
10529 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10531 PR rtl-optimization/57193
10532 * ira-color.c (struct allocno_color_data): Add member
10533 conflict_allocno_hard_prefs.
10534 (update_conflict_allocno_hard_prefs): New.
10535 (bucket_allocno_compare_func): Add a preference based on
10536 conflict_allocno_hard_prefs.
10537 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10538 (color_allocnos): Remove a dead code. Initiate
10539 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10541 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10544 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10545 constraint from =wa to wa. Avoid a subreg on the output operand,
10546 instead use a pseudo and subreg it in a move.
10547 (p9_xxbrd_<mode>): Changed to ...
10548 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10549 (p9_xxbrd_v2df): New expander.
10550 (p9_xxbrw_<mode>): Changed to ...
10551 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10552 (p9_xxbrw_v4sf): New expander.
10554 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10556 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10558 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10561 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10562 multiply in 32-bit mode.
10563 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10564 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10567 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10569 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10570 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10571 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10572 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10574 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10577 * dwarf2out.c (is_trivial_indirect_ref): New function.
10578 (dwarf2out_late_global_decl): Do not generate a location
10579 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10580 and that are not defined in the current unit.
10582 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10584 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10585 instead of a libcall for UNORDERED.
10587 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10590 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10591 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10593 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10596 * config/s390/s390.c (s390_set_current_function): Invoke
10597 s390_indirect_branch_settings also if fndecl didn't change.
10599 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10601 * config/rs6000/rs6000.md (blockage): Set length to zero.
10603 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10605 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10607 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10610 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10611 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10615 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10616 PARALLEL incoming that failed vt_get_decl_and_offset check.
10618 PR middle-end/84237
10619 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10620 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10622 (get_variable_section): For decls in named .bss* sections pass true as
10623 second argument to bss_initializer_p.
10625 2018-02-09 Marek Polacek <polacek@redhat.com>
10626 Jakub Jelinek <jakub@redhat.com>
10629 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10630 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10631 Sync some changes from cxx_fold_indirect_ref.
10633 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10635 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10637 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10638 (BLOCK_INLINE_ENTRY_LABEL): New.
10639 (dwarf2out_var_location): Disregard inline entry markers.
10640 (inline_entry_data): New struct.
10641 (inline_entry_data_hasher): New hashtable type.
10642 (inline_entry_data_hasher::hash): New.
10643 (inline_entry_data_hasher::equal): New.
10644 (inline_entry_data_table): New variable.
10645 (add_high_low_attributes): Add DW_AT_entry_pc and
10646 DW_AT_GNU_entry_view attributes if a pending entry is found
10647 in inline_entry_data_table. Add old entry_pc attribute only
10648 if debug nonbinding markers are disabled.
10649 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10650 markers are enabled.
10651 (block_within_block_p, dwarf2out_inline_entry): New.
10652 (dwarf2out_finish): Check that no entries remained in
10653 inline_entry_data_table.
10654 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10655 (final_scan_insn, notice_source_line): Likewise.
10656 (rest_of_clean_state): Skip inline entry markers.
10657 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10659 * gimple.c (gimple_build_debug_inline_entry): New.
10660 * gimple.h (enum gimple_debug_subcode): Add
10661 GIMPLE_DEBUG_INLINE_ENTRY.
10662 (gimple_build_debug_inline_entry): Declare.
10663 (gimple_debug_inline_entry_p): New.
10664 (gimple_debug_nonbind_marker_p): Adjust.
10665 * insn-notes.def (INLINE_ENTRY): New.
10666 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10667 inline entry marker notes.
10668 (print_insn): Likewise.
10669 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10670 (INSN_DEBUG_MARKER_KIND): Likewise.
10671 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10672 * tree-inline.c (expand_call_inline): Build and insert
10673 debug_inline_entry stmt.
10674 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10675 inline entry blocks early, if nonbind markers are enabled.
10676 (dump_scope_block): Dump fragment info.
10677 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10678 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10679 (gimple_build_debug_inline_entry): New.
10680 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10681 Enable/disable inline entry points too.
10682 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10683 (DEBUG_INSN): Describe inline entry markers.
10685 * common.opt (gvariable-location-views): New.
10686 (gvariable-location-views=incompat5): New.
10687 * config.in: Rebuilt.
10688 * configure: Rebuilt.
10689 * configure.ac: Test assembler for view support.
10690 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10691 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10692 * dwarf2out.c (var_loc_view): New typedef.
10693 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10694 (dwarf2out_locviews_in_attribute): New.
10695 (dwarf2out_locviews_in_loclist): New.
10696 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10697 (enum dw_line_info_opcode): Add LI_adv_address.
10698 (struct dw_line_info_table): Add view.
10699 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10700 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10701 (zero_view_p): New variable.
10702 (ZERO_VIEW_P): New macro.
10703 (output_asm_line_debug_info): New.
10704 (struct var_loc_node): Add view.
10705 (add_AT_view_list, AT_loc_list): New.
10706 (add_var_loc_to_decl): Add view param. Test it against last.
10707 (new_loc_list): Add view params. Record them.
10708 (AT_loc_list_ptr): Handle loc and view lists.
10709 (view_list_to_loc_list_val_node): New.
10710 (print_dw_val): Handle dw_val_class_view_list.
10711 (size_of_die): Likewise.
10712 (value_format): Likewise.
10713 (loc_list_has_views): New.
10714 (gen_llsym): Set vl_symbol too.
10715 (maybe_gen_llsym, skip_loc_list_entry): New.
10716 (dwarf2out_maybe_output_loclist_view_pair): New.
10717 (output_loc_list): Output view list or entries too.
10718 (output_view_list_offset): New.
10719 (output_die): Handle dw_val_class_view_list.
10720 (output_dwarf_version): New.
10721 (output_compilation_unit_header): Use it.
10722 (output_skeleton_debug_sections): Likewise.
10723 (output_rnglists, output_line_info): Likewise.
10724 (output_pubnames, output_aranges): Update version comments.
10725 (output_one_line_info_table): Output view numbers in asm comments.
10726 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10727 Call maybe_gen_llsym.
10728 (loc_list_from_tree_1): Adjust.
10729 (add_AT_location_description): Create view list attribute if
10730 needed, check it's absent otherwise.
10731 (convert_cfa_to_fb_loc_list): Adjust.
10732 (maybe_emit_file): Call output_asm_line_debug_info for test.
10733 (dwarf2out_var_location): Reset views as needed. Precompute
10734 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10735 attribute. Set view.
10736 (new_line_info_table): Reset next view.
10737 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10738 (dwarf2out_source_line): Likewise. Output view resets and labels to
10739 the assembler, or select appropriate line info opcodes.
10740 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10741 (optimize_string_length): Catch it. Adjust.
10742 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10743 dw_val_class_view_list, and remove it if no longer needed.
10744 (hash_loc_list): Hash view numbers.
10745 (loc_list_hasher::equal): Compare them.
10746 (optimize_location_lists): Check whether a view list symbol is
10747 needed, and whether the locview attribute is present, and
10748 whether they match. Remove the locview attribute if no longer
10750 (index_location_lists): Call skip_loc_list_entry for test.
10751 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10752 Use output_dwarf_version.
10753 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10754 (struct dw_val_node): Add val_view_list.
10755 * final.c (SEEN_NEXT_VIEW): New.
10756 (set_next_view_needed): New.
10757 (clear_next_view_needed): New.
10758 (maybe_output_next_view): New.
10759 (final_start_function): Rename to...
10760 (final_start_function_1): ... this. Take pointer to FIRST,
10761 add SEEN parameter. Emit param bindings in the initial view.
10762 (final_start_function): Reintroduce SEEN-less interface.
10763 (final): Rename to...
10764 (final_1): ... this. Take SEEN parameter. Output final pending
10765 next view at the end.
10766 (final): Reintroduce seen-less interface.
10767 (final_scan_insn): Output pending next view before switching
10768 sections or ending a block. Mark the next view as needed when
10769 outputting variable locations. Notify debug backend of section
10770 changes, and of location view changes.
10771 (rest_of_handle_final): Adjust.
10772 * toplev.c (process_options): Autodetect value for debug variable
10773 location views option. Warn on incompat5 without -gdwarf-5.
10774 * doc/invoke.texi (gvariable-location-views): New.
10775 (gvariable-location-views=incompat5): New.
10776 (gno-variable-location-views): New.
10778 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10780 PR tree-optimization/84136
10781 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10782 that the result of find_edge is non-NULL.
10784 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10787 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10788 storing integer register in SImode. Fix cost of 256 and 512
10789 byte aligned SSE register store.
10791 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10793 * config/i386/i386.c (ix86_multiplication_cost): Fix
10794 multiplication cost for TARGET_AVX512DQ.
10796 2018-02-08 Marek Polacek <polacek@redhat.com>
10798 PR tree-optimization/84238
10799 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10802 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10804 PR tree-optimization/84265
10805 * tree-vect-stmts.c (vectorizable_store): Don't treat
10806 VMAT_CONTIGUOUS accesses as grouped.
10807 (vectorizable_load): Likewise.
10809 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10811 PR tree-optimization/81635
10812 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10813 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10814 (test_round_for_mask): New functions.
10815 (wide_int_cc_tests): Call test_round_for_mask.
10816 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10817 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10818 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10819 range returned by get_range_info.
10821 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10824 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10825 * symtab.c: Include builtins.h
10826 (symtab_node::output_to_lto_symbol_table_p): Move here
10827 from lto-streamer-out.c:output_symbol_p.
10828 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10829 (output_symbol_p): Move all logic to symtab.c
10830 (produce_symtab): Update.
10832 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10834 * config/s390/s390-opts.h (enum indirect_branch): Define.
10835 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10836 (s390_indirect_branch_via_thunk)
10837 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10838 (enum s390_indirect_branch_type): Define.
10839 * config/s390/s390.c (struct s390_frame_layout, struct
10840 machine_function): Remove.
10841 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10842 (indirect_branch_table_label_no, indirect_branch_table_name):
10844 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10845 (enum s390_indirect_branch_option): Define.
10846 (s390_return_addr_from_memory): New function.
10847 (s390_handle_string_attribute): New function.
10848 (s390_attribute_table): Add new attribute handler.
10849 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10850 (s390_indirect_branch_via_thunk): New function.
10851 (s390_indirect_branch_via_inline_thunk): New function.
10852 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10853 sibling call optimization for non z10 compiles.
10854 (s390_emit_call): Force indirect branch target to be a single
10855 register. Add r1 clobber for non-z10 compiles.
10856 (s390_emit_epilogue): Emit return jump via return_use expander.
10857 (s390_reorg): Handle JUMP_INSNs as execute targets.
10858 (s390_option_override_internal): Perform validity checks for the
10859 new command line options.
10860 (s390_indirect_branch_attrvalue): New function.
10861 (s390_indirect_branch_settings): New function.
10862 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10863 (s390_output_indirect_thunk_function): New function.
10864 (s390_code_end): Implement target hook.
10865 (s390_case_values_threshold): Implement target hook.
10866 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10868 * config/s390/s390.h (struct s390_frame_layout)
10869 (struct machine_function): Move here from s390.c.
10870 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10871 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10872 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10873 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10874 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10875 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10876 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10877 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10878 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10879 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10880 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10881 (mnemonic attribute): Add values which aren't recognized
10883 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10884 pattern for branch conversion. Fix mnemonic attribute.
10885 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10886 indirect branch via thunk if requested.
10887 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10888 ("*indirect_jump"): Disable for branch conversion using out of
10890 ("indirect_jump_via_thunk<mode>_z10")
10891 ("indirect_jump_via_thunk<mode>")
10892 ("indirect_jump_via_inlinethunk<mode>_z10")
10893 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10894 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10895 ("casesi_jump_via_inlinethunk<mode>_z10")
10896 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10897 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10898 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10899 ("*indirect2_jump"): Disable for branch conversion.
10900 ("casesi_jump"): Turn into expander and expand patterns for branch
10902 ("return_use"): New expander.
10903 ("*return"): Emit return via thunk and rename it to ...
10904 ("*return<mode>"): ... this one.
10905 * config/s390/s390.opt: Add new options and and enum for the
10908 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10910 * lra-constraints.c (match_reload): Unconditionally use
10911 gen_lowpart_SUBREG, rather than selecting between that
10912 and equivalent gen_rtx_SUBREG code.
10914 2018-02-08 Richard Biener <rguenther@suse.de>
10916 PR tree-optimization/84233
10917 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10918 changed flag instead of boguously re-using phi_inserted.
10920 2018-02-08 Martin Jambor <mjambor@suse.cz>
10922 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10923 static local variables.
10925 2018-02-08 Richard Biener <rguenther@suse.de>
10927 PR tree-optimization/84278
10928 * tree-vect-stmts.c (vectorizable_store): When looking for
10929 smaller vector types to perform grouped strided loads/stores
10930 make sure the mode is supported by the target.
10931 (vectorizable_load): Likewise.
10933 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10935 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10936 Increase LDP/STP opportunities by adding adjacent callee-saves.
10938 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10940 PR rtl-optimization/84068
10941 PR rtl-optimization/83459
10942 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10944 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10946 PR tree-optimization/84224
10947 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10948 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10949 non-zero arguments.
10951 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10954 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10955 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10956 count, remove one USE.
10958 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10960 * doc/install.texi (Configuration): Document the
10961 --with-long-double-format={ibm,ieee} PowerPC configuration
10965 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10966 Convert from define_expand to be define_insn_and_split. Rework
10967 float/double/_Float128 conversions to QI/HI/SImode to work with
10968 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10969 conversions to QI/HImode types did a store and then a load to
10970 truncate the value. For conversions to VSX registers, don't split
10971 the insn, instead emit the code directly. Use the code iterator
10972 any_fix to combine signed and unsigned conversions.
10973 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10974 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10975 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10976 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10977 (fix_<mode>di2_hw): Likewise.
10978 (fixuns_<mode>di2_hw): Likewise.
10979 (fix_<mode>si2_hw): Likewise.
10980 (fixuns_<mode>si2_hw): Likewise.
10981 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10982 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10983 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10984 fix<uns>_trunc<SFDF:mode>si2_p8.
10985 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10986 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10987 (fix<uns>_<mode>_mem): Likewise.
10988 (fctiw<u>z_<mode>_mem): Likewise.
10989 (fix<uns>_<mode>_mem): Likewise.
10990 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10991 the register allocator from doing a direct move to the GPRs to do
10992 a store, and instead use the ISA 3.0 store byte/half-word from
10993 vector register instruction. For IEEE 128-bit floating point,
10994 also optimize stores of 32-bit ints.
10995 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10997 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10999 * genextract.c (push_pathstr_operand): New function to support
11001 (walk_rtx): Call push_pathstr_operand.
11002 (print_path): Support [a-zA-Z].
11004 2018-02-07 Richard Biener <rguenther@suse.de>
11006 PR tree-optimization/84037
11007 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11008 (cse_and_gimplify_to_preheader): Declare.
11009 (vect_get_place_in_interleaving_chain): Likewise.
11010 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11012 (_loop_vec_info::~_loop_vec_info): Delete it.
11013 (cse_and_gimplify_to_preheader): New function.
11014 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11015 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11016 (vectorizable_load): Likewise. For grouped stores always base
11017 the IV on the first element.
11018 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11019 condition before gimplifying.
11021 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11023 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11024 *DIV_EXPR and *MOD_EXPR.
11026 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11029 * config/i386/i386.c (ix86_option_override_internal): Mask out
11030 the CF_SET bit when checking -fcf-protection.
11032 2018-02-07 Tom de Vries <tom@codesourcery.com>
11035 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11038 2018-02-07 Richard Biener <rguenther@suse.de>
11040 PR tree-optimization/84204
11041 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11044 PR tree-optimization/84205
11045 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11046 special-case isl_ast_op_zdiv_r.
11048 PR tree-optimization/84223
11049 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11050 Only add conditions from within the region.
11051 (gather_bbs::after_dom_children): Adjust.
11053 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11056 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11057 * config/avr/avr.md: Only post-reload split REG-REG moves if
11058 either register is GENERAL_REG_P.
11060 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11062 PR tree-optimization/84235
11063 * tree-ssa-scopedtables.c
11064 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11065 if the subtraction is performed in floating point type where NaNs are
11066 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11067 build 1. Formatting fix.
11069 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11072 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11073 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11074 and skip it regardless of bb boundaries. Use CALL_P macro,
11075 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11076 unnecessarily, formatting fix.
11078 2018-02-06 Michael Collison <michael.collison@arm.com>
11080 * config/arm/thumb2.md:
11081 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11082 (*thumb_mov_notscc): Ditto.
11084 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11087 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11088 unsigned_fix, not "s".
11090 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11092 * configure.ac (gcc_fn_eh_frame_ro): New function.
11093 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11094 correct .eh_frame permissions.
11095 * configure: Regenerate.
11097 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11099 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11100 irrelevant options.
11102 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11104 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11105 Display warning message for -mno-speculate-indirect-jumps.
11107 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11109 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11111 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11113 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11115 PR tree-optimization/84225
11116 * tree-eh.c (find_trapping_overflow): Only call
11117 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11119 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11122 * config/i386/i386.c: Reimplement the check of possible options
11123 -mibt/-mshstk conbination. Change error messages.
11124 * doc/invoke.texi: Fix a typo: remove extra '='.
11126 2018-02-06 Marek Polacek <polacek@redhat.com>
11128 PR tree-optimization/84228
11129 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11131 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11134 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11135 emitted arch directives.
11136 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11137 __ARM_FEATURE_COPROC before changing architectures.
11139 2018-02-06 Richard Biener <rguenther@suse.de>
11141 * config/i386/i386.c (print_reg): Fix typo.
11142 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11144 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11146 * configure: Regenerate.
11148 2018-02-05 Martin Sebor <msebor@redhat.com>
11150 PR tree-optimization/83369
11151 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11154 2018-02-05 Martin Liska <mliska@suse.cz>
11156 * doc/invoke.texi: Cherry-pick upstream r323995.
11158 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11160 * ira.c (ira_init_register_move_cost): Adjust comment.
11162 2018-02-05 Martin Liska <mliska@suse.cz>
11164 PR gcov-profile/84137
11165 * doc/gcov.texi: Fix typo in documentation.
11167 2018-02-05 Martin Liska <mliska@suse.cz>
11169 PR gcov-profile/83879
11170 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11171 using dlopen functionality.
11173 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11175 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11176 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11177 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11178 _mm_maskz_range_round_ss): New intrinsics.
11179 (__builtin_ia32_rangesd128_round)
11180 (__builtin_ia32_rangess128_round): Remove.
11181 (__builtin_ia32_rangesd128_mask_round,
11182 __builtin_ia32_rangess128_mask_round): New builtins.
11183 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11184 __builtin_ia32_rangess128_round): Remove.
11185 (__builtin_ia32_rangesd128_mask_round,
11186 __builtin_ia32_rangess128_mask_round): New builtins.
11187 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11188 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11189 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11190 "<round_saeonly_constraint>")): Changed to ...
11191 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11192 "<round_saeonly_scalar_constraint>")): ... this.
11193 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11194 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11195 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11196 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11197 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11199 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11201 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11203 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11204 Remove all values except native, 8540 and 8548.
11206 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11208 * config/i386/i386.c (ix86_output_function_return): Pass
11209 INVALID_REGNUM, instead of -1, as invalid register number to
11210 indirect_thunk_name and output_indirect_thunk.
11212 2018-02-02 Julia Koval <julia.koval@intel.com>
11214 * config.gcc: Add -march=icelake.
11215 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11216 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11217 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11218 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11219 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11220 (processor_target_table): Add icelake.
11221 (ix86_option_override_internal): Handle new PTAs.
11222 (get_builtin_code_for_version): Handle icelake.
11223 (M_INTEL_COREI7_ICELAKE): New.
11224 (fold_builtin_cpu): Handle icelake.
11225 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11226 * doc/invoke.texi: Add -march=icelake.
11228 2018-02-02 Julia Koval <julia.koval@intel.com>
11230 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11231 to wide_int_bitmask.
11232 * wide-int-bitmask.h: New.
11234 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11237 * config/i386/i386.md: Replace Pmode with word_mode in
11238 builtin_setjmp_setup and builtin_longjmp to support x32.
11240 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11244 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11246 (rs6000_supported_cpu_names): New static variable.
11247 (linux_cpu_translation_table): Likewise.
11248 (elf_platform) <cpu>: Define new static variable and use it.
11249 Translate kernel AT_PLATFORM name to canonical name if needed.
11250 Error if platform name is unknown.
11252 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11255 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11257 2018-02-01 Jeff Law <law@redhat.com>
11260 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11261 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11262 the scratch if RELEASE_VIA_POP is false.
11263 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11264 If we have to save a temporary register, decrement SIZE appropriately.
11265 Pass new arguments to release_scratch_register_on_entry.
11266 (ix86_adjust_stack_and_probe): Likewise.
11267 (ix86_emit_probe_stack_range): Pass new arguments to
11268 release_scratch_register_on_entry.
11270 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11272 PR rtl-optimization/84157
11273 * combine.c (change_zero_ext): Use REG_P predicate in
11274 front of HARD_REGISTER_P predicate.
11276 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11278 * config/avr/avr.c (avr_option_override): Move disabling of
11279 -fdelete-null-pointer-checks to...
11280 * common/config/avr/avr-common.c (avr_option_optimization_table):
11283 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11285 PR tree-optimization/81635
11286 * tree-data-ref.c (split_constant_offset_1): For types that
11287 wrap on overflow, try to use range info to prove that wrapping
11290 2018-02-01 Renlin Li <renlin.li@arm.com>
11293 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11294 TAILCALL_ADDR_REGS.
11295 (aarch64_register_move_cost): Likewise.
11296 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11297 TAILCALL_ADDR_REGS.
11298 (REG_CLASS_NAMES): Likewise.
11299 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11300 TAILCALL_ADDR_REGS. Remove IP registers.
11301 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11303 2018-02-01 Richard Biener <rguenther@suse.de>
11305 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11306 for specifying RPO order and allow NULL for that.
11307 * domwalk.c (dom_walker::dom_walker): Likewise.
11308 (dom_walker::walk): Handle NULL RPO order.
11309 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11311 (rewrite_update_dom_walker): Likewise.
11312 (mark_def_dom_walker): Likewise.
11314 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11316 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11317 (aarch64_maybe_expand_sve_subreg_move): Declare.
11318 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11319 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11321 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11322 that are semantically a reverse operation.
11323 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11324 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11325 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11327 (aarch64_can_change_mode_class): For big-endian, forbid changes
11328 between two SVE modes if they have different element sizes.
11330 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11332 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11333 the TImode handling for big-endian targets.
11335 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11337 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11338 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11340 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11341 Remove BSWAP handing for big-endian targets and use the form of
11342 LD1RQ appropariate for the mode.
11344 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11346 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11347 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11348 duplicated element.
11350 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11353 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11354 check for operands that need to go through aarch64_sve_reload_be.
11356 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11358 PR tree-optimization/81661
11359 PR tree-optimization/84117
11360 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11361 * tree-eh.c: Include gimplify.h.
11362 (find_trapping_overflow, replace_trapping_overflow,
11363 rewrite_to_non_trapping_overflow): New functions.
11364 * tree-vect-loop.c: Include tree-eh.h.
11365 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11366 * tree-data-ref.c: Include tree-eh.h.
11367 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11369 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11371 PR rtl-optimization/84123
11372 * combine.c (change_zero_ext): Check if hard register satisfies
11373 can_change_dest_mode before calling gen_lowpart_SUBREG.
11375 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11378 * ira.c (ira_init_register_move_cost): Remove assert.
11380 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11382 PR rtl-optimization/84071
11383 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11384 * doc/tm.texi: Regenerate.
11386 2018-01-31 Richard Biener <rguenther@suse.de>
11388 PR tree-optimization/84132
11389 * tree-data-ref.c (analyze_miv_subscript): Properly
11390 check whether evolution_function_is_affine_multivariate_p
11391 before calling gcd_of_steps_may_divide_p.
11393 2018-01-31 Julia Koval <julia.koval@intel.com>
11396 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11397 * config/i386/i386.md (rdpid_rex64) New.
11398 (rdpid): Make 32bit only.
11400 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11403 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11404 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11406 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11409 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11411 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11413 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11415 PR rtl-optimization/84071
11416 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11417 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11419 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11421 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11422 (arc_attribute_table): Add 'aux' attribute.
11423 (arc_in_small_data_p): Consider aux like variables.
11424 (arc_is_aux_reg_p): New function.
11425 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11426 (arc_get_aux_arg): New function.
11427 (prepare_move_operands): Handle aux-register access.
11428 (arc_handle_aux_attribute): New function.
11429 * doc/extend.texi (ARC Variable attributes): Add subsection.
11431 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11433 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11434 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11435 (arc_attribute_table): Add 'uncached' attribute.
11436 (arc_print_operand): Print '.di' flag for uncached memory
11438 (arc_in_small_data_p): Do not consider for small data the uncached
11440 (arc_is_uncached_mem_p): New function.
11441 * config/arc/predicates.md (compact_store_memory_operand): Check
11442 for uncached memory accesses.
11443 (nonvol_nonimm_operand): Likewise.
11444 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11446 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11449 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11450 falign-loops=): Add Optimization flag.
11452 2018-01-30 Jeff Law <law@redhat.com>
11455 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11456 INT_REGISTERS_SAVED. Check it prior to calling
11457 get_scratch_register_on_entry.
11458 (ix86_adjust_stack_and_probe): Similarly.
11459 (ix86_emit_probe_stack_range): Similarly.
11460 (ix86_expand_prologue): Corresponding changes.
11462 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11465 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11466 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11468 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11471 * lra-constraints.c (curr_insn_transform): Process AND in the
11474 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11476 PR rtl-optimization/83986
11477 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11478 dependence against last_pending_memory_flush in addition to
11479 pending_jump_insns.
11481 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11483 PR tree-optimization/81611
11484 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11487 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11490 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11493 2018-01-30 Richard Biener <rguenther@suse.de>
11494 Jakub Jelinek <jakub@redhat.com>
11496 PR tree-optimization/84111
11497 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11498 inner loops added during recursion, as they don't have up-to-date
11501 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11504 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11505 (can_inline_edge_by_limits_p): ... here.
11506 (can_early_inline_edge_p, check_callers,
11507 update_caller_keys, update_callee_keys, recursive_inlining,
11508 add_new_edges_to_heap, speculation_useful_p,
11509 inline_small_functions,
11510 inline_small_functions, flatten_function,
11511 inline_to_all_callers_1): Update.
11513 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11515 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11518 2018-01-30 Richard Biener <rguenther@suse.de>
11520 PR tree-optimization/83008
11521 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11522 invariant and constant vector uses in stmts when they need
11523 more than one stmt.
11525 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11528 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11529 * configure: Regenerate.
11531 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11533 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11535 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11536 Use gen_rtx_REG rather than gen_lowpart.
11538 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11540 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11541 rather than 0 when creating partial subregs.
11543 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11545 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11548 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11551 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11552 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11553 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11554 flags. This restores the settings used before the 2017-07-24.
11555 Turning off pre increment/decrement/modify allows IVOPTS to
11556 optimize DF/SF loops where the index is an int.
11558 2018-01-29 Richard Biener <rguenther@suse.de>
11559 Kelvin Nilsen <kelvin@gcc.gnu.org>
11562 * tree-vect-stmts.c (vectorizable_call): Don't call
11563 targetm.vectorize_builtin_md_vectorized_function if callee is
11566 2018-01-22 Carl Love <cel@us.ibm.com>
11568 * doc/extend.tex: Fix typo in second arg in
11569 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11571 2018-01-29 Richard Biener <rguenther@suse.de>
11573 PR tree-optimization/84086
11574 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11575 (flush_ssaname_freelist): When SSA names were released reset
11576 the SCEV hash table.
11578 2018-01-29 Richard Biener <rguenther@suse.de>
11580 PR tree-optimization/84057
11581 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11582 removed paths when removing edges.
11584 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11586 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11587 -mfunction-return=@var{choice}.
11589 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11591 PR diagnostic/84034
11592 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11593 Handle CR like TAB.
11594 (layout::print_source_line): Likewise.
11595 (test_get_line_width_without_trailing_whitespace): Add test cases.
11597 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11599 PR middle-end/84040
11600 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11603 2018-01-26 Jim Wilson <jimw@sifive.com>
11605 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11607 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11610 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11612 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11613 and CMP + SUB-immediate -> SUBS.
11615 2018-01-26 Martin Sebor <msebor@redhat.com>
11617 PR tree-optimization/83896
11618 * tree-ssa-strlen.c (get_string_len): Rename...
11619 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11620 Avoid assuming length is constant.
11621 (handle_char_store): Use HOST_WIDE_INT for string length.
11623 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11626 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11627 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11629 2018-01-26 Richard Biener <rguenther@suse.de>
11631 PR rtl-optimization/84003
11632 * dse.c (record_store): Only record redundant stores when
11633 the earlier store aliases at least all accesses the later one does.
11635 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11637 PR rtl-optimization/83985
11638 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11639 REG_CFA_RESTORE insns.
11640 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11641 REG_CFA_RESTORE insns here.
11644 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11645 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11647 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11649 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11650 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11651 (arc_init): Likewise.
11652 (arc_override_options): Likewise.
11653 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11655 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11657 * config/arc/arc.h (TARGET_DBNZ): Define.
11658 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11659 properly set the tune attribute.
11660 (dbnz): Use TARGET_DBNZ guard.
11661 * config/arc/arc.opt (mtune): Add core3 option.
11663 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11665 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11666 recognize new pic like addresses.
11667 (arc_delegitimize_address): Clean up.
11669 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11671 * config/arc/arc-arches.def: Option mrf16 valid for all
11673 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11674 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11675 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11676 * config/arc/arc-tables.opt: Regenerate.
11677 * config/arc/arc.c (arc_conditional_register_usage): Handle
11678 reduced register file case.
11679 (arc_file_start): Set must have build attributes.
11680 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11681 mrf16 option value.
11682 * config/arc/arc.opt (mrf16): Add new option.
11683 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11684 * config/arc/genmultilib.awk: Handle new mrf16 option.
11685 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11686 * config/arc/t-multilib: Regenerate.
11687 * doc/invoke.texi (ARC Options): Document mrf16 option.
11689 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11691 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11692 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11693 (arc_attribute_table): Add 'secure_call' attribute.
11694 (arc_print_operand): Print secure call operand.
11695 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11697 (arc_is_secure_call_p): New function. * config/arc/arc.md
11698 (call_i): Add support for sjli instruction.
11699 (call_value_i): Likewise.
11700 * config/arc/constraints.md (Csc): New constraint.
11702 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11703 John Eric Martin <John.Martin@emmicro-us.com>
11705 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11706 * config/arc/arc.c (_arc_jli_section): New struct.
11707 (arc_jli_section): New type.
11708 (rc_jli_sections): New static variable.
11709 (arc_handle_jli_attribute): New function.
11710 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11711 (arc_file_end): New function.
11712 (TARGET_ASM_FILE_END): Define.
11713 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11714 (arc_add_jli_section): New function.
11715 (jli_call_scan): Likewise.
11716 (arc_reorg): Call jli_call_scan.
11717 (arc_output_addsi): Remove 'S' from printing asm operand.
11718 (arc_is_jli_call_p): New function.
11719 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11721 (movhi_insn): Likewise.
11722 (movsi_insn): Likewise.
11723 (movsi_set_cc_insn): Likewise.
11724 (loadqi_update): Likewise.
11725 (load_zeroextendqisi_update): Likewise.
11726 (load_signextendqisi_update): Likewise.
11727 (loadhi_update): Likewise.
11728 (load_zeroextendhisi_update): Likewise.
11729 (load_signextendhisi_update): Likewise.
11730 (loadsi_update): Likewise.
11731 (loadsf_update): Likewise.
11732 (movsicc_insn): Likewise.
11733 (bset_insn): Likewise.
11734 (bxor_insn): Likewise.
11735 (bclr_insn): Likewise.
11736 (bmsk_insn): Likewise.
11737 (bicsi3_insn): Likewise.
11738 (cmpsi_cc_c_insn): Likewise.
11739 (movsi_ne): Likewise.
11740 (movsi_cond_exec): Likewise.
11741 (clrsbsi2): Likewise.
11742 (norm_f): Likewise.
11752 (call_i): Remove 'S' asm letter, add jli instruction.
11753 (call_value_i): Likewise.
11754 * config/arc/arc.op (mjli-always): New option.
11755 * config/arc/constraints.md (Cji): New constraint.
11756 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11758 (subsf3_fpx): Likewise.
11759 (mulsf3_fpx): Likewise.
11760 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11762 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11763 function attrbutes.
11764 * doc/invoke.texi (ARC): Document mjli-always option.
11766 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11768 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11769 avoid addition with 0 and use incw and decw where possible.
11771 2018-01-26 Richard Biener <rguenther@suse.de>
11773 PR tree-optimization/81082
11774 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11775 association if it requires casting to unsigned.
11776 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11777 from fold_plusminus_mult_expr to catch important cases late when
11778 range info is available.
11780 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11782 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11783 * configure.ac (hidden_linkonce): New test.
11784 * configure: Regenerate.
11785 * config.in: Regenerate.
11787 2018-01-26 Julia Koval <julia.koval@intel.com>
11789 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11790 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11791 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11792 _mm_mask_bitshuffle_epi64_mask): Fix type.
11793 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11794 USI_FTYPE_V4DI_V4DI_USI): Remove.
11795 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11796 __builtin_ia32_vpshufbitqmb256_mask,
11797 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11798 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11799 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11801 2018-01-26 Alan Modra <amodra@gmail.com>
11804 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11805 UNSPEC_VBPERMQ. Sort other unspecs.
11807 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11809 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11811 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11813 PR middle-end/83055
11814 * predict.c (drop_profile): Do not push/pop cfun; update also
11816 (handle_missing_profiles): Fix logic looking for zero profiles.
11818 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11820 PR middle-end/83977
11821 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11822 on functions with #pragma omp declare simd or functions with simd
11824 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11825 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11826 Remove trailing \n from warning_at calls.
11828 2018-01-25 Tom de Vries <tom@codesourcery.com>
11831 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11832 for neutered workers.
11834 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11837 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11838 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11840 2018-01-24 Jeff Law <law@redhat.com>
11843 * i386.c (get_probe_interval): Move to earlier point.
11844 (ix86_compute_frame_layout): If -fstack-clash-protection and
11845 the frame is larger than the probe interval, then use pushes
11846 to save registers rather than reg->mem moves.
11847 (ix86_expand_prologue): Remove conditional for int_registers_saved
11850 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11853 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11854 min/max for never referenced object.
11856 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11858 PR middle-end/83977
11859 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11861 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11862 attributes from DECL_ATTRIBUTES (decl) without affecting
11863 DECL_ATTRIBUTES (current_function_decl).
11864 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11865 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11867 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11869 PR tree-optimization/83979
11870 * fold-const.c (fold_comparison): Use constant_boolean_node
11871 instead of boolean_{true,false}_node.
11873 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11875 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11878 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11880 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11881 Simplify the clause that sets the length attribute.
11882 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11883 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11884 clause that sets the length attribute.
11885 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11887 2018-01-24 Tom de Vries <tom@codesourcery.com>
11890 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11891 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11892 Add strict parameter.
11893 (prevent_branch_around_nothing): Insert dummy insn between branch to
11894 label and label with no ptx insn inbetween.
11895 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11897 2018-01-24 Tom de Vries <tom@codesourcery.com>
11900 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11901 for neutered threads in warp.
11902 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11904 2018-01-24 Richard Biener <rguenther@suse.de>
11906 PR tree-optimization/83176
11907 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11910 2018-01-24 Richard Biener <rguenther@suse.de>
11912 PR tree-optimization/82819
11913 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11914 code generating pluses that are no-ops in the target precision.
11916 2018-01-24 Richard Biener <rguenther@suse.de>
11918 PR middle-end/84000
11919 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11921 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11923 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11924 to merge probabilities.
11925 * predict.c (probably_never_executed): Also mark as cold functions
11926 with global 0 profile and guessed local profile.
11927 * profile-count.c (profile_probability::combine_with_count): New
11929 * profile-count.h (profile_probability::operator*,
11930 profile_probability::operator*=, profile_probability::operator/,
11931 profile_probability::operator/=): Reduce precision to adjusted
11932 and set value to guessed on contradictory divisions.
11933 (profile_probability::combine_with_freq): Remove.
11934 (profile_probability::combine_wiht_count): Declare.
11935 (profile_count::force_nonzero):: Set to adjusted.
11936 (profile_count::probability_in):: Set quality to adjusted.
11937 * tree-ssa-tail-merge.c (replace_block_by): Use
11938 combine_with_count.
11940 2018-01-23 Andrew Waterman <andrew@sifive.com>
11941 Jim Wilson <jimw@sifive.com>
11943 * config/riscv/riscv.c (riscv_stack_boundary): New.
11944 (riscv_option_override): Set riscv_stack_boundary. Handle
11945 riscv_preferred_stack_boundary_arg.
11946 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11947 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11948 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11949 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11950 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11951 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11953 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11956 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11957 of struct ix86_frame.
11958 (ix86_expand_epilogue): Likewise. Add a local variable for
11959 the reg_save_offset field in struct ix86_frame.
11961 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11963 PR tree-optimization/82604
11964 * tree-loop-distribution.c (enum partition_kind): New enum item
11965 PKIND_PARTIAL_MEMSET.
11966 (partition_builtin_p): Support above new enum item.
11967 (generate_code_for_partition): Ditto.
11968 (compute_access_range): Differentiate cases that equality can be
11969 proven at all loops, the innermost loops or no loops.
11970 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11971 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11972 (finalize_partitions, distribute_loop): Don't fuse partition of
11973 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11974 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11975 parloop is enabled.
11977 2018-01-23 Martin Liska <mliska@suse.cz>
11979 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11980 order to ignore the predictor.
11981 (PRED_POLYMORPHIC_CALL): Likewise.
11982 (PRED_RECURSIVE_CALL): Likewise.
11984 2018-01-23 Martin Liska <mliska@suse.cz>
11986 * tree-profile.c (tree_profiling): Print function header to
11987 aware reader which function we are working on.
11988 * value-prof.c (gimple_find_values_to_profile): Do not print
11989 not interesting value histograms.
11991 2018-01-23 Martin Liska <mliska@suse.cz>
11993 * profile-count.h (enum profile_quality): Add
11994 profile_uninitialized as the first value. Do not number values
11995 as they are zero based.
11996 (profile_count::verify): Update sanity check.
11997 (profile_probability::verify): Likewise.
11999 2018-01-23 Nathan Sidwell <nathan@acm.org>
12001 * doc/invoke.texi (ffor-scope): Deprecate.
12003 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12005 PR tree-optimization/83510
12006 * domwalk.c (set_all_edges_as_executable): New function.
12007 (dom_walker::dom_walker): Convert bool param
12008 "skip_unreachable_blocks" to enum reachability. Move setup of
12009 edge flags to set_all_edges_as_executable and only do it when
12010 reachability is REACHABLE_BLOCKS.
12011 * domwalk.h (enum dom_walker::reachability): New enum.
12012 (dom_walker::dom_walker): Convert bool param
12013 "skip_unreachable_blocks" to enum reachability.
12014 (set_all_edges_as_executable): New decl.
12015 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12016 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12018 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12019 but converting true to REACHABLE_BLOCKS.
12020 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12022 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12023 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12024 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12026 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12027 if check_all_array_refs will be called.
12029 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12031 * tree.c (selftest::test_location_wrappers): Add more test
12034 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12036 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12037 (selftest::test_bit_in_range): Likewise.
12039 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12042 * doc/sourcebuild.texi (vect_float): Say that the selector
12043 only describes the situation when -funsafe-math-optimizations is on.
12044 (vect_float_strict): Document.
12046 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12048 PR tree-optimization/83965
12049 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12050 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12051 instead of checking only for a reduction.
12052 (vect_recog_widen_sum_pattern): Likewise.
12054 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12056 * predict.c (probably_never_executed): Only use precise profile info.
12057 (compute_function_frequency): Skip after inlining hack since we now
12058 have quality checking.
12060 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12062 * profile-count.h (profile_probability::very_unlikely,
12063 profile_probability::unlikely, profile_probability::even): Set
12064 precision to guessed.
12066 2018-01-23 Richard Biener <rguenther@suse.de>
12068 PR tree-optimization/83963
12069 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12070 Properly terminate dominator walk when crossing the exit edge not
12071 when visiting its source block.
12073 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12076 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12077 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12079 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12081 PR tree-optimization/83957
12082 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12083 semicolon after for body surrounded by braces.
12085 PR tree-optimization/83081
12086 * profile-count.h (profile_probability::split): New method.
12087 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12088 Use profile_probability::split.
12089 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12090 when splitting a single conditional jump into 2.
12092 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12094 PR tree-optimization/69452
12095 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12100 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12101 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12102 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12104 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12106 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12108 * config/rl78/rl78.md (movdi): New define_expand.
12109 * config/rl78/rl78.c (rl78_split_movdi): New function.
12111 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12114 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12116 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12117 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12118 128-bit to produce an UNSPEC move to get the double word with the
12119 signbit and then a shift directly to do signbit.
12120 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12121 implementation with a new version that just does either a direct
12122 move or a regular move. Move memory interface to separate insns.
12123 Move insns so they are next to the expander.
12124 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12125 with signbit move. Split big and little endian case.
12126 (signbit<mode>2_dm_mem_le): Likewise.
12127 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12128 (signbit<mode>2_dm2): Likewise.
12130 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12132 * config/rl78/rl78.md (anddi3): New define_expand.
12134 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12136 * config/rl78/rl78.md (umindi3): New define_expand.
12138 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12140 * config/rl78/rl78.md (smindi3): New define_expand.
12142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12144 * config/rl78/rl78.md (smaxdi3): New define_expand.
12146 2018-01-22 Carl Love <cel@us.ibm.com>
12148 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12149 LVX_V1TI): Add macro expansion.
12150 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12151 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12152 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12153 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12154 Change check to determine if the instruction is a byte reversing
12155 entry. Fix typo in comment.
12156 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12157 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12158 Add def_builtin calls for new builtins.
12159 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12160 Add define_insn expansion.
12162 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12164 * config/rl78/rl78.md (umaxdi3): New define_expand.
12166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12168 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12169 for non-QImode registers.
12171 2018-01-22 Richard Biener <rguenther@suse.de>
12173 PR tree-optimization/83963
12174 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12175 including the loop exit block.
12176 (scop_detection::merge_sese): Likewise.
12177 (scop_detection::add_scop): Do it here instead.
12179 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12181 * doc/sourcebuild.texi (arm_softfloat): Document.
12183 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12186 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12187 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12188 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12190 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12191 David Edelsohn <dje.gcc@gmail.com>
12194 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12195 Change "crset eq" to "crset 2".
12196 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12197 (*call_indirect_aix<mode>_nospec): Likewise.
12198 (*call_value_indirect_aix<mode>_nospec): Likewise.
12199 (*call_indirect_elfv2<mode>_nospec): Likewise.
12200 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12201 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12202 change assembly output from . to $.
12203 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12204 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12205 (*tablejump<mode>_internal1_nospec): Likewise.
12207 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12210 * config/sh/sh_optimize_sett_clrt.cc:
12211 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12213 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12215 PR tree-optimization/83940
12216 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12217 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12218 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12219 use it to pass back the definition type.
12220 (vect_check_store_rhs): Likewise rhs_dt_out.
12221 (vect_build_gather_load_calls): Add a mask_dt argument and use
12222 it instead of a call to vect_is_simple_use.
12223 (vectorizable_store): Update calls to vect_check_load_store_mask
12224 and vect_check_store_rhs. Use the dt returned by the latter instead
12225 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12226 instead of calls to vect_is_simple_use. Pass the scalar rather
12227 than the vector operand to vect_is_simple_use when handling
12228 second and subsequent copies of an rhs value.
12229 (vectorizable_load): Update calls to vect_check_load_store_mask
12230 and vect_build_gather_load_calls. Use the cached mask_dt and
12231 gs_info.offset_dt instead of calls to vect_is_simple_use.
12233 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12235 PR middle-end/83945
12236 * tree-emutls.c: Include gimplify.h.
12237 (lower_emutls_2): New function.
12238 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12239 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12240 it before further processing.
12243 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12244 UINTVAL (trueop1) instead of INTVAL (op1).
12246 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12250 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12251 INCOMING_FRAME_SP_OFFSET if not defined.
12252 (scan_trace): Add ENTRY argument. If true and
12253 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12254 emit a note to adjust the CFA offset.
12255 (create_cfi_notes): Adjust scan_trace callers.
12256 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12257 INCOMING_FRAME_SP_OFFSET in the CIE.
12258 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12259 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12261 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12262 * doc/tm.texi: Regenerated.
12264 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12266 PR rtl-optimization/83147
12267 * lra-constraints.c (remove_inheritance_pseudos): Use
12268 lra_substitute_pseudo_within_insn.
12270 2018-01-19 Tom de Vries <tom@codesourcery.com>
12271 Cesar Philippidis <cesar@codesourcery.com>
12274 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12276 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12279 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12280 spaces for function labels.
12282 2018-01-19 Martin Liska <mliska@suse.cz>
12284 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12285 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12286 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12287 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12288 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12289 (PRED_CONST_RETURN): Change from 69 to 65.
12290 (PRED_NULL_RETURN): Change from 91 to 71.
12291 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12292 (PRED_LOOP_GUARD): Change from 66 to 73.
12294 2018-01-19 Martin Liska <mliska@suse.cz>
12296 * predict.c (predict_insn_def): Add new assert.
12297 (struct branch_predictor): Change type to signed integer.
12298 (test_prediction_value_range): Amend test to cover
12299 PROB_UNINITIALIZED.
12300 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12301 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12302 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12303 (PRED_LOOP_IV_COMPARE): Likewise.
12304 * predict.h (PROB_UNINITIALIZED): Define new constant.
12306 2018-01-19 Martin Liska <mliska@suse.cz>
12308 * predict.c (dump_prediction): Add new format for
12309 analyze_brprob.py script which is enabled with -details
12311 * profile-count.h (precise_p): New function.
12313 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12315 PR tree-optimization/83922
12316 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12317 there are no statements that need masking.
12318 (vect_active_double_reduction_p): New function.
12319 (vect_analyze_loop_operations): Use it when handling phis that
12320 are not in the loop header.
12322 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12324 PR tree-optimization/83914
12325 * tree-vect-loop.c (vectorizable_induction): Don't convert
12326 init_expr or apply the peeling adjustment for inductions
12327 that are nested within the vectorized loop.
12329 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12331 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12334 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12338 * function.h (gimplify_parameters): Add gimple_seq * argument.
12339 * function.c: Include gimple.h and options.h.
12340 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12341 for the added local temporaries if needed.
12342 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12343 if there are any parameter cleanups, wrap whole body into a
12344 try/finally with the cleanups.
12346 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12349 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12350 Use GET_MODE_CLASS for scalar floating point.
12352 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12356 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12357 Fix call of call_cgraph_insertion_hooks.
12359 2018-01-18 Martin Sebor <msebor@redhat.com>
12361 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12363 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12366 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12369 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12372 * common.opt: (-ffile-prefix-map): New option.
12373 * opts.c (common_handle_option): Defer it.
12374 * opts-global.c (handle_common_deferred_options): Handle it.
12375 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12376 * file-prefix-map.h: New file.
12377 (remap_debug_filename, add_debug_prefix_map): ...here.
12378 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12379 * final.c (debug_prefix_map, add_debug_prefix_map
12380 remap_debug_filename): Move to...
12381 * file-prefix-map.c: New file.
12382 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12383 generalize, get rid of alloca(), use strrchr() instead of strchr().
12384 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12385 Implement in terms of add_prefix_map().
12386 (remap_macro_filename, remap_debug_filename): Implement in term of
12388 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12389 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12390 * dbxout.c: Include file-prefix-map.h.
12391 * varasm.c: Likewise.
12392 * vmsdbgout.c: Likewise.
12393 * xcoffout.c: Likewise.
12394 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12395 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12396 * doc/invoke.texi (-ffile-prefix-map): Document.
12397 (-fdebug-prefix-map): Update description.
12399 2018-01-18 Martin Liska <mliska@suse.cz>
12401 * config/i386/i386.c (indirect_thunk_name): Document that also
12403 (output_indirect_thunk): Document why both instructions
12404 (pause and lfence) are generated.
12406 2018-01-18 Richard Biener <rguenther@suse.de>
12408 PR tree-optimization/83887
12409 * graphite-scop-detection.c
12410 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12411 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12412 (scop_detection::merge_sese): Re-implement with a flood-fill
12413 algorithm that properly finds a SESE region if it exists.
12415 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12418 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12419 pointer_diff optimizations use view_convert instead of convert.
12421 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12423 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12424 Generate different code for -mno-speculate-indirect-jumps.
12425 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12426 (*call_indirect_aix<mode>): Disable for
12427 -mno-speculate-indirect-jumps.
12428 (*call_indirect_aix<mode>_nospec): New define_insn.
12429 (*call_value_indirect_aix<mode>): Disable for
12430 -mno-speculate-indirect-jumps.
12431 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12432 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12433 -mno-speculate-indirect-jumps.
12434 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12436 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12438 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12439 long double type, set the flags for noting the default long double
12440 type, even if we don't pass or return a long double type.
12442 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12445 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12448 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12450 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12451 support for merge[hl].
12452 (fold_mergehl_helper): New helper function.
12453 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12454 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12455 (altivec_vmrglw_direct): Add xxmrglw insn.
12457 2018-01-17 Andrew Waterman <andrew@sifive.com>
12459 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12460 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12462 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12465 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12466 call the lto_location_cache before reading the
12467 DECL_SOURCE_LOCATION of the types.
12469 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12470 Richard Sandiford <richard.sandiford@linaro.org>
12472 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12473 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12474 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12475 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12476 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12478 * config/aarch64/constraints.md (aarch64_movti_operand):
12480 * config/aarch64/predicates.md (Uti): Add new constraint.
12482 2018-01-17 Carl Love <cel@us.ibm.com>
12484 * config/rs6000/vsx.md (define_expand xl_len_r,
12485 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12486 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12488 (define_expand, define_insn): Move the shift left from the
12489 define_insn to the define_expand for lxvl and stxvl instructions.
12490 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12491 and XL_LEN_R definitions to PURE.
12493 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12495 * config/i386/i386.c (indirect_thunk_name): Declare regno
12496 as unsigned int. Compare regno with INVALID_REGNUM.
12497 (output_indirect_thunk): Ditto.
12498 (output_indirect_thunk_function): Ditto.
12499 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12500 in the call to output_indirect_thunk_function.
12502 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12504 PR middle-end/83884
12505 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12506 rather than the size of inner_type to determine the stack slot size
12507 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12509 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12512 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12515 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12517 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12518 endian Linux systems to optionally enable multilibs for selecting
12519 the long double type if the user configured an explicit type.
12520 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12521 have no long double multilibs if not defined.
12522 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12523 warn if the user used -mabi={ieee,ibm}longdouble and we built
12524 multilibs for long double.
12525 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12526 appropriate multilib option.
12527 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12529 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12530 for building long double multilibs.
12531 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12533 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12535 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12538 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12540 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12543 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12546 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12549 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12551 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12552 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12554 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12556 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12557 different rtl trees depending on TARGET_64BIT.
12558 (rs6000_gen_lvx): Likewise.
12560 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12562 * config/visium/visium.md (nop): Tweak comment.
12563 (hazard_nop): Likewise.
12565 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12567 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12568 -mspeculate-indirect-jumps.
12569 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12570 for -mno-speculate-indirect-jumps.
12571 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12572 (*call_value_indirect_elfv2<mode>): Disable for
12573 -mno-speculate-indirect-jumps.
12574 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12575 (indirect_jump): Emit different RTL for
12576 -mno-speculate-indirect-jumps.
12577 (*indirect_jump<mode>): Disable for
12578 -mno-speculate-indirect-jumps.
12579 (*indirect_jump<mode>_nospec): New define_insn.
12580 (tablejump): Emit different RTL for
12581 -mno-speculate-indirect-jumps.
12582 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12583 (tablejumpsi_nospec): New define_expand.
12584 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12585 (tablejumpdi_nospec): New define_expand.
12586 (*tablejump<mode>_internal1): Disable for
12587 -mno-speculate-indirect-jumps.
12588 (*tablejump<mode>_internal1_nospec): New define_insn.
12589 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12592 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12594 * caller-save.c (insert_save): Drop unnecessary parameter. All
12597 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12598 Richard Biener <rguenth@suse.de>
12601 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12602 return early, inline manually is_gimple_sizepos. Make sure if we
12603 call gimplify_expr we don't end up with a gimple constant.
12604 * tree.c (variably_modified_type_p): Don't return true for
12605 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12606 * gimplify.h (is_gimple_sizepos): Remove.
12608 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12610 PR tree-optimization/83857
12611 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12612 vectorizable_live_operation for pure SLP statements.
12613 (vectorizable_live_operation): Handle PHIs.
12615 2018-01-16 Richard Biener <rguenther@suse.de>
12617 PR tree-optimization/83867
12618 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12619 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12621 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12624 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12625 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12626 If off is not INTEGER_CST, issue a may not be aligned warning
12627 rather than isn't aligned. Use isn%'t rather than isn't.
12628 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12630 <case MULT_EXPR>: Improve the case when bottom and one of the
12631 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12632 operand, in that case check if the other operand is multiple of
12633 bottom divided by the INTEGER_CST operand.
12635 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12638 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12639 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12640 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12641 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12642 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12643 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12644 (pa_function_arg_size): New function.
12646 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12648 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12649 in a separate statement.
12651 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12653 PR tree-optimization/83847
12654 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12655 group gathers and scatters.
12657 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12659 PR rtl-optimization/86620
12660 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12662 PR rtl-optimization/83213
12663 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12664 to last if both are JUMP_INSNs.
12666 PR tree-optimization/83843
12667 * gimple-ssa-store-merging.c
12668 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12669 store_immediate_info for bswap/nop orig_stores.
12671 2018-01-15 Andrew Waterman <andrew@sifive.com>
12673 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12675 <UDIV>: Increase cost if !TARGET_DIV.
12677 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12679 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12680 (define_attr "cr_logical_3op"): New.
12681 (cceq_ior_compare): Adjust.
12682 (cceq_ior_compare_complement): Adjust.
12683 (*cceq_rev_compare): Adjust.
12684 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12685 (is_cracked_insn): Adjust.
12686 (insn_must_be_first_in_group): Adjust.
12687 * config/rs6000/40x.md: Adjust.
12688 * config/rs6000/440.md: Adjust.
12689 * config/rs6000/476.md: Adjust.
12690 * config/rs6000/601.md: Adjust.
12691 * config/rs6000/603.md: Adjust.
12692 * config/rs6000/6xx.md: Adjust.
12693 * config/rs6000/7450.md: Adjust.
12694 * config/rs6000/7xx.md: Adjust.
12695 * config/rs6000/8540.md: Adjust.
12696 * config/rs6000/cell.md: Adjust.
12697 * config/rs6000/e300c2c3.md: Adjust.
12698 * config/rs6000/e500mc.md: Adjust.
12699 * config/rs6000/e500mc64.md: Adjust.
12700 * config/rs6000/e5500.md: Adjust.
12701 * config/rs6000/e6500.md: Adjust.
12702 * config/rs6000/mpc.md: Adjust.
12703 * config/rs6000/power4.md: Adjust.
12704 * config/rs6000/power5.md: Adjust.
12705 * config/rs6000/power6.md: Adjust.
12706 * config/rs6000/power7.md: Adjust.
12707 * config/rs6000/power8.md: Adjust.
12708 * config/rs6000/power9.md: Adjust.
12709 * config/rs6000/rs64.md: Adjust.
12710 * config/rs6000/titan.md: Adjust.
12712 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12714 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12715 ix86_indirect_branch_register logic.
12717 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12719 * config/i386/constraints.md (Bs): Update
12720 ix86_indirect_branch_register check. Don't check
12721 ix86_indirect_branch_register with GOT_memory_operand.
12723 * config/i386/predicates.md (GOT_memory_operand): Don't check
12724 ix86_indirect_branch_register here.
12725 (GOT32_symbol_operand): Likewise.
12727 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12729 * config/i386/predicates.md (constant_call_address_operand):
12730 Rewrite ix86_indirect_branch_register logic.
12731 (sibcall_insn_operand): Likewise.
12733 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12735 * config/i386/constraints.md (Bs): Replace
12736 ix86_indirect_branch_thunk_register with
12737 ix86_indirect_branch_register.
12739 * config/i386/i386.md (indirect_jump): Likewise.
12740 (tablejump): Likewise.
12741 (*sibcall_memory): Likewise.
12742 (*sibcall_value_memory): Likewise.
12743 Peepholes of indirect call and jump via memory: Likewise.
12744 * config/i386/i386.opt: Likewise.
12745 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12746 (GOT_memory_operand): Likewise.
12747 (call_insn_operand): Likewise.
12748 (sibcall_insn_operand): Likewise.
12749 (GOT32_symbol_operand): Likewise.
12751 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12753 PR middle-end/83837
12754 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12755 type rather than type addr's type points to.
12756 (expand_omp_atomic_mutex): Likewise.
12757 (expand_omp_atomic): Likewise.
12759 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12762 * config/i386/i386.c (output_indirect_thunk_function): Use
12763 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12764 for __x86_return_thunk.
12766 2018-01-15 Richard Biener <rguenther@suse.de>
12768 PR middle-end/83850
12769 * expmed.c (extract_bit_field_1): Fix typo.
12771 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12774 * config/arm/iterators.md (VF): New mode iterator.
12775 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12776 Remove integer-related logic from pattern.
12777 (neon_vabd<mode>_3): Likewise.
12779 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12781 PR middle-end/82694
12782 * common.opt (fstrict-overflow): No longer an alias.
12783 (fwrapv-pointer): New option.
12784 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12785 also for pointer types based on flag_wrapv_pointer.
12786 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12787 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12788 opts->x_flag_wrapv got set.
12789 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12790 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12791 POINTER_TYPE_OVERFLOW_UNDEFINED.
12792 * match.pd: Likewise in address comparison pattern.
12793 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12795 2018-01-15 Richard Biener <rguenther@suse.de>
12798 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12799 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12800 Reset type names to their identifier if their TYPE_DECL doesn't
12801 have linkage (and thus is used for ODR and devirt).
12802 (save_debug_info_for_decl): Remove.
12803 (save_debug_info_for_type): Likewise.
12804 (add_tree_to_fld_list): Adjust.
12805 * tree-pretty-print.c (dump_generic_node): Make dumping of
12806 type names more robust.
12808 2018-01-15 Richard Biener <rguenther@suse.de>
12810 * BASE-VER: Bump to 8.0.1.
12812 2018-01-14 Martin Sebor <msebor@redhat.com>
12815 * builtins.c (check_access): Avoid warning when the no-warning bit
12818 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12820 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12821 * ira-color (allocno_hard_regs_compare): Likewise.
12823 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12826 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12827 Use .pushsection/.popsection.
12829 2018-01-14 Martin Sebor <msebor@redhat.com>
12832 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12834 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12836 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12837 entry from extra_headers.
12838 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12839 extra_headers, make the list bitwise identical to the i?86-*-* one.
12841 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12843 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12844 -mcmodel=large with -mindirect-branch=thunk,
12845 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12846 -mfunction-return=thunk-extern.
12847 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12848 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12849 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12851 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12853 * config/i386/i386.c (print_reg): Print the name of the full
12854 integer register without '%'.
12855 (ix86_print_operand): Handle 'V'.
12856 * doc/extend.texi: Document 'V' modifier.
12858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12860 * config/i386/constraints.md (Bs): Disallow memory operand for
12861 -mindirect-branch-register.
12863 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12864 (GOT_memory_operand): Likewise.
12865 (call_insn_operand): Likewise.
12866 (sibcall_insn_operand): Likewise.
12867 (GOT32_symbol_operand): Likewise.
12868 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12869 for -mindirect-branch-register.
12870 (tablejump): Likewise.
12871 (*sibcall_memory): Likewise.
12872 (*sibcall_value_memory): Likewise.
12873 Disallow peepholes of indirect call and jump via memory for
12874 -mindirect-branch-register.
12875 (*call_pop): Replace m with Bw.
12876 (*call_value_pop): Likewise.
12877 (*sibcall_pop_memory): Replace m with Bs.
12878 * config/i386/i386.opt (mindirect-branch-register): New option.
12879 * doc/invoke.texi: Document -mindirect-branch-register option.
12881 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12883 * config/i386/i386-protos.h (ix86_output_function_return): New.
12884 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12885 set function_return_type.
12886 (indirect_thunk_name): Add ret_p to indicate thunk for function
12888 (output_indirect_thunk_function): Pass false to
12889 indirect_thunk_name.
12890 (ix86_output_indirect_branch_via_reg): Likewise.
12891 (ix86_output_indirect_branch_via_push): Likewise.
12892 (output_indirect_thunk_function): Create alias for function
12893 return thunk if regno < 0.
12894 (ix86_output_function_return): New function.
12895 (ix86_handle_fndecl_attribute): Handle function_return.
12896 (ix86_attribute_table): Add function_return.
12897 * config/i386/i386.h (machine_function): Add
12898 function_return_type.
12899 * config/i386/i386.md (simple_return_internal): Use
12900 ix86_output_function_return.
12901 (simple_return_internal_long): Likewise.
12902 * config/i386/i386.opt (mfunction-return=): New option.
12903 (indirect_branch): Mention -mfunction-return=.
12904 * doc/extend.texi: Document function_return function attribute.
12905 * doc/invoke.texi: Document -mfunction-return= option.
12907 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12909 * config/i386/i386-opts.h (indirect_branch): New.
12910 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12911 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12912 with local indirect jump when converting indirect call and jump.
12913 (ix86_set_indirect_branch_type): New.
12914 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12915 (indirectlabelno): New.
12916 (indirect_thunk_needed): Likewise.
12917 (indirect_thunk_bnd_needed): Likewise.
12918 (indirect_thunks_used): Likewise.
12919 (indirect_thunks_bnd_used): Likewise.
12920 (INDIRECT_LABEL): Likewise.
12921 (indirect_thunk_name): Likewise.
12922 (output_indirect_thunk): Likewise.
12923 (output_indirect_thunk_function): Likewise.
12924 (ix86_output_indirect_branch_via_reg): Likewise.
12925 (ix86_output_indirect_branch_via_push): Likewise.
12926 (ix86_output_indirect_branch): Likewise.
12927 (ix86_output_indirect_jmp): Likewise.
12928 (ix86_code_end): Call output_indirect_thunk_function if needed.
12929 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12931 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12932 (ix86_attribute_table): Add indirect_branch.
12933 * config/i386/i386.h (machine_function): Add indirect_branch_type
12934 and has_local_indirect_jump.
12935 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12937 (tablejump): Likewise.
12938 (*indirect_jump): Use ix86_output_indirect_jmp.
12939 (*tablejump_1): Likewise.
12940 (simple_return_indirect_internal): Likewise.
12941 * config/i386/i386.opt (mindirect-branch=): New option.
12942 (indirect_branch): New.
12945 (thunk-inline): Likewise.
12946 (thunk-extern): Likewise.
12947 * doc/extend.texi: Document indirect_branch function attribute.
12948 * doc/invoke.texi: Document -mindirect-branch= option.
12950 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12953 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12955 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12957 * ipa-inline.c (want_inline_small_function_p): Return false if
12958 inlining has already failed with CIF_FINAL_ERROR.
12959 (update_caller_keys): Call want_inline_small_function_p before
12961 (update_callee_keys): Likewise.
12963 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12965 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12967 (rs6000_quadword_masked_address_p): Likewise.
12968 (quad_aligned_load_p): Likewise.
12969 (quad_aligned_store_p): Likewise.
12970 (const_load_sequence_p): Add comment to describe the outer-most loop.
12971 (mimic_memory_attributes_and_flags): New function.
12972 (rs6000_gen_stvx): Likewise.
12973 (replace_swapped_aligned_store): Likewise.
12974 (rs6000_gen_lvx): Likewise.
12975 (replace_swapped_aligned_load): Likewise.
12976 (replace_swapped_load_constant): Capitalize argument name in
12977 comment describing this function.
12978 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12979 and stores that access quad-word aligned addresses and replace
12980 with stvx or lvx instructions when appropriate.
12981 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12982 New function prototype.
12983 (rs6000_quadword_masked_address_p): Likewise.
12984 (rs6000_gen_lvx): Likewise.
12985 (rs6000_gen_stvx): Likewise.
12986 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12987 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12988 when memory address is aligned.
12989 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12990 this split to select lvx instruction when memory address is aligned.
12991 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12992 instruction when memory address is aligned.
12993 (*vsx_le_perm_load_v16qi): Likewise.
12994 (four unnamed splitters): Modify to select the stvx instruction
12995 when memory is aligned.
12997 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12999 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13000 which appears in the queue multiple times.
13002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13003 Alan Hayward <alan.hayward@arm.com>
13004 David Sherwood <david.sherwood@arm.com>
13006 * tree-vectorizer.h (vec_lower_bound): New structure.
13007 (_loop_vec_info): Add check_nonzero and lower_bounds.
13008 (LOOP_VINFO_CHECK_NONZERO): New macro.
13009 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13010 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13011 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13012 fields. Make seg_len the distance travelled, not including the
13014 (dr_direction_indicator): Declare.
13015 (dr_zero_step_indicator): Likewise.
13016 (dr_known_forward_stride_p): Likewise.
13017 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13019 (runtime_alias_check_p): Allow runtime alias checks with
13021 (operator ==): Compare access_size and align.
13022 (prune_runtime_alias_test_list): Rework for new distinction between
13023 the access_size and seg_len.
13024 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13026 (get_segment_min_max): New function.
13027 (create_intersect_range_checks): Use it.
13028 (dr_step_indicator): New function.
13029 (dr_direction_indicator): Likewise.
13030 (dr_zero_step_indicator): Likewise.
13031 (dr_known_forward_stride_p): Likewise.
13032 * tree-loop-distribution.c (data_ref_segment_size): Return
13033 DR_STEP * (niters - 1).
13034 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13036 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13037 (vect_preserves_scalar_order_p): New function, split out from...
13038 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13039 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13040 (vect_vfa_access_size): New function.
13041 (vect_vfa_align): Likewise.
13042 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13043 (dump_lower_bound): New function.
13044 (vect_check_lower_bound): Likewise.
13045 (vect_small_gap_p): Likewise.
13046 (vectorizable_with_step_bound_p): Likewise.
13047 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13048 depencies if the vectorization factor is 1. Convert the checks
13049 for nonzero steps into checks on the bounds of DR_STEP. Try using
13050 a bunds check for variable steps if the minimum required step is
13051 relatively small. Update calls to the dr_with_seg_len
13052 constructor and to vect_compile_time_alias.
13053 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13055 (vect_loop_versioning): Call it.
13056 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13058 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13061 Alan Hayward <alan.hayward@arm.com>
13062 David Sherwood <david.sherwood@arm.com>
13064 * doc/sourcebuild.texi (vect_scatter_store): Document.
13065 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13067 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13069 * genopinit.c (main): Add supports_vec_scatter_store and
13070 supports_vec_scatter_store_cached to target_optabs.
13071 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13072 IFN_MASK_SCATTER_STORE.
13073 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13075 * internal-fn.h (internal_store_fn_p): Declare.
13076 (internal_fn_stored_value_index): Likewise.
13077 * internal-fn.c (scatter_store_direct): New macro.
13078 (expand_scatter_store_optab_fn): New function.
13079 (direct_scatter_store_optab_supported_p): New macro.
13080 (internal_store_fn_p): New function.
13081 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13082 IFN_MASK_SCATTER_STORE.
13083 (internal_fn_mask_index): Likewise.
13084 (internal_fn_stored_value_index): New function.
13085 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13086 for scatter stores.
13087 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13088 * optabs-query.c (supports_vec_scatter_store_p): New function.
13089 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13090 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13091 true for scatter stores.
13092 (vect_gather_scatter_fn_p): Handle scatter stores too.
13093 (vect_check_gather_scatter): Consider using scatter stores if
13094 supports_vec_scatter_store_p.
13095 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13096 scatter stores too.
13097 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13098 internal_fn_stored_value_index.
13099 (check_load_store_masking): Handle scatter stores too.
13100 (vect_get_store_rhs): Make public.
13101 (vectorizable_call): Use internal_store_fn_p.
13102 (vectorizable_store): Handle scatter store internal functions.
13103 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13104 when deciding whether the end of the group has been reached.
13105 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13106 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13107 (mask_scatter_store<mode>): New insns.
13109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13110 Alan Hayward <alan.hayward@arm.com>
13111 David Sherwood <david.sherwood@arm.com>
13113 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13114 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13115 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13117 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13118 Use vect_truncate_gather_scatter_offset if we can't treat the
13119 operation as a normal gather load or scatter store.
13120 (get_group_load_store_type): Take the gather_scatter_info
13121 as argument. Try using a gather load or scatter store for
13122 single-element groups.
13123 (get_load_store_type): Update calls to get_group_load_store_type
13124 and vect_use_strided_gather_scatters_p.
13126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13127 Alan Hayward <alan.hayward@arm.com>
13128 David Sherwood <david.sherwood@arm.com>
13130 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13131 optional tree argument.
13132 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13134 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13135 but continue to use the current value as a fallback.
13136 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13137 to compare the updates.
13138 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13139 (get_load_store_type): Use it when handling a strided access.
13140 (vect_get_strided_load_store_ops): New function.
13141 (vect_get_data_ptr_increment): Likewise.
13142 (vectorizable_load): Handle strided gather loads. Always pass
13143 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13146 Alan Hayward <alan.hayward@arm.com>
13147 David Sherwood <david.sherwood@arm.com>
13149 * doc/md.texi (gather_load@var{m}): Document.
13150 (mask_gather_load@var{m}): Likewise.
13151 * genopinit.c (main): Add supports_vec_gather_load and
13152 supports_vec_gather_load_cached to target_optabs.
13153 * optabs-tree.c (init_tree_optimization_optabs): Use
13154 ggc_cleared_alloc to allocate target_optabs.
13155 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13156 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13158 * internal-fn.h (internal_load_fn_p): Declare.
13159 (internal_gather_scatter_fn_p): Likewise.
13160 (internal_fn_mask_index): Likewise.
13161 (internal_gather_scatter_fn_supported_p): Likewise.
13162 * internal-fn.c (gather_load_direct): New macro.
13163 (expand_gather_load_optab_fn): New function.
13164 (direct_gather_load_optab_supported_p): New macro.
13165 (direct_internal_fn_optab): New function.
13166 (internal_load_fn_p): Likewise.
13167 (internal_gather_scatter_fn_p): Likewise.
13168 (internal_fn_mask_index): Likewise.
13169 (internal_gather_scatter_fn_supported_p): Likewise.
13170 * optabs-query.c (supports_at_least_one_mode_p): New function.
13171 (supports_vec_gather_load_p): Likewise.
13172 * optabs-query.h (supports_vec_gather_load_p): Declare.
13173 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13174 and memory_type field.
13175 (NUM_PATTERNS): Bump to 15.
13176 * tree-vect-data-refs.c: Include internal-fn.h.
13177 (vect_gather_scatter_fn_p): New function.
13178 (vect_describe_gather_scatter_call): Likewise.
13179 (vect_check_gather_scatter): Try using internal functions for
13180 gather loads. Recognize existing calls to a gather load function.
13181 (vect_analyze_data_refs): Consider using gather loads if
13182 supports_vec_gather_load_p.
13183 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13184 (vect_get_gather_scatter_offset_type): Likewise.
13185 (vect_convert_mask_for_vectype): Likewise.
13186 (vect_add_conversion_to_patterm): Likewise.
13187 (vect_try_gather_scatter_pattern): Likewise.
13188 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13189 (vect_vect_recog_func_ptrs): Add it.
13190 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13191 internal_fn_mask_index and internal_gather_scatter_fn_p.
13192 (check_load_store_masking): Take the gather_scatter_info as an
13193 argument and handle gather loads.
13194 (vect_get_gather_scatter_ops): New function.
13195 (vectorizable_call): Check internal_load_fn_p.
13196 (vectorizable_load): Likewise. Handle gather load internal
13198 (vectorizable_store): Update call to check_load_store_masking.
13199 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13200 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13201 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13202 (aarch64_gather_scale_operand_d): New predicates.
13203 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13204 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
13211 * doc/md.texi (fold_left_plus_@var{m}): Document.
13212 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13213 * internal-fn.c (fold_left_direct): Define.
13214 (expand_fold_left_optab_fn): Likewise.
13215 (direct_fold_left_optab_supported_p): Likewise.
13216 * fold-const-call.c (fold_const_fold_left): New function.
13217 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13218 * tree-parloops.c (valid_reduction_p): New function.
13219 (gather_scalar_reductions): Use it.
13220 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13221 (vect_finish_replace_stmt): Declare.
13222 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13223 (needs_fold_left_reduction_p): New function, split out from...
13224 (vect_is_simple_reduction): ...here. Accept reductions that
13225 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13226 (vect_force_simple_reduction): Also store the reduction type in
13227 the assignment's STMT_VINFO_REDUC_TYPE.
13228 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13229 (merge_with_identity): New function.
13230 (vect_expand_fold_left): Likewise.
13231 (vectorize_fold_left_reduction): Likewise.
13232 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13233 scalar phi in place for it. Check for target support and reject
13234 cases that would reassociate the operation. Defer the transform
13235 phase to vectorize_fold_left_reduction.
13236 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13237 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13238 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13242 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13243 call to ifc_temp_var.
13245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13246 Alan Hayward <alan.hayward@arm.com>
13247 David Sherwood <david.sherwood@arm.com>
13249 * target.def (legitimize_address_displacement): Take the original
13250 offset as a poly_int.
13251 * targhooks.h (default_legitimize_address_displacement): Update
13253 * targhooks.c (default_legitimize_address_displacement): Likewise.
13254 * doc/tm.texi: Regenerate.
13255 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13256 as an argument, moving assert of ad->disp == ad->disp_term to...
13257 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13258 Try calling targetm.legitimize_address_displacement before expanding
13259 the address rather than afterwards, and adjust for the new interface.
13260 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13261 Match the new hook interface. Handle SVE addresses.
13262 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13263 new hook interface.
13265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13267 * Makefile.in (OBJS): Add early-remat.o.
13268 * target.def (select_early_remat_modes): New hook.
13269 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13270 * doc/tm.texi: Regenerate.
13271 * targhooks.h (default_select_early_remat_modes): Declare.
13272 * targhooks.c (default_select_early_remat_modes): New function.
13273 * timevar.def (TV_EARLY_REMAT): New timevar.
13274 * passes.def (pass_early_remat): New pass.
13275 * tree-pass.h (make_pass_early_remat): Declare.
13276 * early-remat.c: New file.
13277 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13279 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13282 Alan Hayward <alan.hayward@arm.com>
13283 David Sherwood <david.sherwood@arm.com>
13285 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13286 vfm1 with a bound_epilog parameter.
13287 (vect_do_peeling): Update calls accordingly, and move the prologue
13288 call earlier in the function. Treat the base bound_epilog as 0 for
13289 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13290 this base when peeling for gaps.
13291 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13292 with fully-masked loops.
13293 (vect_estimate_min_profitable_iters): Handle the single peeled
13294 iteration in that case.
13296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13297 Alan Hayward <alan.hayward@arm.com>
13298 David Sherwood <david.sherwood@arm.com>
13300 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13301 single-element interleaving even if the size is not a power of 2.
13302 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13303 accesses for single-element interleaving if the group size is
13306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13307 Alan Hayward <alan.hayward@arm.com>
13308 David Sherwood <david.sherwood@arm.com>
13310 * doc/md.texi (fold_extract_last_@var{m}): Document.
13311 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13312 * optabs.def (fold_extract_last_optab): New optab.
13313 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13314 * internal-fn.c (fold_extract_direct): New macro.
13315 (expand_fold_extract_optab_fn): Likewise.
13316 (direct_fold_extract_optab_supported_p): Likewise.
13317 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13318 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13319 EXTRACT_LAST_REDUCTION.
13320 (get_initial_def_for_reduction): Do not create an initial vector
13321 for EXTRACT_LAST_REDUCTION reductions.
13322 (vectorizable_reduction): Leave the scalar phi in place for
13323 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13324 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13325 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13326 transform phase to vectorizable_condition.
13327 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13329 (vect_finish_stmt_generation): ...here.
13330 (vect_finish_replace_stmt): New function.
13331 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13332 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13334 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13337 Alan Hayward <alan.hayward@arm.com>
13338 David Sherwood <david.sherwood@arm.com>
13340 * doc/md.texi (extract_last_@var{m}): Document.
13341 * optabs.def (extract_last_optab): New optab.
13342 * internal-fn.def (EXTRACT_LAST): New internal function.
13343 * internal-fn.c (cond_unary_direct): New macro.
13344 (expand_cond_unary_optab_fn): Likewise.
13345 (direct_cond_unary_optab_supported_p): Likewise.
13346 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13347 loops using EXTRACT_LAST.
13348 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13349 (extract_last_<mode>): ...this optab.
13350 (vec_extract<mode><Vel>): Update accordingly.
13352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13353 Alan Hayward <alan.hayward@arm.com>
13354 David Sherwood <david.sherwood@arm.com>
13356 * target.def (empty_mask_is_expensive): New hook.
13357 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13358 * doc/tm.texi: Regenerate.
13359 * targhooks.h (default_empty_mask_is_expensive): Declare.
13360 * targhooks.c (default_empty_mask_is_expensive): New function.
13361 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13362 if the target says that empty masks are expensive.
13363 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13365 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13368 Alan Hayward <alan.hayward@arm.com>
13369 David Sherwood <david.sherwood@arm.com>
13371 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13372 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13373 (vect_use_loop_mask_for_alignment_p): New function.
13374 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13375 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13376 niters_skip argument. Make sure that the first niters_skip elements
13377 of the first iteration are inactive.
13378 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13379 Update call to vect_set_loop_masks_directly.
13380 (get_misalign_in_elems): New function, split out from...
13381 (vect_gen_prolog_loop_niters): ...here.
13382 (vect_update_init_of_dr): Take a code argument that specifies whether
13383 the adjustment should be added or subtracted.
13384 (vect_update_init_of_drs): Likewise.
13385 (vect_prepare_for_masked_peels): New function.
13386 (vect_do_peeling): Skip prologue peeling if we're using a mask
13387 instead. Update call to vect_update_inits_of_drs.
13388 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13390 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13391 alignment. Do not include the number of peeled iterations in
13392 the minimum threshold in that case.
13393 (vectorizable_induction): Adjust the start value down by
13394 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13395 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13396 Take the number of skipped iterations into account when calculating
13398 * tree-vect-stmts.c (vect_gen_while_not): New function.
13400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13401 Alan Hayward <alan.hayward@arm.com>
13402 David Sherwood <david.sherwood@arm.com>
13404 * doc/sourcebuild.texi (vect_fully_masked): Document.
13405 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13406 default value to 0.
13407 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13409 (vect_analyze_loop_2): ...here. Don't check the vectorization
13410 factor against the number of loop iterations if the loop is
13413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13414 Alan Hayward <alan.hayward@arm.com>
13415 David Sherwood <david.sherwood@arm.com>
13417 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13418 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13419 (dump_groups): Update accordingly.
13420 (iv_use::mem_type): New member variable.
13421 (address_p): New function.
13422 (record_use): Add a mem_type argument and initialize the new
13424 (record_group_use): Add a mem_type argument. Use address_p.
13425 Remove obsolete null checks of base_object. Update call to record_use.
13426 (find_interesting_uses_op): Update call to record_group_use.
13427 (find_interesting_uses_cond): Likewise.
13428 (find_interesting_uses_address): Likewise.
13429 (get_mem_type_for_internal_fn): New function.
13430 (find_address_like_use): Likewise.
13431 (find_interesting_uses_stmt): Try find_address_like_use before
13432 calling find_interesting_uses_op.
13433 (addr_offset_valid_p): Use the iv mem_type field as the type
13434 of the addressed memory.
13435 (add_autoinc_candidates): Likewise.
13436 (get_address_cost): Likewise.
13437 (split_small_address_groups_p): Use address_p.
13438 (split_address_groups): Likewise.
13439 (add_iv_candidate_for_use): Likewise.
13440 (autoinc_possible_for_pair): Likewise.
13441 (rewrite_groups): Likewise.
13442 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13443 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13444 (get_alias_ptr_type_for_ptr_address): New function.
13445 (rewrite_use_address): Rewrite address uses in calls that were
13446 identified by find_address_like_use.
13448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13449 Alan Hayward <alan.hayward@arm.com>
13450 David Sherwood <david.sherwood@arm.com>
13452 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13454 * gimple-expr.h (is_gimple_addressable: Likewise.
13455 * gimple-expr.c (is_gimple_address): Likewise.
13456 * internal-fn.c (expand_call_mem_ref): New function.
13457 (expand_mask_load_optab_fn): Use it.
13458 (expand_mask_store_optab_fn): Likewise.
13460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13461 Alan Hayward <alan.hayward@arm.com>
13462 David Sherwood <david.sherwood@arm.com>
13464 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13465 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13466 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13467 (cond_umax@var{mode}): Document.
13468 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13469 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13470 (cond_umin_optab, cond_umax_optab): New optabs.
13471 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13472 (COND_IOR, COND_XOR): New internal functions.
13473 * internal-fn.h (get_conditional_internal_fn): Declare.
13474 * internal-fn.c (cond_binary_direct): New macro.
13475 (expand_cond_binary_optab_fn): Likewise.
13476 (direct_cond_binary_optab_supported_p): Likewise.
13477 (get_conditional_internal_fn): New function.
13478 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13479 Cope with reduction statements that are vectorized as calls rather
13481 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13482 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13483 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13484 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13485 (UNSPEC_COND_EOR): New unspecs.
13486 (optab): Add mappings for them.
13487 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13488 (sve_int_op, sve_fp_op): New int attributes.
13490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13491 Alan Hayward <alan.hayward@arm.com>
13492 David Sherwood <david.sherwood@arm.com>
13494 * optabs.def (while_ult_optab): New optab.
13495 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13496 * internal-fn.def (WHILE_ULT): New internal function.
13497 * internal-fn.h (direct_internal_fn_supported_p): New override
13498 that takes two types as argument.
13499 * internal-fn.c (while_direct): New macro.
13500 (expand_while_optab_fn): New function.
13501 (convert_optab_supported_p): Likewise.
13502 (direct_while_optab_supported_p): New macro.
13503 * wide-int.h (wi::udiv_ceil): New function.
13504 * tree-vectorizer.h (rgroup_masks): New structure.
13505 (vec_loop_masks): New typedef.
13506 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13507 and fully_masked_p.
13508 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13509 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13510 (vect_max_vf): New function.
13511 (slpeel_make_loop_iterate_ntimes): Delete.
13512 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13513 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13514 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13515 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13516 internal-fn.h, stor-layout.h and optabs-query.h.
13517 (vect_set_loop_mask): New function.
13518 (add_preheader_seq): Likewise.
13519 (add_header_seq): Likewise.
13520 (interleave_supported_p): Likewise.
13521 (vect_maybe_permute_loop_masks): Likewise.
13522 (vect_set_loop_masks_directly): Likewise.
13523 (vect_set_loop_condition_masked): Likewise.
13524 (vect_set_loop_condition_unmasked): New function, split out from
13525 slpeel_make_loop_iterate_ntimes.
13526 (slpeel_make_loop_iterate_ntimes): Rename to..
13527 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13528 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13529 (vect_do_peeling): Update call accordingly.
13530 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13532 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13533 mask_compare_type, can_fully_mask_p and fully_masked_p.
13534 (release_vec_loop_masks): New function.
13535 (_loop_vec_info): Use it to free the loop masks.
13536 (can_produce_all_loop_masks_p): New function.
13537 (vect_get_max_nscalars_per_iter): Likewise.
13538 (vect_verify_full_masking): Likewise.
13539 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13540 retries, and free the mask rgroups before retrying. Check loop-wide
13541 reasons for disallowing fully-masked loops. Make the final decision
13542 about whether use a fully-masked loop or not.
13543 (vect_estimate_min_profitable_iters): Do not assume that peeling
13544 for the number of iterations will be needed for fully-masked loops.
13545 (vectorizable_reduction): Disable fully-masked loops.
13546 (vectorizable_live_operation): Likewise.
13547 (vect_halve_mask_nunits): New function.
13548 (vect_double_mask_nunits): Likewise.
13549 (vect_record_loop_mask): Likewise.
13550 (vect_get_loop_mask): Likewise.
13551 (vect_transform_loop): Handle the case in which the final loop
13552 iteration might handle a partial vector. Call vect_set_loop_condition
13553 instead of slpeel_make_loop_iterate_ntimes.
13554 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13555 (check_load_store_masking): New function.
13556 (prepare_load_store_mask): Likewise.
13557 (vectorizable_store): Handle fully-masked loops.
13558 (vectorizable_load): Likewise.
13559 (supportable_widening_operation): Use vect_halve_mask_nunits for
13561 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13562 (vect_gen_while): New function.
13563 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13564 (aarch64_uqdec<mode>): New insn.
13566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13567 Alan Hayward <alan.hayward@arm.com>
13568 David Sherwood <david.sherwood@arm.com>
13570 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13571 (reduc_xor_scal_optab): New optabs.
13572 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13573 (reduc_xor_scal_@var{m}): Document.
13574 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13575 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13576 internal functions.
13577 * fold-const-call.c (fold_const_call): Handle them.
13578 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13579 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13580 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13581 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13582 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13583 (UNSPEC_XORV): New unspecs.
13584 (optab): Add entries for them.
13585 (BITWISEV): New int iterator.
13586 (bit_reduc_op): New int attributes.
13588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13589 Alan Hayward <alan.hayward@arm.com>
13590 David Sherwood <david.sherwood@arm.com>
13592 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13593 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13594 * optabs.def (vec_shl_insert_optab): New optab.
13595 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13596 (duplicate_and_interleave): Likewise.
13597 * tree-vect-loop.c: Include internal-fn.h.
13598 (neutral_op_for_slp_reduction): New function, split out from
13599 get_initial_defs_for_reduction.
13600 (get_initial_def_for_reduction): Handle option 2 for variable-length
13601 vectors by loading the neutral value into a vector and then shifting
13602 the initial value into element 0.
13603 (get_initial_defs_for_reduction): Replace the code argument with
13604 the neutral value calculated by neutral_op_for_slp_reduction.
13605 Use gimple_build_vector for constant-length vectors.
13606 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13607 but the first group_size elements have a neutral value.
13608 Use duplicate_and_interleave otherwise.
13609 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13610 Update call to get_initial_defs_for_reduction. Handle SLP
13611 reductions for variable-length vectors by creating one vector
13612 result for each scalar result, with the elements associated
13613 with other scalar results stubbed out with the neutral value.
13614 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13615 Require IFN_VEC_SHL_INSERT for double reductions on
13616 variable-length vectors, or SLP reductions that have
13617 a neutral value. Require can_duplicate_and_interleave_p
13618 support for variable-length unchained SLP reductions if there
13619 is no neutral value, such as for MIN/MAX reductions. Also require
13620 the number of vector elements to be a multiple of the number of
13621 SLP statements when doing variable-length unchained SLP reductions.
13622 Update call to vect_create_epilog_for_reduction.
13623 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13624 and remove initial values.
13625 (duplicate_and_interleave): Make public.
13626 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13627 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13630 Alan Hayward <alan.hayward@arm.com>
13631 David Sherwood <david.sherwood@arm.com>
13633 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13634 (can_duplicate_and_interleave_p): New function.
13635 (vect_get_and_check_slp_defs): Take the vector of statements
13636 rather than just the current one. Remove excess parentheses.
13637 Restriction rejectinon of vect_constant_def and vect_external_def
13638 for variable-length vectors to boolean types, or types for which
13639 can_duplicate_and_interleave_p is false.
13640 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13641 (duplicate_and_interleave): New function.
13642 (vect_get_constant_vectors): Use gimple_build_vector for
13643 constant-length vectors and suitable variable-length constant
13644 vectors. Use duplicate_and_interleave for other variable-length
13645 vectors. Don't defer the update when inserting new statements.
13647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13648 Alan Hayward <alan.hayward@arm.com>
13649 David Sherwood <david.sherwood@arm.com>
13651 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13652 min_profitable_iters doesn't go negative.
13654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13655 Alan Hayward <alan.hayward@arm.com>
13656 David Sherwood <david.sherwood@arm.com>
13658 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13659 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13660 * optabs.def (vec_mask_load_lanes_optab): New optab.
13661 (vec_mask_store_lanes_optab): Likewise.
13662 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13663 (MASK_STORE_LANES): Likewise.
13664 * internal-fn.c (mask_load_lanes_direct): New macro.
13665 (mask_store_lanes_direct): Likewise.
13666 (expand_mask_load_optab_fn): Handle masked operations.
13667 (expand_mask_load_lanes_optab_fn): New macro.
13668 (expand_mask_store_optab_fn): Handle masked operations.
13669 (expand_mask_store_lanes_optab_fn): New macro.
13670 (direct_mask_load_lanes_optab_supported_p): Likewise.
13671 (direct_mask_store_lanes_optab_supported_p): Likewise.
13672 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13674 (vect_load_lanes_supported): Likewise.
13675 * tree-vect-data-refs.c (strip_conversion): New function.
13676 (can_group_stmts_p): Likewise.
13677 (vect_analyze_data_ref_accesses): Use it instead of checking
13678 for a pair of assignments.
13679 (vect_store_lanes_supported): Take a masked_p parameter.
13680 (vect_load_lanes_supported): Likewise.
13681 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13682 vect_store_lanes_supported and vect_load_lanes_supported.
13683 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13684 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13685 parameter. Don't allow gaps for masked accesses.
13686 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13687 and vect_load_lanes_supported.
13688 (get_load_store_type): Take a masked_p parameter and update
13689 call to get_group_load_store_type.
13690 (vectorizable_store): Update call to get_load_store_type.
13691 Handle IFN_MASK_STORE_LANES.
13692 (vectorizable_load): Update call to get_load_store_type.
13693 Handle IFN_MASK_LOAD_LANES.
13695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13696 Alan Hayward <alan.hayward@arm.com>
13697 David Sherwood <david.sherwood@arm.com>
13699 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13701 * config/aarch64/aarch64-protos.h
13702 (aarch64_sve_struct_memory_operand_p): Declare.
13703 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13704 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13705 (VPRED, vpred): Handle SVE structure modes.
13706 * config/aarch64/constraints.md (Utx): New constraint.
13707 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13708 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13709 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13710 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13711 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13712 structure modes. Split into pieces after RA.
13713 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13714 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13716 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13717 SVE structure modes.
13718 (aarch64_classify_address): Likewise.
13719 (sizetochar): Move earlier in file.
13720 (aarch64_print_operand): Handle SVE register lists.
13721 (aarch64_array_mode): New function.
13722 (aarch64_sve_struct_memory_operand_p): Likewise.
13723 (TARGET_ARRAY_MODE): Redefine.
13725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13726 Alan Hayward <alan.hayward@arm.com>
13727 David Sherwood <david.sherwood@arm.com>
13729 * target.def (array_mode): New target hook.
13730 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13731 * doc/tm.texi: Regenerate.
13732 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13733 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13734 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13735 targetm.array_mode.
13736 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13740 Alan Hayward <alan.hayward@arm.com>
13741 David Sherwood <david.sherwood@arm.com>
13743 * fold-const.c (fold_binary_loc): Check the argument types
13744 rather than the result type when testing for a vector operation.
13746 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13748 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13749 * doc/tm.texi: Regenerate.
13751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13752 Alan Hayward <alan.hayward@arm.com>
13753 David Sherwood <david.sherwood@arm.com>
13755 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13756 (sve): Document new AArch64 extension.
13757 * doc/md.texi (w): Extend the description of the AArch64
13758 constraint to include SVE vectors.
13759 (Upl, Upa): Document new AArch64 predicate constraints.
13760 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13762 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13763 (msve-vector-bits=): New option.
13764 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13765 SVE when these are disabled.
13766 (sve): New extension.
13767 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13768 modes. Adjust their number of units based on aarch64_sve_vg.
13769 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13770 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13771 aarch64_addr_query_type.
13772 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13773 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13774 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13775 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13776 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13777 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13778 (aarch64_simd_imm_zero_p): Delete.
13779 (aarch64_check_zero_based_sve_index_immediate): Declare.
13780 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13781 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13782 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13783 (aarch64_sve_float_mul_immediate_p): Likewise.
13784 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13785 rather than an rtx.
13786 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13787 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13788 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13789 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13790 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13791 (aarch64_regmode_natural_size): Likewise.
13792 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13793 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13795 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13796 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13797 for VG and the SVE predicate registers.
13798 (V_ALIASES): Add a "z"-prefixed alias.
13799 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13800 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13801 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13802 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13803 (REG_CLASS_NAMES): Add entries for them.
13804 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13805 and the predicate registers.
13806 (aarch64_sve_vg): Declare.
13807 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13808 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13809 (REGMODE_NATURAL_SIZE): Define.
13810 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13812 * config/aarch64/aarch64.c: Include cfgrtl.h.
13813 (simd_immediate_info): Add a constructor for series vectors,
13814 and an associated step field.
13815 (aarch64_sve_vg): New variable.
13816 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13817 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13818 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13819 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13820 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13821 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13822 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13823 (aarch64_get_mask_mode): New functions.
13824 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13825 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13826 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13827 predicate modes and predicate registers. Explicitly restrict
13828 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13829 to store a vector mode if it is recognized by
13830 aarch64_classify_vector_mode.
13831 (aarch64_regmode_natural_size): New function.
13832 (aarch64_hard_regno_caller_save_mode): Return the original mode
13834 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13835 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13836 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13837 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13839 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13840 does not overlap dest if the function is frame-related. Handle
13842 (aarch64_split_add_offset): New function.
13843 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13844 them aarch64_add_offset.
13845 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13846 and update call to aarch64_sub_sp.
13847 (aarch64_add_cfa_expression): New function.
13848 (aarch64_expand_prologue): Pass extra temporary registers to the
13849 functions above. Handle the case in which we need to emit new
13850 DW_CFA_expressions for registers that were originally saved
13851 relative to the stack pointer, but now have to be expressed
13852 relative to the frame pointer.
13853 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13855 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13856 IP0 and IP1 values for SVE frames.
13857 (aarch64_expand_vec_series): New function.
13858 (aarch64_expand_sve_widened_duplicate): Likewise.
13859 (aarch64_expand_sve_const_vector): Likewise.
13860 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13861 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13862 into the register, rather than emitting a SET directly.
13863 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13864 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13865 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13866 (offset_9bit_signed_scaled_p): New functions.
13867 (aarch64_replicate_bitmask_imm): New function.
13868 (aarch64_bitmask_imm): Use it.
13869 (aarch64_cannot_force_const_mem): Reject expressions involving
13870 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13871 (aarch64_classify_index): Handle SVE indices, by requiring
13872 a plain register index with a scale that matches the element size.
13873 (aarch64_classify_address): Handle SVE addresses. Assert that
13874 the mode of the address is VOIDmode or an integer mode.
13875 Update call to aarch64_classify_symbol.
13876 (aarch64_classify_symbolic_expression): Update call to
13877 aarch64_classify_symbol.
13878 (aarch64_const_vec_all_in_range_p): New function.
13879 (aarch64_print_vector_float_operand): Likewise.
13880 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13881 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13882 and the FP immediates 1.0 and 0.5.
13883 (aarch64_print_address_internal): Handle SVE addresses.
13884 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13885 (aarch64_regno_regclass): Handle predicate registers.
13886 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13888 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13889 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13890 (aarch64_convert_sve_vector_bits): New function.
13891 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13892 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13893 rather than an rtx.
13894 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13895 Handle SVE vector and predicate modes. Accept VL-based constants
13896 that need only one temporary register, and VL offsets that require
13897 no temporary registers.
13898 (aarch64_conditional_register_usage): Mark the predicate registers
13899 as fixed if SVE isn't available.
13900 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13901 Return true for SVE vector and predicate modes.
13902 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13903 rather than an unsigned int. Handle SVE modes.
13904 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13906 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13908 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13909 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13910 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13911 (aarch64_sve_float_mul_immediate_p): New functions.
13912 (aarch64_sve_valid_immediate): New function.
13913 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13914 Explicitly reject structure modes. Check for INDEX constants.
13915 Handle PTRUE and PFALSE constants.
13916 (aarch64_check_zero_based_sve_index_immediate): New function.
13917 (aarch64_simd_imm_zero_p): Delete.
13918 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13919 vector modes. Accept constants in the range of CNT[BHWD].
13920 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13921 ask for an Advanced SIMD mode.
13922 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13923 (aarch64_simd_vector_alignment): Handle SVE predicates.
13924 (aarch64_vectorize_preferred_vector_alignment): New function.
13925 (aarch64_simd_vector_alignment_reachable): Use it instead of
13927 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13928 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13930 (MAX_VECT_LEN): Delete.
13931 (expand_vec_perm_d): Add a vec_flags field.
13932 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13933 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13934 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13936 (aarch64_evpc_rev): Rename to...
13937 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13938 (aarch64_evpc_rev_global): New function.
13939 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13940 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13942 (aarch64_evpc_sve_tbl): New function.
13943 (aarch64_expand_vec_perm_const_1): Update after rename of
13944 aarch64_evpc_rev. Handle SVE permutes too, trying
13945 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13946 than aarch64_evpc_tbl.
13947 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13948 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13949 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13950 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13951 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13952 (aarch64_expand_sve_vcond): New functions.
13953 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13954 of aarch64_vector_mode_p.
13955 (aarch64_dwarf_poly_indeterminate_value): New function.
13956 (aarch64_compute_pressure_classes): Likewise.
13957 (aarch64_can_change_mode_class): Likewise.
13958 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13959 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13960 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13961 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13962 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13963 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13964 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13965 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13967 (Dn, Dl, Dr): Accept const as well as const_vector.
13968 (Dz): Likewise. Compare against CONST0_RTX.
13969 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13970 of "vector" where appropriate.
13971 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13972 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13973 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13974 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13975 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13976 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13977 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13978 (v_int_equiv): Extend to SVE modes.
13979 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13981 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13982 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13983 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13984 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13985 (SVE_COND_FP_CMP): New int iterators.
13986 (perm_hilo): Handle the new unpack unspecs.
13987 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13989 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13990 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13991 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13992 (aarch64_equality_operator, aarch64_constant_vector_operand)
13993 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13994 (aarch64_sve_nonimmediate_operand): Likewise.
13995 (aarch64_sve_general_operand): Likewise.
13996 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13997 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13998 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13999 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14000 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14001 (aarch64_sve_float_arith_immediate): Likewise.
14002 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14003 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14004 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14005 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14006 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14007 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14008 (aarch64_sve_float_arith_operand): Likewise.
14009 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14010 (aarch64_sve_float_mul_operand): Likewise.
14011 (aarch64_sve_vec_perm_operand): Likewise.
14012 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14013 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14014 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14015 as well as const_vector.
14016 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14017 in file. Use CONST0_RTX and CONSTM1_RTX.
14018 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14019 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14020 Use aarch64_simd_imm_zero.
14021 * config/aarch64/aarch64-sve.md: New file.
14022 * config/aarch64/aarch64.md: Include it.
14023 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14024 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14025 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14026 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14027 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14028 (sve): New attribute.
14029 (enabled): Disable instructions with the sve attribute unless
14031 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14032 aarch64_expand_mov_immediate.
14033 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14034 CNT[BHSD] immediates.
14035 (movti): Split CONST_POLY_INT moves into two halves.
14036 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14037 Split additions that need a temporary here if the destination
14038 is the stack pointer.
14039 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14040 (*add<mode>3_poly_1): New instruction.
14041 (set_clobber_cc): New expander.
14043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14045 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14046 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14047 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14048 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14049 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14050 Change innermode from fixed_mode_size to machine_mode.
14051 (simplify_subreg): Update call accordingly. Handle a constant-sized
14052 subreg of a variable-length CONST_VECTOR.
14054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14055 Alan Hayward <alan.hayward@arm.com>
14056 David Sherwood <david.sherwood@arm.com>
14058 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14059 (add_offset_to_base): New function, split out from...
14060 (create_mem_ref): ...here. When handling a scale other than 1,
14061 check first whether the address is valid without the offset.
14062 Add it into the base if so, leaving the index and scale as-is.
14064 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14067 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14068 fold_for_warn before checking if arg2 is INTEGER_CST.
14070 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14072 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14073 (store_multiple_operation): Delete.
14074 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14075 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14076 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14077 guarded by TARGET_STRING.
14078 (rs6000_output_load_multiple): Delete.
14079 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14080 OPTION_MASK_STRING / TARGET_STRING handling.
14081 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14082 (const rs6000_opt_masks) <"string">: Change mask to 0.
14083 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14084 (MASK_STRING): Delete.
14085 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14087 (load_multiple): Delete.
14094 (store_multiple): Delete.
14101 (movmemsi_8reg): Delete.
14102 (corresponding unnamed define_insn): Delete.
14103 (movmemsi_6reg): Delete.
14104 (corresponding unnamed define_insn): Delete.
14105 (movmemsi_4reg): Delete.
14106 (corresponding unnamed define_insn): Delete.
14107 (movmemsi_2reg): Delete.
14108 (corresponding unnamed define_insn): Delete.
14109 (movmemsi_1reg): Delete.
14110 (corresponding unnamed define_insn): Delete.
14111 * config/rs6000/rs6000.opt (mno-string): New.
14112 (mstring): Replace by deprecation warning stub.
14113 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14115 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14117 * regrename.c (regrename_do_replace): If replacing the same
14118 reg multiple times, try to reuse last created gen_raw_REG.
14121 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14122 main to workaround a bug in GDB.
14124 2018-01-12 Tom de Vries <tom@codesourcery.com>
14127 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14129 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14131 PR rtl-optimization/80481
14132 * ira-color.c (get_cap_member): New function.
14133 (allocnos_conflict_by_live_ranges_p): Use it.
14134 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14135 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14137 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14140 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14141 (*saddl_se_1): Ditto.
14142 (*ssubsi_1): Ditto.
14143 (*ssubl_se_1): Ditto.
14145 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14147 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14148 rather than wi::to_widest for DR_INITs.
14149 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14150 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14151 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14153 (vect_analyze_group_access_1): Note that here.
14155 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14157 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14158 polynomial type sizes.
14160 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14162 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14163 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14164 (gimple_add_tmp_var): Likewise.
14166 2018-01-12 Martin Liska <mliska@suse.cz>
14168 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14169 (gimple_alloc_sizes): Likewise.
14170 (dump_gimple_statistics): Use PRIu64 in printf format.
14171 * gimple.h: Change uint64_t to int.
14173 2018-01-12 Martin Liska <mliska@suse.cz>
14175 * tree-core.h: Use uint64_t instead of int.
14176 * tree.c (tree_node_counts): Likewise.
14177 (tree_node_sizes): Likewise.
14178 (dump_tree_statistics): Use PRIu64 in printf format.
14180 2018-01-12 Martin Liska <mliska@suse.cz>
14182 * Makefile.in: As qsort_chk is implemented in vec.c, add
14183 vec.o to linkage of gencfn-macros.
14184 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14185 passing the info to record_node_allocation_statistics.
14186 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14188 * ggc-common.c (struct ggc_usage): Add operator== and use
14189 it in operator< and compare function.
14190 * mem-stats.h (struct mem_usage): Likewise.
14191 * vec.c (struct vec_usage): Remove operator< and compare
14192 function. Can be simply inherited.
14194 2018-01-12 Martin Jambor <mjambor@suse.cz>
14197 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14198 * tree-ssa-math-opts.c: Include domwalk.h.
14199 (convert_mult_to_fma_1): New function.
14200 (fma_transformation_info): New type.
14201 (fma_deferring_state): Likewise.
14202 (cancel_fma_deferring): New function.
14203 (result_of_phi): Likewise.
14204 (last_fma_candidate_feeds_initial_phi): Likewise.
14205 (convert_mult_to_fma): Added deferring logic, split actual
14206 transformation to convert_mult_to_fma_1.
14207 (math_opts_dom_walker): New type.
14208 (math_opts_dom_walker::after_dom_children): New method, body moved
14209 here from pass_optimize_widening_mul::execute, added deferring logic
14211 (pass_optimize_widening_mul::execute): Moved most of code to
14212 math_opts_dom_walker::after_dom_children.
14213 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14214 * config/i386/i386.c (ix86_option_override_internal): Added
14215 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14217 2018-01-12 Richard Biener <rguenther@suse.de>
14220 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14221 inline instance vars.
14223 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14226 * config/rx/rx.c (rx_is_restricted_memory_address):
14227 Handle SUBREG case.
14229 2018-01-12 Richard Biener <rguenther@suse.de>
14231 PR tree-optimization/80846
14232 * target.def (split_reduction): New target hook.
14233 * targhooks.c (default_split_reduction): New function.
14234 * targhooks.h (default_split_reduction): Declare.
14235 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14236 target requests first reduce vectors by combining low and high
14238 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14239 (get_vectype_for_scalar_type_and_size): Export.
14240 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14241 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14242 * doc/tm.texi: Regenerate.
14243 * config/i386/i386.c (ix86_split_reduction): Implement
14244 TARGET_VECTORIZE_SPLIT_REDUCTION.
14246 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14249 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14250 in PIC mode except for TARGET_VXWORKS_RTP.
14251 * config/sparc/sparc.c: Include cfgrtl.h.
14252 (TARGET_INIT_PIC_REG): Define.
14253 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14254 (sparc_pic_register_p): New predicate.
14255 (sparc_legitimate_address_p): Use it.
14256 (sparc_legitimize_pic_address): Likewise.
14257 (sparc_delegitimize_address): Likewise.
14258 (sparc_mode_dependent_address_p): Likewise.
14259 (gen_load_pcrel_sym): Remove 4th parameter.
14260 (load_got_register): Adjust call to above. Remove obsolete stuff.
14261 (sparc_expand_prologue): Do not call load_got_register here.
14262 (sparc_flat_expand_prologue): Likewise.
14263 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14264 (sparc_use_pseudo_pic_reg): New function.
14265 (sparc_init_pic_reg): Likewise.
14266 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14267 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14269 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14271 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14272 Add item for branch_cost.
14274 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14276 PR rtl-optimization/83565
14277 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14278 not extend the result to a larger mode for rotate operations.
14279 (num_sign_bit_copies1): Likewise.
14281 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14284 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14286 Use values-Xc.o for -pedantic.
14287 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14289 2018-01-12 Martin Liska <mliska@suse.cz>
14292 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14294 (possible_polymorphic_call_targets): Use it.
14295 (ipa_devirt): Likewise.
14297 2018-01-12 Martin Liska <mliska@suse.cz>
14299 * profile-count.h (enum profile_quality): Use 0 as invalid
14300 enum value of profile_quality.
14302 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14304 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14305 -mext-string options.
14307 2018-01-12 Richard Biener <rguenther@suse.de>
14309 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14310 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14311 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14313 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14315 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14317 * configure.ac (--with-long-double-format): Add support for the
14318 configuration option to change the default long double format on
14320 * config.gcc (powerpc*-linux*-*): Likewise.
14321 * configure: Regenerate.
14322 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14323 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14324 used without modification.
14326 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14328 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14329 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14330 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14331 MISC_BUILTIN_SPEC_BARRIER.
14332 (rs6000_init_builtins): Likewise.
14333 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14335 (speculation_barrier): New define_insn.
14336 * doc/extend.texi: Document __builtin_speculation_barrier.
14338 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14341 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14342 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14343 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14345 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14346 integral modes instead of "ss" and "sd".
14347 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14348 vectors with 32-bit and 64-bit elements.
14349 (vecdupssescalarmodesuffix): New mode attribute.
14350 (vec_dup<mode>): Use it.
14352 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14355 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14356 frame if argument is passed on stack.
14358 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14361 * ree.c (combine_reaching_defs): Optimize also
14362 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14363 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14365 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14367 PR middle-end/83189
14368 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14370 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14372 PR middle-end/83718
14373 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14374 after they are computed.
14376 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14378 PR tree-optimization/83695
14379 * gimple-loop-linterchange.cc
14380 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14381 reset cached scev information after interchange.
14382 (pass_linterchange::execute): Remove call to scev_reset_htab.
14384 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14386 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14387 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14388 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14389 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14390 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14391 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14392 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14393 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14394 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14395 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14396 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14397 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14398 (V_lane_reg): Likewise.
14399 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14401 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14402 (vfmal_lane_low<mode>_intrinsic,
14403 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14404 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14405 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14406 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14407 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14408 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14410 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14412 * config/arm/arm-cpus.in (fp16fml): New feature.
14413 (ALL_SIMD): Add fp16fml.
14414 (armv8.2-a): Add fp16fml as an option.
14415 (armv8.3-a): Likewise.
14416 (armv8.4-a): Add fp16fml as part of fp16.
14417 * config/arm/arm.h (TARGET_FP16FML): Define.
14418 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14420 * config/arm/arm-modes.def (V2HF): Define.
14421 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14422 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14423 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14424 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14425 vfmsl_low, vfmsl_high): New set of builtins.
14426 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14427 (vfml_op): New code attribute.
14428 (VFMLHALVES): New int iterator.
14429 (VFML, VFMLSEL): New mode attributes.
14430 (V_reg): Define mapping for V2HF.
14431 (V_hi, V_lo): New mode attributes.
14432 (VF_constraint): Likewise.
14433 (vfml_half, vfml_half_selector): New int attributes.
14434 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14436 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14437 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14439 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14440 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14441 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14442 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14444 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14445 Document new effective target and option set.
14447 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14449 * config/arm/arm-cpus.in (armv8_4): New feature.
14450 (ARMv8_4a): New fgroup.
14451 (armv8.4-a): New arch.
14452 * config/arm/arm-tables.opt: Regenerate.
14453 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14454 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14455 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14456 Add matching rules for -march=armv8.4-a and extensions.
14457 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14459 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14462 * config/rx/rx.md (BW): New mode attribute.
14463 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14465 2018-01-11 Richard Biener <rguenther@suse.de>
14467 PR tree-optimization/83435
14468 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14469 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14470 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14472 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14473 Alan Hayward <alan.hayward@arm.com>
14474 David Sherwood <david.sherwood@arm.com>
14476 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14478 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14479 (aarch64_print_address_internal): Use it to check for a zero offset.
14481 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14482 Alan Hayward <alan.hayward@arm.com>
14483 David Sherwood <david.sherwood@arm.com>
14485 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14486 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14487 Return a poly_int64 rather than a HOST_WIDE_INT.
14488 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14489 rather than a HOST_WIDE_INT.
14490 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14491 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14492 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14493 final_offset from HOST_WIDE_INT to poly_int64.
14494 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14495 to_constant when getting the number of units in an Advanced SIMD
14497 (aarch64_builtin_vectorized_function): Check for a constant number
14499 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14501 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14502 attribute instead of GET_MODE_NUNITS.
14503 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14504 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14505 GET_MODE_SIZE for fixed-size registers.
14506 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14507 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14508 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14509 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14510 (aarch64_print_operand, aarch64_print_address_internal)
14511 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14512 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14513 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14514 Handle polynomial GET_MODE_SIZE.
14515 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14516 wider than SImode without modification.
14517 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14518 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14519 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14520 passing and returning SVE modes.
14521 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14522 rather than GEN_INT.
14523 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14524 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14525 (aarch64_allocate_and_probe_stack_space): Likewise.
14526 (aarch64_layout_frame): Cope with polynomial offsets.
14527 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14528 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14529 polynomial offsets.
14530 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14531 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14532 poly_int64 rather than a HOST_WIDE_INT.
14533 (aarch64_get_separate_components, aarch64_process_components)
14534 (aarch64_expand_prologue, aarch64_expand_epilogue)
14535 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14536 (aarch64_anchor_offset): New function, split out from...
14537 (aarch64_legitimize_address): ...here.
14538 (aarch64_builtin_vectorization_cost): Handle polynomial
14539 TYPE_VECTOR_SUBPARTS.
14540 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14542 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14543 number of elements from the PARALLEL rather than the mode.
14544 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14545 rather than GET_MODE_BITSIZE.
14546 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14547 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14548 (aarch64_expand_vec_perm_const_1): Handle polynomial
14549 d->perm.length () and d->perm elements.
14550 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14551 Apply to_constant to d->perm elements.
14552 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14553 polynomial CONST_VECTOR_NUNITS.
14554 (aarch64_move_pointer): Take amount as a poly_int64 rather
14556 (aarch64_progress_pointer): Avoid temporary variable.
14557 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14558 the mode attribute instead of GET_MODE.
14560 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14565 x exists before using it.
14566 (aarch64_add_constant_internal): Rename to...
14567 (aarch64_add_offset_1): ...this. Replace regnum with separate
14568 src and dest rtxes. Handle the case in which they're different,
14569 including when the offset is zero. Replace scratchreg with an rtx.
14570 Use 2 additions if there is no spare register into which we can
14571 move a 16-bit constant.
14572 (aarch64_add_constant): Delete.
14573 (aarch64_add_offset): Replace reg with separate src and dest
14574 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14575 Use aarch64_add_offset_1.
14576 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14577 an rtx rather than an int. Take the delta as a poly_int64
14578 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14579 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14580 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14581 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14582 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14583 and aarch64_add_sp.
14584 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14585 aarch64_add_constant.
14587 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14589 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14590 Use scalar_float_mode.
14592 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14594 * config/aarch64/aarch64-simd.md
14595 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14596 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14597 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14598 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14599 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14600 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14601 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14602 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14603 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14604 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14606 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14609 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14610 targ_options->x_arm_arch_string is non NULL.
14612 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14614 * config/aarch64/aarch64.h
14615 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14617 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14620 * expmed.c (emit_store_flag_force): Swap if const op0
14621 and change VOIDmode to mode of op0.
14623 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14625 PR rtl-optimization/83761
14626 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14627 than bytes to mode_for_size.
14629 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14631 PR middle-end/83189
14632 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14633 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14636 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14638 PR middle-end/83575
14639 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14640 when in layout mode.
14641 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14642 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14645 2018-01-10 Michael Collison <michael.collison@arm.com>
14647 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14648 * config/aarch64/aarch64-option-extension.def: Add
14649 AARCH64_OPT_EXTENSION of 'fp16fml'.
14650 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14651 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14652 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14653 * config/aarch64/constraints.md (Ui7): New constraint.
14654 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14655 (VFMLA_SEL_W): Ditto.
14658 (VFMLA16_LOW): New int iterator.
14659 (VFMLA16_HIGH): Ditto.
14660 (UNSPEC_FMLAL): New unspec.
14661 (UNSPEC_FMLSL): Ditto.
14662 (UNSPEC_FMLAL2): Ditto.
14663 (UNSPEC_FMLSL2): Ditto.
14664 (f16mac): New code attribute.
14665 * config/aarch64/aarch64-simd-builtins.def
14666 (aarch64_fmlal_lowv2sf): Ditto.
14667 (aarch64_fmlsl_lowv2sf): Ditto.
14668 (aarch64_fmlalq_lowv4sf): Ditto.
14669 (aarch64_fmlslq_lowv4sf): Ditto.
14670 (aarch64_fmlal_highv2sf): Ditto.
14671 (aarch64_fmlsl_highv2sf): Ditto.
14672 (aarch64_fmlalq_highv4sf): Ditto.
14673 (aarch64_fmlslq_highv4sf): Ditto.
14674 (aarch64_fmlal_lane_lowv2sf): Ditto.
14675 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14676 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14677 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14678 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14679 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14680 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14681 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14682 (aarch64_fmlal_lane_highv2sf): Ditto.
14683 (aarch64_fmlsl_lane_highv2sf): Ditto.
14684 (aarch64_fmlal_laneq_highv2sf): Ditto.
14685 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14686 (aarch64_fmlalq_lane_highv4sf): Ditto.
14687 (aarch64_fmlsl_lane_highv4sf): Ditto.
14688 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14689 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14690 * config/aarch64/aarch64-simd.md:
14691 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14692 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14693 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14694 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14695 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14696 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14697 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14698 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14699 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14700 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14701 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14702 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14703 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14704 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14705 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14706 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14707 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14708 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14709 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14710 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14711 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14712 (vfmlsl_low_u32): Ditto.
14713 (vfmlalq_low_u32): Ditto.
14714 (vfmlslq_low_u32): Ditto.
14715 (vfmlal_high_u32): Ditto.
14716 (vfmlsl_high_u32): Ditto.
14717 (vfmlalq_high_u32): Ditto.
14718 (vfmlslq_high_u32): Ditto.
14719 (vfmlal_lane_low_u32): Ditto.
14720 (vfmlsl_lane_low_u32): Ditto.
14721 (vfmlal_laneq_low_u32): Ditto.
14722 (vfmlsl_laneq_low_u32): Ditto.
14723 (vfmlalq_lane_low_u32): Ditto.
14724 (vfmlslq_lane_low_u32): Ditto.
14725 (vfmlalq_laneq_low_u32): Ditto.
14726 (vfmlslq_laneq_low_u32): Ditto.
14727 (vfmlal_lane_high_u32): Ditto.
14728 (vfmlsl_lane_high_u32): Ditto.
14729 (vfmlal_laneq_high_u32): Ditto.
14730 (vfmlsl_laneq_high_u32): Ditto.
14731 (vfmlalq_lane_high_u32): Ditto.
14732 (vfmlslq_lane_high_u32): Ditto.
14733 (vfmlalq_laneq_high_u32): Ditto.
14734 (vfmlslq_laneq_high_u32): Ditto.
14735 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14736 (AARCH64_FL_FOR_ARCH8_4): New.
14737 (AARCH64_ISA_F16FML): New ISA flag.
14738 (TARGET_F16FML): New feature flag for fp16fml.
14739 (doc/invoke.texi): Document new fp16fml option.
14741 2018-01-10 Michael Collison <michael.collison@arm.com>
14743 * config/aarch64/aarch64-builtins.c:
14744 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14745 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14746 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14747 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14748 (AARCH64_ISA_SHA3): New ISA flag.
14749 (TARGET_SHA3): New feature flag for sha3.
14750 * config/aarch64/iterators.md (sha512_op): New int attribute.
14751 (CRYPTO_SHA512): New int iterator.
14752 (UNSPEC_SHA512H): New unspec.
14753 (UNSPEC_SHA512H2): Ditto.
14754 (UNSPEC_SHA512SU0): Ditto.
14755 (UNSPEC_SHA512SU1): Ditto.
14756 * config/aarch64/aarch64-simd-builtins.def
14757 (aarch64_crypto_sha512hqv2di): New builtin.
14758 (aarch64_crypto_sha512h2qv2di): Ditto.
14759 (aarch64_crypto_sha512su0qv2di): Ditto.
14760 (aarch64_crypto_sha512su1qv2di): Ditto.
14761 (aarch64_eor3qv8hi): Ditto.
14762 (aarch64_rax1qv2di): Ditto.
14763 (aarch64_xarqv2di): Ditto.
14764 (aarch64_bcaxqv8hi): Ditto.
14765 * config/aarch64/aarch64-simd.md:
14766 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14767 (aarch64_crypto_sha512su0qv2di): Ditto.
14768 (aarch64_crypto_sha512su1qv2di): Ditto.
14769 (aarch64_eor3qv8hi): Ditto.
14770 (aarch64_rax1qv2di): Ditto.
14771 (aarch64_xarqv2di): Ditto.
14772 (aarch64_bcaxqv8hi): Ditto.
14773 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14774 (vsha512h2q_u64): Ditto.
14775 (vsha512su0q_u64): Ditto.
14776 (vsha512su1q_u64): Ditto.
14777 (veor3q_u16): Ditto.
14778 (vrax1q_u64): Ditto.
14779 (vxarq_u64): Ditto.
14780 (vbcaxq_u16): Ditto.
14781 * config/arm/types.md (crypto_sha512): New type attribute.
14782 (crypto_sha3): Ditto.
14783 (doc/invoke.texi): Document new sha3 option.
14785 2018-01-10 Michael Collison <michael.collison@arm.com>
14787 * config/aarch64/aarch64-builtins.c:
14788 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14789 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14790 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14791 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14792 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14793 (AARCH64_ISA_SM4): New ISA flag.
14794 (TARGET_SM4): New feature flag for sm4.
14795 * config/aarch64/aarch64-simd-builtins.def
14796 (aarch64_sm3ss1qv4si): Ditto.
14797 (aarch64_sm3tt1aq4si): Ditto.
14798 (aarch64_sm3tt1bq4si): Ditto.
14799 (aarch64_sm3tt2aq4si): Ditto.
14800 (aarch64_sm3tt2bq4si): Ditto.
14801 (aarch64_sm3partw1qv4si): Ditto.
14802 (aarch64_sm3partw2qv4si): Ditto.
14803 (aarch64_sm4eqv4si): Ditto.
14804 (aarch64_sm4ekeyqv4si): Ditto.
14805 * config/aarch64/aarch64-simd.md:
14806 (aarch64_sm3ss1qv4si): Ditto.
14807 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14808 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14809 (aarch64_sm4eqv4si): Ditto.
14810 (aarch64_sm4ekeyqv4si): Ditto.
14811 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14812 (sm3part_op): Ditto.
14813 (CRYPTO_SM3TT): Ditto.
14814 (CRYPTO_SM3PART): Ditto.
14815 (UNSPEC_SM3SS1): New unspec.
14816 (UNSPEC_SM3TT1A): Ditto.
14817 (UNSPEC_SM3TT1B): Ditto.
14818 (UNSPEC_SM3TT2A): Ditto.
14819 (UNSPEC_SM3TT2B): Ditto.
14820 (UNSPEC_SM3PARTW1): Ditto.
14821 (UNSPEC_SM3PARTW2): Ditto.
14822 (UNSPEC_SM4E): Ditto.
14823 (UNSPEC_SM4EKEY): Ditto.
14824 * config/aarch64/constraints.md (Ui2): New constraint.
14825 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14826 * config/arm/types.md (crypto_sm3): New type attribute.
14827 (crypto_sm4): Ditto.
14828 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14829 (vsm3tt1aq_u32): Ditto.
14830 (vsm3tt1bq_u32): Ditto.
14831 (vsm3tt2aq_u32): Ditto.
14832 (vsm3tt2bq_u32): Ditto.
14833 (vsm3partw1q_u32): Ditto.
14834 (vsm3partw2q_u32): Ditto.
14835 (vsm4eq_u32): Ditto.
14836 (vsm4ekeyq_u32): Ditto.
14837 (doc/invoke.texi): Document new sm4 option.
14839 2018-01-10 Michael Collison <michael.collison@arm.com>
14841 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14842 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14843 (AARCH64_FL_FOR_ARCH8_4): New.
14844 (AARCH64_FL_V8_4): New flag.
14845 (doc/invoke.texi): Document new armv8.4-a option.
14847 2018-01-10 Michael Collison <michael.collison@arm.com>
14849 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14850 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14851 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14852 * config/aarch64/aarch64-option-extension.def: Add
14853 AARCH64_OPT_EXTENSION of 'sha2'.
14854 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14855 (crypto): Disable sha2 and aes if crypto disabled.
14856 (crypto): Enable aes and sha2 if enabled.
14857 (simd): Disable sha2 and aes if simd disabled.
14858 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14860 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14861 (TARGET_SHA2): New feature flag for sha2.
14862 (TARGET_AES): New feature flag for aes.
14863 * config/aarch64/aarch64-simd.md:
14864 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14865 conditional on TARGET_AES.
14866 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14867 (aarch64_crypto_sha1hsi): Make pattern conditional
14869 (aarch64_crypto_sha1hv4si): Ditto.
14870 (aarch64_be_crypto_sha1hv4si): Ditto.
14871 (aarch64_crypto_sha1su1v4si): Ditto.
14872 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14873 (aarch64_crypto_sha1su0v4si): Ditto.
14874 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14875 (aarch64_crypto_sha256su0v4si): Ditto.
14876 (aarch64_crypto_sha256su1v4si): Ditto.
14877 (doc/invoke.texi): Document new aes and sha2 options.
14879 2018-01-10 Martin Sebor <msebor@redhat.com>
14881 PR tree-optimization/83781
14882 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14885 2018-01-11 Martin Sebor <msebor@gmail.com>
14886 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14888 PR tree-optimization/83501
14889 PR tree-optimization/81703
14891 * tree-ssa-strlen.c (get_string_cst): Rename...
14892 (get_string_len): ...to this. Handle global constants.
14893 (handle_char_store): Adjust.
14895 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14896 Jim Wilson <jimw@sifive.com>
14898 * config/riscv/riscv-protos.h (riscv_output_return): New.
14899 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14900 (riscv_attribute_table, riscv_output_return),
14901 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14902 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14903 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14905 (riscv_expand_prologue): Add early return for naked function.
14906 (riscv_expand_epilogue): Likewise.
14907 (riscv_function_ok_for_sibcall): Return false for naked function.
14908 (riscv_set_current_function): New.
14909 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14910 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14911 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14912 * doc/extend.texi (RISC-V Function Attributes): New.
14914 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14916 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14917 check for 128-bit long double before checking TCmode.
14918 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14919 128-bit long doubles before checking TFmode or TCmode.
14920 (FLOAT128_IBM_P): Likewise.
14922 2018-01-10 Martin Sebor <msebor@redhat.com>
14924 PR tree-optimization/83671
14925 * builtins.c (c_strlen): Unconditionally return zero for the empty
14927 Use -Warray-bounds for warnings.
14928 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14929 for non-constant array indices with COMPONENT_REF, arrays of
14930 arrays, and pointers to arrays.
14931 (gimple_fold_builtin_strlen): Determine and set length range for
14932 non-constant character arrays.
14934 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14936 PR middle-end/81897
14937 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14940 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14942 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14944 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14947 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14948 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14949 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14950 indexed_or_indirect_operand predicate.
14951 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14952 (*vsx_le_perm_load_v8hi): Likewise.
14953 (*vsx_le_perm_load_v16qi): Likewise.
14954 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14955 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14956 (*vsx_le_perm_store_v8hi): Likewise.
14957 (*vsx_le_perm_store_v16qi): Likewise.
14958 (eight unnamed splitters): Likewise.
14960 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14962 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14963 * config/rs6000/emmintrin.h: Likewise.
14964 * config/rs6000/mmintrin.h: Likewise.
14965 * config/rs6000/xmmintrin.h: Likewise.
14967 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14970 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14972 * tree.c (tree_nop_conversion): Return true for location wrapper
14974 (maybe_wrap_with_location): New function.
14975 (selftest::check_strip_nops): New function.
14976 (selftest::test_location_wrappers): New function.
14977 (selftest::tree_c_tests): Call it.
14978 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14979 (maybe_wrap_with_location): New decl.
14980 (EXPR_LOCATION_WRAPPER_P): New macro.
14981 (location_wrapper_p): New inline function.
14982 (tree_strip_any_location_wrapper): New inline function.
14984 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14987 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14988 stack_realign_offset for the largest alignment of stack slot
14990 (ix86_find_max_used_stack_alignment): New function.
14991 (ix86_finalize_stack_frame_flags): Use it. Set
14992 max_used_stack_alignment if we don't realign stack.
14993 * config/i386/i386.h (machine_function): Add
14994 max_used_stack_alignment.
14996 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14998 * config/arm/arm.opt (-mbranch-cost): New option.
14999 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15002 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15005 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15006 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15008 2018-01-10 Richard Biener <rguenther@suse.de>
15011 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15012 early out so it also covers the case where we have a non-NULL
15015 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15017 PR tree-optimization/83753
15018 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15019 for non-strided grouped accesses if the number of elements is 1.
15021 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15024 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15025 * i386.h (TARGET_USE_GATHER): Define.
15026 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15028 2018-01-10 Martin Liska <mliska@suse.cz>
15031 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15032 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15034 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15035 CLEANUP_NO_PARTITIONING is not set.
15037 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15039 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15040 for vectors, as a partial revert of r254296.
15041 * rtl.h (const_vec_p): Delete.
15042 (const_vec_duplicate_p): Don't test for vector CONSTs.
15043 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15044 * expmed.c (make_tree): Likewise.
15047 * common.md (E, F): Use CONSTANT_P instead of checking for
15049 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15050 checking for CONST_VECTOR.
15052 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15054 PR middle-end/83575
15055 * predict.c (force_edge_cold): Handle in more sane way edges
15056 with no prediction.
15058 2018-01-09 Carl Love <cel@us.ibm.com>
15060 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15062 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15063 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15064 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15065 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15066 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15067 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15068 * config/rs6000/rs6000-protos.h: Add extern defition for
15069 rs6000_generate_float2_double_code.
15070 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15072 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15073 (float2_v2df): Add define_expand.
15075 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15078 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15079 op_mode in the force_to_mode call.
15081 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15083 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15084 instead of checking each element individually.
15085 (aarch64_evpc_uzp): Likewise.
15086 (aarch64_evpc_zip): Likewise.
15087 (aarch64_evpc_ext): Likewise.
15088 (aarch64_evpc_rev): Likewise.
15089 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15090 instead of checking each element individually. Return true without
15092 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15093 whether all selected elements come from the same input, instead of
15094 checking each element individually. Remove calls to gen_rtx_REG,
15095 start_sequence and end_sequence and instead assert that no rtl is
15098 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15100 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15101 order of HIGH and CONST checks.
15103 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15105 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15106 if the destination isn't an SSA_NAME.
15108 2018-01-09 Richard Biener <rguenther@suse.de>
15110 PR tree-optimization/83668
15111 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15113 (canonicalize_loop_form): ... here, renamed from ...
15114 (canonicalize_loop_closed_ssa_form): ... this and amended to
15115 swap successor edges for loop exit blocks to make us use
15116 the RPO order we need for initial schedule generation.
15118 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15120 PR tree-optimization/64811
15121 * match.pd: When optimizing comparisons with Inf, avoid
15122 introducing or losing exceptions from comparisons with NaN.
15124 2018-01-09 Martin Liska <mliska@suse.cz>
15127 * asan.c (shadow_mem_size): Add gcc_assert.
15129 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15131 Don't save registers in main().
15134 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15135 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15136 * config/avr/avr.c (avr_set_current_function): Don't error if
15137 naked, OS_task or OS_main are specified at the same time.
15138 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15140 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15142 * common/config/avr/avr-common.c (avr_option_optimization_table):
15143 Switch on -mmain-is-OS_task for optimizing compilations.
15145 2018-01-09 Richard Biener <rguenther@suse.de>
15147 PR tree-optimization/83572
15148 * graphite.c: Include cfganal.h.
15149 (graphite_transform_loops): Connect infinite loops to exit
15150 and remove fake edges at the end.
15152 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15154 * ipa-inline.c (edge_badness): Revert accidental checkin.
15156 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15159 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15160 symbols; not inline clones.
15162 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15165 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15166 hard registers. Formatting fixes.
15168 PR preprocessor/83722
15169 * gcc.c (try_generate_repro): Pass
15170 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15171 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15174 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15175 Kito Cheng <kito.cheng@gmail.com>
15177 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15178 (riscv_leaf_function_p): Delete.
15179 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15181 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15183 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15185 (do_ifelse): New function.
15186 (do_isel): New function.
15187 (do_sub3): New function.
15188 (do_add3): New function.
15189 (do_load_mask_compare): New function.
15190 (do_overlap_load_compare): New function.
15191 (expand_compare_loop): New function.
15192 (expand_block_compare): Call expand_compare_loop() when appropriate.
15193 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15194 option description.
15195 (-mblock-compare-inline-loop-limit): New option.
15197 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15200 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15201 Reverse order of second and third operands in first alternative.
15202 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15203 of first and second elements in UNSPEC_VPERMR vector.
15204 (altivec_expand_vec_perm_le): Likewise.
15206 2018-01-08 Jeff Law <law@redhat.com>
15208 PR rtl-optimizatin/81308
15209 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15210 (process_switch): If group_case_labels makes a change, then set
15212 (pass_convert_switch::execute): If a switch is converted, then
15213 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15215 PR rtl-optimization/81308
15216 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15219 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15221 PR target/83663 - Revert r255946
15222 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15223 generation for cases where splatting a value is not useful.
15224 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15225 across a vec_duplicate and a paradoxical subreg forming a vector
15226 mode to a vec_concat.
15228 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15230 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15231 -march=armv8.3-a variants.
15232 * config/arm/t-multilib: Likewise.
15233 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15235 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15237 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15239 (cceq_ior_compare_complement): Give it a name so I can use it, and
15240 change boolean_or_operator predicate to boolean_operator so it can
15241 be used to generate a crand.
15242 (eqne): New code iterator.
15243 (bd/bd_neg): New code_attrs.
15244 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15245 a single define_insn.
15246 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15247 decrement (bdnzt/bdnzf/bdzt/bdzf).
15248 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15249 with the new names of the branch decrement patterns, and added the
15250 names of the branch decrement conditional patterns.
15252 2018-01-08 Richard Biener <rguenther@suse.de>
15254 PR tree-optimization/83563
15255 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15258 2018-01-08 Richard Biener <rguenther@suse.de>
15260 PR middle-end/83713
15261 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15263 2018-01-08 Richard Biener <rguenther@suse.de>
15265 PR tree-optimization/83685
15266 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15267 references to abnormals.
15269 2018-01-08 Richard Biener <rguenther@suse.de>
15272 * dwarf2out.c (output_indirect_strings): Handle empty
15273 skeleton_debug_str_hash.
15274 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15276 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15278 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15279 (emit_store_direct): Likewise.
15280 (arc_trampoline_adjust_address): Likewise.
15281 (arc_asm_trampoline_template): New function.
15282 (arc_initialize_trampoline): Use asm_trampoline_template.
15283 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15284 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15285 * config/arc/arc.md (flush_icache): Delete pattern.
15287 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15289 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15290 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15293 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15296 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15297 by not USED_FOR_TARGET.
15298 (make_pass_resolve_sw_modes): Likewise.
15300 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15302 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15305 2018-01-08 Richard Biener <rguenther@suse.de>
15307 PR middle-end/83580
15308 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15310 2018-01-08 Richard Biener <rguenther@suse.de>
15312 PR middle-end/83517
15313 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15315 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15317 PR middle-end/81897
15318 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15319 basic blocks with a small number of successors.
15320 (convert_control_dep_chain_into_preds): Improve handling of
15322 (dump_predicates): Split apart into...
15323 (dump_pred_chain): ...here...
15324 (dump_pred_info): ...and here.
15325 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15326 (can_chain_union_be_invalidated_p): Improve check for invalidation
15328 (uninit_uses_cannot_happen): Avoid unnecessary if
15329 convert_control_dep_chain_into_preds yielded nothing.
15331 2018-01-06 Martin Sebor <msebor@redhat.com>
15333 PR tree-optimization/83640
15334 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15335 subtracting negative offset from size.
15336 (builtin_access::overlap): Adjust offset bounds of the access to fall
15337 within the size of the object if possible.
15339 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15341 PR rtl-optimization/83699
15342 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15343 extract_bit_field_as_subreg to cases in which the extracted
15344 value is also a vector.
15346 * lra-constraints.c (process_alt_operands): Test for the equivalence
15347 substitutions when detecting a possible reload cycle.
15349 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15352 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15353 by default if flag_selective_schedling{,2}. Formatting fixes.
15355 PR rtl-optimization/83682
15356 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15357 if it has non-VECTOR_MODE element mode.
15358 (vec_duplicate_p): Likewise.
15360 PR middle-end/83694
15361 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15362 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15364 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15367 * config/i386/i386-builtin.def
15368 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15369 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15370 Require also OPTION_MASK_ISA_AVX512F in addition to
15371 OPTION_MASK_ISA_GFNI.
15372 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15373 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15374 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15375 to OPTION_MASK_ISA_GFNI.
15376 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15377 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15378 OPTION_MASK_ISA_AVX512BW.
15379 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15380 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15381 addition to OPTION_MASK_ISA_GFNI.
15382 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15383 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15384 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15385 to OPTION_MASK_ISA_GFNI.
15386 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15387 a requirement for all ISAs rather than any of them with a few
15389 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15391 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15392 bitmasks to be enabled with 3 exceptions, instead of requiring any
15393 enabled ISA with lots of exceptions.
15394 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15395 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15396 Change avx512bw in isa attribute to avx512f.
15397 * config/i386/sgxintrin.h: Add license boilerplate.
15398 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15399 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15400 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15401 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15403 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15404 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15405 temporarily sse2 rather than sse if not enabled already.
15408 * config/i386/sse.md (VI248_VLBW): Rename to ...
15409 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15410 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15411 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15412 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15413 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15414 mode iterator instead of VI248_VLBW.
15416 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15418 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15419 (record_modified): Skip clobbers; add debug output.
15420 (param_change_prob): Use sreal frequencies.
15422 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15424 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15425 punt for user-aligned variables.
15427 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15429 * tree-chrec.c (chrec_contains_symbols): Return true for
15432 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15435 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15436 of (x|y) == x for BICS pattern.
15438 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15440 PR tree-optimization/83605
15441 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15442 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15445 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15447 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15448 * config/epiphany/rtems.h: New file.
15450 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15451 Uros Bizjak <ubizjak@gmail.com>
15454 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15455 QIreg_operand instead of register_operand predicate.
15456 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15457 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15458 comments instead of -fmitigate[-_]rop.
15460 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15463 * cgraphunit.c (symbol_table::compile): Switch to text_section
15464 before calling assembly_start debug hook.
15465 * run-rtl-passes.c (run_rtl_passes): Likewise.
15468 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15470 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15471 range_int_cst_p rather than !symbolic_range_p before calling
15472 extract_range_from_multiplicative_op_1.
15474 2018-01-04 Jeff Law <law@redhat.com>
15476 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15477 redundant test in assertion.
15479 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15481 * doc/rtl.texi: Document machine_mode wrapper classes.
15483 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15485 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15486 using tree_to_uhwi.
15488 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15490 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15491 the VEC_PERM_EXPR fold to fail.
15493 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15496 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15497 to switched_sections.
15499 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15502 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15503 test for d.testing.
15505 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15508 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15509 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15511 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15514 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15515 is BLKmode and bitpos not zero or mode change is needed.
15517 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15520 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15523 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15526 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15527 instead of MULT rtx. Update all corresponding splitters.
15528 (*saddl_se): Ditto.
15529 (*ssub<modesuffix>): Ditto.
15530 (*ssubl_se): Ditto.
15531 (*cmp_sadd_di): Update split patterns.
15532 (*cmp_sadd_si): Ditto.
15533 (*cmp_sadd_sidi): Ditto.
15534 (*cmp_ssub_di): Ditto.
15535 (*cmp_ssub_si): Ditto.
15536 (*cmp_ssub_sidi): Ditto.
15537 * config/alpha/predicates.md (const23_operand): New predicate.
15538 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15539 Look for ASHIFT, not MULT inner operand.
15540 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15542 2018-01-04 Martin Liska <mliska@suse.cz>
15544 PR gcov-profile/83669
15545 * gcov.c (output_intermediate_file): Add version to intermediate
15547 * doc/gcov.texi: Document new field 'version' in intermediate
15548 file format. Fix location of '-k' option of gcov command.
15550 2018-01-04 Martin Liska <mliska@suse.cz>
15553 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15555 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15557 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15559 2018-01-03 Martin Sebor <msebor@redhat.com>
15561 PR tree-optimization/83655
15562 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15563 checking calls with invalid arguments.
15565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15567 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15568 (vectorizable_mask_load_store): Delete.
15569 (vectorizable_call): Return false for masked loads and stores.
15570 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15571 instead of gimple_assign_rhs1.
15572 (vectorizable_load): Handle IFN_MASK_LOAD.
15573 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15577 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15579 (vectorizable_mask_load_store): ...here.
15580 (vectorizable_load): ...and here.
15582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15584 * tree-vect-stmts.c (vect_build_all_ones_mask)
15585 (vect_build_zero_merge_argument): New functions, split out from...
15586 (vectorizable_load): ...here.
15588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15590 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15592 (vectorizable_mask_load_store): ...here.
15593 (vectorizable_store): ...and here.
15595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15597 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15599 (vectorizable_mask_load_store): ...here.
15601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15603 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15604 (vect_model_store_cost): Take a vec_load_store_type instead of a
15606 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15607 (vect_model_store_cost): Take a vec_load_store_type instead of a
15609 (vectorizable_mask_load_store): Update accordingly.
15610 (vectorizable_store): Likewise.
15611 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15615 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15616 IFN_MASK_LOAD calls here rather than...
15617 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15620 Alan Hayward <alan.hayward@arm.com>
15621 David Sherwood <david.sherwood@arm.com>
15623 * expmed.c (extract_bit_field_1): For vector extracts,
15624 fall back to extract_bit_field_as_subreg if vec_extract
15627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15628 Alan Hayward <alan.hayward@arm.com>
15629 David Sherwood <david.sherwood@arm.com>
15631 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15632 they are variable or constant sized.
15633 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15634 slots for constant-sized data.
15636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15637 Alan Hayward <alan.hayward@arm.com>
15638 David Sherwood <david.sherwood@arm.com>
15640 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15641 handling COND_EXPRs with boolean comparisons, try to find a better
15642 basis for the mask type than the boolean itself.
15644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15646 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15647 is calculated and how it can be overridden.
15648 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15649 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15651 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15655 Alan Hayward <alan.hayward@arm.com>
15656 David Sherwood <david.sherwood@arm.com>
15658 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15659 Remove the mode argument.
15660 (aarch64_simd_valid_immediate): Remove the mode and inverse
15662 * config/aarch64/iterators.md (bitsize): New iterator.
15663 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15664 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15665 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15666 aarch64_simd_valid_immediate.
15667 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15668 (aarch64_reg_or_bic_imm): Likewise.
15669 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15670 with an insn_type enum and msl with a modifier_type enum.
15671 Replace element_width with a scalar_mode. Change the shift
15672 to unsigned int. Add constructors for scalar_float_mode and
15673 scalar_int_mode elements.
15674 (aarch64_vect_float_const_representable_p): Delete.
15675 (aarch64_can_const_movi_rtx_p)
15676 (aarch64_simd_scalar_immediate_valid_for_move)
15677 (aarch64_simd_make_constant): Update call to
15678 aarch64_simd_valid_immediate.
15679 (aarch64_advsimd_valid_immediate_hs): New function.
15680 (aarch64_advsimd_valid_immediate): Likewise.
15681 (aarch64_simd_valid_immediate): Remove mode and inverse
15682 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15683 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15684 and aarch64_float_const_representable_p on the result.
15685 (aarch64_output_simd_mov_immediate): Remove mode argument.
15686 Update call to aarch64_simd_valid_immediate and use of
15687 simd_immediate_info.
15688 (aarch64_output_scalar_simd_mov_immediate): Update call
15691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15692 Alan Hayward <alan.hayward@arm.com>
15693 David Sherwood <david.sherwood@arm.com>
15695 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15696 (mode_nunits): Likewise CONST_MODE_NUNITS.
15697 * machmode.def (ADJUST_NUNITS): Document.
15698 * genmodes.c (mode_data::need_nunits_adj): New field.
15699 (blank_mode): Update accordingly.
15700 (adj_nunits): New variable.
15701 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15703 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15704 listed in adj_nunits.
15705 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15706 listed in adj_nunits. Don't emit case statements for such modes.
15707 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15708 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15709 nothing if adj_nunits is nonnull.
15710 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15711 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15712 (emit_mode_fbit): Update use of print_maybe_const_decl.
15713 (emit_move_size): Likewise. Treat the array as non-const
15715 (emit_mode_adjustments): Handle adj_nunits.
15717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15719 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15720 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15721 (VECTOR_MODES): Use it.
15722 (make_vector_modes): Take the prefix as an argument.
15724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15725 Alan Hayward <alan.hayward@arm.com>
15726 David Sherwood <david.sherwood@arm.com>
15728 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15729 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15730 for MODE_VECTOR_BOOL.
15731 * machmode.def (VECTOR_BOOL_MODE): Document.
15732 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15733 (make_vector_bool_mode): New function.
15734 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15736 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15737 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15739 * stor-layout.c (int_mode_for_mode): Likewise.
15740 * tree.c (build_vector_type_for_mode): Likewise.
15741 * varasm.c (output_constant_pool_2): Likewise.
15742 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15743 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15744 for MODE_VECTOR_BOOL.
15745 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15746 of mode class checks.
15747 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15748 instead of a list of mode class checks.
15749 (expand_vector_scalar_condition): Likewise.
15750 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15753 Alan Hayward <alan.hayward@arm.com>
15754 David Sherwood <david.sherwood@arm.com>
15756 * machmode.h (mode_size): Change from unsigned short to
15758 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15759 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15760 or if measurement_type is not polynomial.
15761 (fixed_size_mode::includes_p): Check for constant-sized modes.
15762 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15763 return a poly_uint16 rather than an unsigned short.
15764 (emit_mode_size): Change the type of mode_size from unsigned short
15765 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15766 (emit_mode_adjustments): Cope with polynomial vector sizes.
15767 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15769 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15771 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15772 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15773 * caller-save.c (setup_save_areas): Likewise.
15774 (replace_reg_with_saved_mem): Likewise.
15775 * calls.c (emit_library_call_value_1): Likewise.
15776 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15777 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15778 (gen_lowpart_for_combine): Likewise.
15779 * convert.c (convert_to_integer_1): Likewise.
15780 * cse.c (equiv_constant, cse_insn): Likewise.
15781 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15782 (cselib_subst_to_values): Likewise.
15783 * dce.c (word_dce_process_block): Likewise.
15784 * df-problems.c (df_word_lr_mark_ref): Likewise.
15785 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15786 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15787 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15788 (rtl_for_decl_location): Likewise.
15789 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15790 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15791 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15792 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15793 (expand_expr_real_1): Likewise.
15794 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15795 (pad_below): Likewise.
15796 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15797 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15798 * ira.c (get_subreg_tracking_sizes): Likewise.
15799 * ira-build.c (ira_create_allocno_objects): Likewise.
15800 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15801 (ira_sort_regnos_for_alter_reg): Likewise.
15802 * ira-costs.c (record_operand_costs): Likewise.
15803 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15804 (resolve_simple_move): Likewise.
15805 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15806 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15807 (lra_constraints): Likewise.
15808 (CONST_POOL_OK_P): Reject variable-sized modes.
15809 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15810 (add_pseudo_to_slot, lra_spill): Likewise.
15811 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15812 * optabs-query.c (get_best_extraction_insn): Likewise.
15813 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15814 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15815 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15816 * recog.c (offsettable_address_addr_space_p): Likewise.
15817 * regcprop.c (maybe_mode_change): Likewise.
15818 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15819 * regrename.c (build_def_use): Likewise.
15820 * regstat.c (dump_reg_info): Likewise.
15821 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15822 (find_reloads, find_reloads_subreg_address): Likewise.
15823 * reload1.c (eliminate_regs_1): Likewise.
15824 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15825 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15826 (simplify_binary_operation_1, simplify_subreg): Likewise.
15827 * targhooks.c (default_function_arg_padding): Likewise.
15828 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15829 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15830 (verify_gimple_assign_ternary): Likewise.
15831 * tree-inline.c (estimate_move_cost): Likewise.
15832 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15833 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15834 (get_address_cost_ainc): Likewise.
15835 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15836 (vect_supportable_dr_alignment): Likewise.
15837 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15838 (vectorizable_reduction): Likewise.
15839 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15840 (vectorizable_operation, vectorizable_load): Likewise.
15841 * tree.c (build_same_sized_truth_vector_type): Likewise.
15842 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15843 * var-tracking.c (emit_note_insn_var_location): Likewise.
15844 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15845 (ADDR_VEC_ALIGN): Likewise.
15847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15848 Alan Hayward <alan.hayward@arm.com>
15849 David Sherwood <david.sherwood@arm.com>
15851 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15853 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15854 or if measurement_type is polynomial.
15855 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15856 * combine.c (make_extraction): Likewise.
15857 * dse.c (find_shift_sequence): Likewise.
15858 * dwarf2out.c (mem_loc_descriptor): Likewise.
15859 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15860 (extract_bit_field, extract_low_bits): Likewise.
15861 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15862 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15863 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15864 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15865 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15866 * reload.c (find_reloads): Likewise.
15867 * reload1.c (alter_reg): Likewise.
15868 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15869 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15870 * tree-if-conv.c (predicate_mem_writes): Likewise.
15871 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15872 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15873 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15874 * valtrack.c (dead_debug_insert_temp): Likewise.
15875 * varasm.c (mergeable_constant_section): Likewise.
15876 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15879 Alan Hayward <alan.hayward@arm.com>
15880 David Sherwood <david.sherwood@arm.com>
15882 * expr.c (expand_assignment): Cope with polynomial mode sizes
15883 when assigning to a CONCAT.
15885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15886 Alan Hayward <alan.hayward@arm.com>
15887 David Sherwood <david.sherwood@arm.com>
15889 * machmode.h (mode_precision): Change from unsigned short to
15891 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15893 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15894 or if measurement_type is not polynomial.
15895 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15896 in which the mode is already known to be a scalar_int_mode.
15897 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15898 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15900 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15901 for GET_MODE_PRECISION.
15902 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15903 for GET_MODE_PRECISION.
15904 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15906 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15907 (expand_field_assignment, make_extraction): Likewise.
15908 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15909 (get_last_value): Likewise.
15910 * convert.c (convert_to_integer_1): Likewise.
15911 * cse.c (cse_insn): Likewise.
15912 * expr.c (expand_expr_real_1): Likewise.
15913 * lra-constraints.c (simplify_operand_subreg): Likewise.
15914 * optabs-query.c (can_atomic_load_p): Likewise.
15915 * optabs.c (expand_atomic_load): Likewise.
15916 (expand_atomic_store): Likewise.
15917 * ree.c (combine_reaching_defs): Likewise.
15918 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15919 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15920 * tree.h (type_has_mode_precision_p): Likewise.
15921 * ubsan.c (instrument_si_overflow): Likewise.
15923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15924 Alan Hayward <alan.hayward@arm.com>
15925 David Sherwood <david.sherwood@arm.com>
15927 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15928 polynomial numbers of units.
15929 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15930 (valid_vector_subparts_p): New function.
15931 (build_vector_type): Remove temporary shim and take the number
15932 of units as a poly_uint64 rather than an int.
15933 (build_opaque_vector_type): Take the number of units as a
15934 poly_uint64 rather than an int.
15935 * tree.c (build_vector_from_ctor): Handle polynomial
15936 TYPE_VECTOR_SUBPARTS.
15937 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15938 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15939 (build_vector_from_val): If the number of units is variable,
15940 use build_vec_duplicate_cst for constant operands and
15941 VEC_DUPLICATE_EXPR otherwise.
15942 (make_vector_type): Remove temporary is_constant ().
15943 (build_vector_type, build_opaque_vector_type): Take the number of
15944 units as a poly_uint64 rather than an int.
15945 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15947 * cfgexpand.c (expand_debug_expr): Likewise.
15948 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15949 (store_constructor, expand_expr_real_1): Likewise.
15950 (const_scalar_mask_from_tree): Likewise.
15951 * fold-const-call.c (fold_const_reduction): Likewise.
15952 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15953 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15954 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15955 (fold_relational_const): Likewise.
15956 (native_interpret_vector): Likewise. Change the size from an
15957 int to an unsigned int.
15958 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15959 TYPE_VECTOR_SUBPARTS.
15960 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15961 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15962 duplicating a non-constant operand into a variable-length vector.
15963 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15964 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15965 * ipa-icf.c (sem_variable::equals): Likewise.
15966 * match.pd: Likewise.
15967 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15968 * print-tree.c (print_node): Likewise.
15969 * stor-layout.c (layout_type): Likewise.
15970 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15971 * tree-cfg.c (verify_gimple_comparison): Likewise.
15972 (verify_gimple_assign_binary): Likewise.
15973 (verify_gimple_assign_ternary): Likewise.
15974 (verify_gimple_assign_single): Likewise.
15975 * tree-pretty-print.c (dump_generic_node): Likewise.
15976 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15977 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15978 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15979 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15980 (vect_shift_permute_load_chain): Likewise.
15981 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15982 (expand_vector_condition, optimize_vector_constructor): Likewise.
15983 (lower_vec_perm, get_compute_type): Likewise.
15984 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15985 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15986 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15987 (vect_recog_mask_conversion_pattern): Likewise.
15988 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15989 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15990 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15991 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15992 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15993 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15994 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15995 (supportable_widening_operation): Likewise.
15996 (supportable_narrowing_operation): Likewise.
15997 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15999 * varasm.c (output_constant): Likewise.
16001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16002 Alan Hayward <alan.hayward@arm.com>
16003 David Sherwood <david.sherwood@arm.com>
16005 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16006 so that both the length == 3 and length != 3 cases set up their
16007 own permute vectors. Add comments explaining why we know the
16008 number of elements is constant.
16009 (vect_permute_load_chain): Likewise.
16011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16012 Alan Hayward <alan.hayward@arm.com>
16013 David Sherwood <david.sherwood@arm.com>
16015 * machmode.h (mode_nunits): Change from unsigned char to
16017 (ONLY_FIXED_SIZE_MODES): New macro.
16018 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16019 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16020 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16022 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16023 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16024 or if measurement_type is not polynomial.
16025 * genmodes.c (ZERO_COEFFS): New macro.
16026 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16028 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16029 Use ZERO_COEFFS when emitting initializers.
16030 * data-streamer.h (bp_pack_poly_value): New function.
16031 (bp_unpack_poly_value): Likewise.
16032 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16033 for GET_MODE_NUNITS.
16034 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16035 for GET_MODE_NUNITS.
16036 * tree.c (make_vector_type): Remove temporary shim and make
16037 the real function take the number of units as a poly_uint64
16038 rather than an int.
16039 (build_vector_type_for_mode): Handle polynomial nunits.
16040 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16041 * emit-rtl.c (const_vec_series_p_1): Likewise.
16042 (gen_rtx_CONST_VECTOR): Likewise.
16043 * fold-const.c (test_vec_duplicate_folding): Likewise.
16044 * genrecog.c (validate_pattern): Likewise.
16045 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16046 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16047 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16048 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16049 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16050 * rtlanal.c (subreg_get_info): Likewise.
16051 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16052 (vect_grouped_load_supported): Likewise.
16053 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16054 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16055 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16056 (simplify_const_unary_operation, simplify_binary_operation_1)
16057 (simplify_const_binary_operation, simplify_ternary_operation)
16058 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16059 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16060 instead of CONST_VECTOR_NUNITS.
16061 * varasm.c (output_constant_pool_2): Likewise.
16062 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16063 explicit-encoded elements in the XVEC for variable-length vectors.
16065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16067 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16070 Alan Hayward <alan.hayward@arm.com>
16071 David Sherwood <david.sherwood@arm.com>
16073 * coretypes.h (fixed_size_mode): Declare.
16074 (fixed_size_mode_pod): New typedef.
16075 * builtins.h (target_builtins::x_apply_args_mode)
16076 (target_builtins::x_apply_result_mode): Change type to
16077 fixed_size_mode_pod.
16078 * builtins.c (apply_args_size, apply_result_size, result_vector)
16079 (expand_builtin_apply_args_1, expand_builtin_apply)
16080 (expand_builtin_return): Update accordingly.
16082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16084 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16085 * cselib.c (cselib_hash_rtx): Likewise.
16086 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16087 CONST_VECTOR encoding.
16089 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16090 Jeff Law <law@redhat.com>
16093 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16094 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16095 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16096 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16099 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16100 explicitly probe *sp in a noreturn function if there were any callee
16101 register saves or frame pointer is needed.
16103 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16106 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16107 BLKmode for ternary, binary or unary expressions.
16110 * var-tracking.c (delete_vta_debug_insn): New inline function.
16111 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16112 insns from get_insns () to NULL instead of each bb separately.
16113 Use delete_vta_debug_insn. No longer static.
16114 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16115 delete_vta_debug_insns callers.
16116 * rtl.h (delete_vta_debug_insns): Declare.
16117 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16118 instead of variable_tracking_main.
16120 2018-01-03 Martin Sebor <msebor@redhat.com>
16122 PR tree-optimization/83603
16123 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16124 arguments past the endof the argument list in functions declared
16125 without a prototype.
16126 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16127 Avoid checking when arguments are null.
16129 2018-01-03 Martin Sebor <msebor@redhat.com>
16132 * doc/extend.texi (attribute const): Fix a typo.
16133 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16134 issuing -Wsuggest-attribute for void functions.
16136 2018-01-03 Martin Sebor <msebor@redhat.com>
16138 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16139 offset_int::from instead of wide_int::to_shwi.
16140 (maybe_diag_overlap): Remove assertion.
16141 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16142 * gimple-ssa-sprintf.c (format_directive): Same.
16143 (parse_directive): Same.
16144 (sprintf_dom_walker::compute_format_length): Same.
16145 (try_substitute_return_value): Same.
16147 2018-01-03 Jeff Law <law@redhat.com>
16149 PR middle-end/83654
16150 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16151 non-constant residual for zero at runtime and avoid probing in
16152 that case. Reorganize code for trailing problem to mirror handling
16155 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16157 PR tree-optimization/83501
16158 * tree-ssa-strlen.c (get_string_cst): New.
16159 (handle_char_store): Call get_string_cst.
16161 2018-01-03 Martin Liska <mliska@suse.cz>
16163 PR tree-optimization/83593
16164 * tree-ssa-strlen.c: Include tree-cfg.h.
16165 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16166 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16167 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16169 (strlen_dom_walker::before_dom_children): Call
16170 gimple_purge_dead_eh_edges. Dump tranformation with details
16172 (strlen_dom_walker::before_dom_children): Update call by adding
16173 new argument cleanup_eh.
16174 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16176 2018-01-03 Martin Liska <mliska@suse.cz>
16179 * cif-code.def (VARIADIC_THUNK): New enum value.
16180 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16183 2018-01-03 Jan Beulich <jbeulich@suse.com>
16185 * sse.md (mov<mode>_internal): Tighten condition for when to use
16186 vmovdqu<ssescalarsize> for TI and OI modes.
16188 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16190 Update copyright years.
16192 2018-01-03 Martin Liska <mliska@suse.cz>
16195 * ipa-visibility.c (function_and_variable_visibility): Skip
16196 functions with noipa attribure.
16198 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16200 * gcc.c (process_command): Update copyright notice dates.
16201 * gcov-dump.c (print_version): Ditto.
16202 * gcov.c (print_version): Ditto.
16203 * gcov-tool.c (print_version): Ditto.
16204 * gengtype.c (create_file): Ditto.
16205 * doc/cpp.texi: Bump @copying's copyright year.
16206 * doc/cppinternals.texi: Ditto.
16207 * doc/gcc.texi: Ditto.
16208 * doc/gccint.texi: Ditto.
16209 * doc/gcov.texi: Ditto.
16210 * doc/install.texi: Ditto.
16211 * doc/invoke.texi: Ditto.
16213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16215 * vector-builder.h (vector_builder::m_full_nelts): Change from
16216 unsigned int to poly_uint64.
16217 (vector_builder::full_nelts): Update prototype accordingly.
16218 (vector_builder::new_vector): Likewise.
16219 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16220 (vector_builder::operator ==): Likewise.
16221 (vector_builder::finalize): Likewise.
16222 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16223 Take the number of elements as a poly_uint64 rather than an
16225 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16226 from unsigned int to poly_uint64.
16227 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16228 (vec_perm_indices::new_vector): Likewise.
16229 (vec_perm_indices::length): Likewise.
16230 (vec_perm_indices::nelts_per_input): Likewise.
16231 (vec_perm_indices::input_nelts): Likewise.
16232 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16233 number of elements per input as a poly_uint64 rather than an
16234 unsigned int. Use the original encoding for variable-length
16235 vectors, rather than clamping each individual element.
16236 For the second and subsequent elements in each pattern,
16237 clamp the step and base before clamping their sum.
16238 (vec_perm_indices::series_p): Handle polynomial element counts.
16239 (vec_perm_indices::all_in_range_p): Likewise.
16240 (vec_perm_indices_to_tree): Likewise.
16241 (vec_perm_indices_to_rtx): Likewise.
16242 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16243 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16244 (tree_vector_builder::new_binary_operation): Handle polynomial
16245 element counts. Return false if we need to know the number
16246 of elements at compile time.
16247 * fold-const.c (fold_vec_perm): Punt if the number of elements
16248 isn't known at compile time.
16250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16252 * vec-perm-indices.h (vec_perm_builder): Change element type
16253 from HOST_WIDE_INT to poly_int64.
16254 (vec_perm_indices::element_type): Update accordingly.
16255 (vec_perm_indices::clamp): Handle polynomial element_types.
16256 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16257 (vec_perm_indices::all_in_range_p): Likewise.
16258 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16260 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16261 polynomial vec_perm_indices element types.
16262 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16263 * fold-const.c (fold_vec_perm): Likewise.
16264 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16265 * tree-vect-generic.c (lower_vec_perm): Likewise.
16266 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16267 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16268 element type to HOST_WIDE_INT.
16270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16271 Alan Hayward <alan.hayward@arm.com>
16272 David Sherwood <david.sherwood@arm.com>
16274 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16275 rather than an int. Use plus_constant.
16276 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16277 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16280 Alan Hayward <alan.hayward@arm.com>
16281 David Sherwood <david.sherwood@arm.com>
16283 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16284 a HOST_WIDE_INT to a poly_int64.
16286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16287 Alan Hayward <alan.hayward@arm.com>
16288 David Sherwood <david.sherwood@arm.com>
16290 * calls.c (load_register_parameters): Cope with polynomial
16291 mode sizes. Require a constant size for BLKmode parameters
16292 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16293 forces a parameter to be padded at the lsb end in order to
16294 fill a complete number of words, require the parameter size
16295 to be ordered wrt UNITS_PER_WORD.
16297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16298 Alan Hayward <alan.hayward@arm.com>
16299 David Sherwood <david.sherwood@arm.com>
16301 * reload1.c (spill_stack_slot_width): Change element type
16302 from unsigned int to poly_uint64_pod.
16303 (alter_reg): Treat mode sizes as polynomial.
16305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16306 Alan Hayward <alan.hayward@arm.com>
16307 David Sherwood <david.sherwood@arm.com>
16309 * reload.c (complex_word_subreg_p): New function.
16310 (reload_inner_reg_of_subreg, push_reload): Use it.
16312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16313 Alan Hayward <alan.hayward@arm.com>
16314 David Sherwood <david.sherwood@arm.com>
16316 * lra-constraints.c (process_alt_operands): Reject matched
16317 operands whose sizes aren't ordered.
16318 (match_reload): Refer to this check here.
16320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16321 Alan Hayward <alan.hayward@arm.com>
16322 David Sherwood <david.sherwood@arm.com>
16324 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16325 that the mode size is in the set {1, 2, 4, 8, 16}.
16327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16328 Alan Hayward <alan.hayward@arm.com>
16329 David Sherwood <david.sherwood@arm.com>
16331 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16332 Use plus_constant instead of gen_rtx_PLUS.
16334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16335 Alan Hayward <alan.hayward@arm.com>
16336 David Sherwood <david.sherwood@arm.com>
16338 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16339 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16340 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16341 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16342 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16343 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16344 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16345 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16346 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16347 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16349 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16350 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16351 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16352 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16353 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16354 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16355 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16356 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16357 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16358 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16360 * expr.c (emit_move_resolve_push): Treat the input and result
16361 of PUSH_ROUNDING as a poly_int64.
16362 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16363 (emit_push_insn): Likewise.
16364 * lra-eliminations.c (mark_not_eliminable): Likewise.
16365 * recog.c (push_operand): Likewise.
16366 * reload1.c (elimination_effects): Likewise.
16367 * rtlanal.c (nonzero_bits1): Likewise.
16368 * calls.c (store_one_arg): Likewise. Require the padding to be
16369 known at compile time.
16371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16372 Alan Hayward <alan.hayward@arm.com>
16373 David Sherwood <david.sherwood@arm.com>
16375 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16376 Use plus_constant instead of gen_rtx_PLUS.
16378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16379 Alan Hayward <alan.hayward@arm.com>
16380 David Sherwood <david.sherwood@arm.com>
16382 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16383 rather than an int.
16385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16386 Alan Hayward <alan.hayward@arm.com>
16387 David Sherwood <david.sherwood@arm.com>
16389 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16390 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16391 via stack temporaries. Treat the mode size as polynomial too.
16393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16394 Alan Hayward <alan.hayward@arm.com>
16395 David Sherwood <david.sherwood@arm.com>
16397 * expr.c (expand_expr_real_2): When handling conversions involving
16398 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16399 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16400 as a poly_uint64 too.
16402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16403 Alan Hayward <alan.hayward@arm.com>
16404 David Sherwood <david.sherwood@arm.com>
16406 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16409 Alan Hayward <alan.hayward@arm.com>
16410 David Sherwood <david.sherwood@arm.com>
16412 * combine.c (can_change_dest_mode): Handle polynomial
16413 REGMODE_NATURAL_SIZE.
16414 * expmed.c (store_bit_field_1): Likewise.
16415 * expr.c (store_constructor): Likewise.
16416 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16417 and polynomial REGMODE_NATURAL_SIZE.
16418 (gen_lowpart_common): Likewise.
16419 * reginfo.c (record_subregs_of_mode): Likewise.
16420 * rtlanal.c (read_modify_subreg_p): Likewise.
16422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16423 Alan Hayward <alan.hayward@arm.com>
16424 David Sherwood <david.sherwood@arm.com>
16426 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16427 numbers of elements.
16429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16430 Alan Hayward <alan.hayward@arm.com>
16431 David Sherwood <david.sherwood@arm.com>
16433 * match.pd: Cope with polynomial numbers of vector elements.
16435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16436 Alan Hayward <alan.hayward@arm.com>
16437 David Sherwood <david.sherwood@arm.com>
16439 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16440 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
16447 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16448 (ipa_simd_modify_function_body): Likewise.
16450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16451 Alan Hayward <alan.hayward@arm.com>
16452 David Sherwood <david.sherwood@arm.com>
16454 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16455 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16456 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16457 (expand_vector_condition, vector_element): Likewise.
16458 (subparts_gt): New function.
16459 (get_compute_type): Use subparts_gt.
16460 (count_type_subparts): Delete.
16461 (expand_vector_operations_1): Use subparts_gt instead of
16462 count_type_subparts.
16464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16465 Alan Hayward <alan.hayward@arm.com>
16466 David Sherwood <david.sherwood@arm.com>
16468 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16469 (vect_compile_time_alias): ...this new function. Do the calculation
16470 on poly_ints rather than trees.
16471 (vect_prune_runtime_alias_test_list): Update call accordingly.
16473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16474 Alan Hayward <alan.hayward@arm.com>
16475 David Sherwood <david.sherwood@arm.com>
16477 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16479 (vect_schedule_slp_instance): Likewise.
16481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16482 Alan Hayward <alan.hayward@arm.com>
16483 David Sherwood <david.sherwood@arm.com>
16485 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16486 constant and extern definitions for variable-length vectors.
16487 (vect_get_constant_vectors): Note that the number of units
16488 is known to be constant.
16490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16491 Alan Hayward <alan.hayward@arm.com>
16492 David Sherwood <david.sherwood@arm.com>
16494 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16495 of units as polynomial. Choose between WIDE and NARROW based
16498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16499 Alan Hayward <alan.hayward@arm.com>
16500 David Sherwood <david.sherwood@arm.com>
16502 * tree-vect-stmts.c (simd_clone_subparts): New function.
16503 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16506 Alan Hayward <alan.hayward@arm.com>
16507 David Sherwood <david.sherwood@arm.com>
16509 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16510 vectors as polynomial. Use build_index_vector for
16511 IFN_GOMP_SIMD_LANE.
16513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16514 Alan Hayward <alan.hayward@arm.com>
16515 David Sherwood <david.sherwood@arm.com>
16517 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16518 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16519 for variable-length vectors.
16520 (vectorizable_mask_load_store): Treat the number of units as
16521 polynomial, asserting that it is constant if the condition has
16522 already been enforced.
16523 (vectorizable_store, vectorizable_load): Likewise.
16525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16526 Alan Hayward <alan.hayward@arm.com>
16527 David Sherwood <david.sherwood@arm.com>
16529 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16530 of units as polynomial. Punt if we can't tell at compile time
16531 which vector contains the final result.
16533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16534 Alan Hayward <alan.hayward@arm.com>
16535 David Sherwood <david.sherwood@arm.com>
16537 * tree-vect-loop.c (vectorizable_induction): Treat the number
16538 of units as polynomial. Punt on SLP inductions. Use an integer
16539 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16540 cast of such a series for variable-length floating-point
16543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16544 Alan Hayward <alan.hayward@arm.com>
16545 David Sherwood <david.sherwood@arm.com>
16547 * tree.h (build_index_vector): Declare.
16548 * tree.c (build_index_vector): New function.
16549 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16550 of units as polynomial, forcibly converting it to a constant if
16551 vectorizable_reduction has already enforced the condition.
16552 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16553 to create a {1,2,3,...} vector.
16554 (vectorizable_reduction): Treat the number of units as polynomial.
16555 Choose vectype_in based on the largest scalar element size rather
16556 than the smallest number of units. Enforce the restrictions
16559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16560 Alan Hayward <alan.hayward@arm.com>
16561 David Sherwood <david.sherwood@arm.com>
16563 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16564 number of units as polynomial.
16566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16567 Alan Hayward <alan.hayward@arm.com>
16568 David Sherwood <david.sherwood@arm.com>
16570 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16571 * target.def (autovectorize_vector_sizes): Return the vector sizes
16572 by pointer, using vector_sizes rather than a bitmask.
16573 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16574 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16575 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16577 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16578 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16579 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16580 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16581 * omp-general.c (omp_max_vf): Likewise.
16582 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16583 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16584 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16585 * tree-vect-slp.c (vect_slp_bb): Likewise.
16586 * doc/tm.texi: Regenerate.
16587 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16589 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16590 the vector size as a poly_uint64 rather than an unsigned int.
16591 (current_vector_size): Change from an unsigned int to a poly_uint64.
16592 (get_vectype_for_scalar_type): Update accordingly.
16593 * tree.h (build_truth_vector_type): Take the size and number of
16594 units as a poly_uint64 rather than an unsigned int.
16595 (build_vector_type): Add a temporary overload that takes
16596 the number of units as a poly_uint64 rather than an unsigned int.
16597 * tree.c (make_vector_type): Likewise.
16598 (build_truth_vector_type): Take the number of units as a poly_uint64
16599 rather than an unsigned int.
16601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16602 Alan Hayward <alan.hayward@arm.com>
16603 David Sherwood <david.sherwood@arm.com>
16605 * target.def (get_mask_mode): Take the number of units and length
16606 as poly_uint64s rather than unsigned ints.
16607 * targhooks.h (default_get_mask_mode): Update accordingly.
16608 * targhooks.c (default_get_mask_mode): Likewise.
16609 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16610 * doc/tm.texi: Regenerate.
16612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16613 Alan Hayward <alan.hayward@arm.com>
16614 David Sherwood <david.sherwood@arm.com>
16616 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16617 * omp-general.c (omp_max_vf): Likewise.
16618 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16619 (expand_omp_simd): Handle polynomial safelen.
16620 * omp-low.c (omplow_simd_context): Add a default constructor.
16621 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16622 (lower_rec_simd_input_clauses): Update accordingly.
16623 (lower_rec_input_clauses): Likewise.
16625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16626 Alan Hayward <alan.hayward@arm.com>
16627 David Sherwood <david.sherwood@arm.com>
16629 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16630 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16631 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16632 (vect_analyze_slp_cost): Likewise.
16633 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16634 (vect_model_load_cost): Likewise.
16636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16637 Alan Hayward <alan.hayward@arm.com>
16638 David Sherwood <david.sherwood@arm.com>
16640 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16641 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16642 from an unsigned int * to a poly_uint64_pod *.
16643 (calculate_unrolling_factor): New function.
16644 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16647 Alan Hayward <alan.hayward@arm.com>
16648 David Sherwood <david.sherwood@arm.com>
16650 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16651 from an unsigned int to a poly_uint64.
16652 (_loop_vec_info::slp_unrolling_factor): Likewise.
16653 (_loop_vec_info::vectorization_factor): Change from an int
16655 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16656 (vect_get_num_vectors): New function.
16657 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16658 (vect_get_num_copies): Use vect_get_num_vectors.
16659 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16660 to an unsigned int *.
16661 (vect_analyze_data_refs): Change min_vf from an int * to a
16663 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16664 than an unsigned HOST_WIDE_INT.
16665 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16666 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16667 to an unsigned int *.
16668 (vect_analyze_data_ref_dependences): Likewise.
16669 (vect_compute_data_ref_alignment): Handle polynomial vf.
16670 (vect_enhance_data_refs_alignment): Likewise.
16671 (vect_prune_runtime_alias_test_list): Likewise.
16672 (vect_shift_permute_load_chain): Likewise.
16673 (vect_supportable_dr_alignment): Likewise.
16674 (dependence_distance_ge_vf): Take the vectorization factor as a
16675 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16676 (vect_analyze_data_refs): Change min_vf from an int * to a
16678 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16679 vfm1 as a poly_uint64 rather than an int. Make the same change
16680 for the returned bound_scalar.
16681 (vect_gen_vector_loop_niters): Handle polynomial vf.
16682 (vect_do_peeling): Likewise. Update call to
16683 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16684 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16686 * tree-vect-loop.c (vect_determine_vectorization_factor)
16687 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16688 (vect_get_known_peeling_cost): Likewise.
16689 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16690 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16691 (vect_transform_loop): Likewise. Use the lowest possible VF when
16692 updating the upper bounds of the loop.
16693 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16694 rather than an int.
16695 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16696 polynomial unroll factors.
16697 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16698 (vect_make_slp_decision): Likewise.
16699 (vect_supported_load_permutation_p): Likewise, and polynomial
16701 (vect_analyze_slp_cost): Handle polynomial vf.
16702 (vect_slp_analyze_node_operations): Likewise.
16703 (vect_slp_analyze_bb_1): Likewise.
16704 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16705 than an unsigned HOST_WIDE_INT.
16706 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16707 (vectorizable_load): Handle polynomial vf.
16708 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16710 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16713 Alan Hayward <alan.hayward@arm.com>
16714 David Sherwood <david.sherwood@arm.com>
16716 * match.pd: Handle bit operations involving three constants
16717 and try to fold one pair.
16719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16721 * tree-vect-loop-manip.c: Include gimple-fold.h.
16722 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16723 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16724 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16725 Add a path that uses a step of VF instead of 1, but disable it
16727 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16728 and niters_no_overflow parameters. Update calls to
16729 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16730 Create a new SSA name if the latter choses to use a ste other
16731 than zero, and return it via niters_vector_mult_vf_var.
16732 * tree-vect-loop.c (vect_transform_loop): Update calls to
16733 vect_do_peeling, vect_gen_vector_loop_niters and
16734 slpeel_make_loop_iterate_ntimes.
16735 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16736 (vect_gen_vector_loop_niters): Update declarations after above changes.
16738 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16740 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16741 128-bit round to integer instructions.
16742 (ceil<mode>2): Likewise.
16743 (btrunc<mode>2): Likewise.
16744 (round<mode>2): Likewise.
16746 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16748 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16749 unaligned VSX load/store on P8/P9.
16750 (expand_block_clear): Allow the use of unaligned VSX
16751 load/store on P8/P9.
16753 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16755 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16757 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16758 swap associated with both a load and a store.
16760 2018-01-02 Andrew Waterman <andrew@sifive.com>
16762 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16763 * config/riscv/riscv.md (clear_cache): Use it.
16765 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16767 * web.c: Remove out-of-date comment.
16769 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16771 * expr.c (fixup_args_size_notes): Check that any existing
16772 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16773 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16774 (emit_single_push_insn): ...here.
16776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16778 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16779 (const_vector_encoded_nelts): New function.
16780 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16781 (const_vector_int_elt, const_vector_elt): Declare.
16782 * emit-rtl.c (const_vector_int_elt_1): New function.
16783 (const_vector_elt): Likewise.
16784 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16785 of CONST_VECTOR_ELT.
16787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16789 * expr.c: Include rtx-vector-builder.h.
16790 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16791 directly on the tree encoding.
16792 (const_vector_from_tree): Likewise.
16793 * optabs.c: Include rtx-vector-builder.h.
16794 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16795 sequence of "u" values.
16796 * vec-perm-indices.c: Include rtx-vector-builder.h.
16797 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16798 directly on the vec_perm_indices encoding.
16800 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16802 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16803 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16804 * rtx-vector-builder.h: New file.
16805 * rtx-vector-builder.c: Likewise.
16806 * rtl.h (rtx_def::u2): Add a const_vector field.
16807 (CONST_VECTOR_NPATTERNS): New macro.
16808 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16809 (CONST_VECTOR_DUPLICATE_P): Likewise.
16810 (CONST_VECTOR_STEPPED_P): Likewise.
16811 (CONST_VECTOR_ENCODED_ELT): Likewise.
16812 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16813 (unwrap_const_vec_duplicate): Likewise.
16814 (const_vec_series_p): Check for a non-duplicated vector encoding.
16815 Say that the function only returns true for integer vectors.
16816 * emit-rtl.c: Include rtx-vector-builder.h.
16817 (gen_const_vec_duplicate_1): Delete.
16818 (gen_const_vector): Call gen_const_vec_duplicate instead of
16819 gen_const_vec_duplicate_1.
16820 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16821 (gen_const_vec_duplicate): Use rtx_vector_builder.
16822 (gen_const_vec_series): Likewise.
16823 (gen_rtx_CONST_VECTOR): Likewise.
16824 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16825 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16826 Build a new vector rather than modifying a CONST_VECTOR in-place.
16827 (handle_special_swappables): Update call accordingly.
16828 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16829 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16830 Build a new vector rather than modifying a CONST_VECTOR in-place.
16831 (handle_special_swappables): Update call accordingly.
16833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16835 * simplify-rtx.c (simplify_const_binary_operation): Use
16836 CONST_VECTOR_ELT instead of XVECEXP.
16838 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16840 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16841 the selector elements to be different from the data elements
16842 if the selector is a VECTOR_CST.
16843 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16844 ssizetype for the selector.
16846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16848 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16849 before testing each element individually.
16850 * tree-vect-generic.c (lower_vec_perm): Likewise.
16852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16854 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16855 * selftest-run-tests.c (selftest::run_tests): Call it.
16856 * vector-builder.h (vector_builder::operator ==): New function.
16857 (vector_builder::operator !=): Likewise.
16858 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16859 (vec_perm_indices::all_from_input_p): New function.
16860 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16861 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16862 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16863 instead of reading the VECTOR_CST directly. Detect whether both
16864 vector inputs are the same before constructing the vec_perm_indices,
16865 and update the number of inputs argument accordingly. Use the
16866 utility functions added above. Only construct sel2 if we need to.
16868 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16870 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16871 the broadcast of the low byte.
16872 (expand_mult_highpart): Use an explicit encoding for the permutes.
16873 * optabs-query.c (can_mult_highpart_p): Likewise.
16874 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16875 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16876 (vectorizable_bswap): Likewise.
16877 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16878 explicit encoding for the power-of-2 permutes.
16879 (vect_permute_store_chain): Likewise.
16880 (vect_grouped_load_supported): Likewise.
16881 (vect_permute_load_chain): Likewise.
16883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16885 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16886 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16887 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16888 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16889 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16890 (vect_gen_perm_mask_any): Likewise.
16892 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16894 * int-vector-builder.h: New file.
16895 * vec-perm-indices.h: Include int-vector-builder.h.
16896 (vec_perm_indices): Redefine as an int_vector_builder.
16897 (auto_vec_perm_indices): Delete.
16898 (vec_perm_builder): Redefine as a stand-alone class.
16899 (vec_perm_indices::vec_perm_indices): New function.
16900 (vec_perm_indices::clamp): Likewise.
16901 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16902 (vec_perm_indices::new_vector): New function.
16903 (vec_perm_indices::new_expanded_vector): Update for new
16904 vec_perm_indices class.
16905 (vec_perm_indices::rotate_inputs): New function.
16906 (vec_perm_indices::all_in_range_p): Operate directly on the
16907 encoded form, without computing elided elements.
16908 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16909 encoding. Update for new vec_perm_indices class.
16910 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16911 the given vec_perm_builder.
16912 (expand_vec_perm_var): Update vec_perm_builder constructor.
16913 (expand_mult_highpart): Use vec_perm_builder instead of
16914 auto_vec_perm_indices.
16915 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16916 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16917 or double series encoding as appropriate.
16918 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16919 vec_perm_indices instead of auto_vec_perm_indices.
16920 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16921 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16922 (vect_permute_store_chain): Likewise.
16923 (vect_grouped_load_supported): Likewise.
16924 (vect_permute_load_chain): Likewise.
16925 (vect_shift_permute_load_chain): Likewise.
16926 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16927 (vect_transform_slp_perm_load): Likewise.
16928 (vect_schedule_slp_instance): Likewise.
16929 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16930 (vectorizable_mask_load_store): Likewise.
16931 (vectorizable_bswap): Likewise.
16932 (vectorizable_store): Likewise.
16933 (vectorizable_load): Likewise.
16934 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16935 vec_perm_indices instead of auto_vec_perm_indices. Use
16936 tree_to_vec_perm_builder to read the vector from a tree.
16937 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16938 vec_perm_builder instead of a vec_perm_indices.
16939 (have_whole_vector_shift): Use vec_perm_builder and
16940 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16941 truncation to calc_vec_perm_mask_for_shift.
16942 (vect_create_epilog_for_reduction): Likewise.
16943 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16944 from auto_vec_perm_indices to vec_perm_indices.
16945 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16946 instead of changing individual elements.
16947 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16948 the vector in d.perm.
16949 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16950 from auto_vec_perm_indices to vec_perm_indices.
16951 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16952 instead of changing individual elements.
16953 (arm_vectorize_vec_perm_const): Use new_vector to install
16954 the vector in d.perm.
16955 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16956 Update vec_perm_builder constructor.
16957 (rs6000_expand_interleave): Likewise.
16958 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16959 (rs6000_expand_interleave): Likewise.
16961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16963 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16964 to qimode could truncate the indices.
16965 * optabs.c (expand_vec_perm_var): Likewise.
16967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16969 * Makefile.in (OBJS): Add vec-perm-indices.o.
16970 * vec-perm-indices.h: New file.
16971 * vec-perm-indices.c: Likewise.
16972 * target.h (vec_perm_indices): Replace with a forward class
16974 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16975 * optabs.h: Include vec-perm-indices.h.
16976 (expand_vec_perm): Delete.
16977 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16978 (expand_vec_perm_const): Declare.
16979 * target.def (vec_perm_const_ok): Replace with...
16980 (vec_perm_const): ...this new hook.
16981 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16982 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16983 * doc/tm.texi: Regenerate.
16984 * optabs.def (vec_perm_const): Delete.
16985 * doc/md.texi (vec_perm_const): Likewise.
16986 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16987 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16988 expand_vec_perm for constant permutation vectors. Assert that
16989 the mode of variable permutation vectors is the integer equivalent
16990 of the mode that is being permuted.
16991 * optabs-query.h (selector_fits_mode_p): Declare.
16992 * optabs-query.c: Include vec-perm-indices.h.
16993 (selector_fits_mode_p): New function.
16994 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16995 is defined, instead of checking whether the vec_perm_const_optab
16996 exists. Use targetm.vectorize.vec_perm_const instead of
16997 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16998 fit in the vector mode before using a variable permute.
16999 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17000 vec_perm_indices instead of an rtx.
17001 (expand_vec_perm): Replace with...
17002 (expand_vec_perm_const): ...this new function. Take the selector
17003 as a vec_perm_indices rather than an rtx. Also take the mode of
17004 the selector. Update call to shift_amt_for_vec_perm_mask.
17005 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17006 Use vec_perm_indices::new_expanded_vector to expand the original
17007 selector into bytes. Check whether the indices fit in the vector
17008 mode before using a variable permute.
17009 (expand_vec_perm_var): Make global.
17010 (expand_mult_highpart): Use expand_vec_perm_const.
17011 * fold-const.c: Includes vec-perm-indices.h.
17012 * tree-ssa-forwprop.c: Likewise.
17013 * tree-vect-data-refs.c: Likewise.
17014 * tree-vect-generic.c: Likewise.
17015 * tree-vect-loop.c: Likewise.
17016 * tree-vect-slp.c: Likewise.
17017 * tree-vect-stmts.c: Likewise.
17018 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17020 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17021 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17022 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17023 (aarch64_vectorize_vec_perm_const): ...this new function.
17024 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17025 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17026 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17027 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17028 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17029 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17030 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17032 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17033 check for NEON modes.
17034 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17035 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17036 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17037 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17039 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17040 the old VEC_PERM_CONST conditions.
17041 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17042 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17043 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17044 (ia64_vectorize_vec_perm_const_ok): Merge into...
17045 (ia64_vectorize_vec_perm_const): ...this new function.
17046 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17047 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17048 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17049 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17050 * config/mips/mips.c (mips_expand_vec_perm_const)
17051 (mips_vectorize_vec_perm_const_ok): Merge into...
17052 (mips_vectorize_vec_perm_const): ...this new function.
17053 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17054 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17055 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17056 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17057 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17058 (rs6000_expand_vec_perm_const): Delete.
17059 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17061 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17062 (altivec_expand_vec_perm_const_le): Take each operand individually.
17063 Operate on constant selectors rather than rtxes.
17064 (altivec_expand_vec_perm_const): Likewise. Update call to
17065 altivec_expand_vec_perm_const_le.
17066 (rs6000_expand_vec_perm_const): Delete.
17067 (rs6000_vectorize_vec_perm_const_ok): Delete.
17068 (rs6000_vectorize_vec_perm_const): New function.
17069 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17070 an element count and rtx array.
17071 (rs6000_expand_extract_even): Update call accordingly.
17072 (rs6000_expand_interleave): Likewise.
17073 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17074 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17075 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17076 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17077 (rs6000_expand_vec_perm_const): Delete.
17078 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17079 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17080 (altivec_expand_vec_perm_const_le): Take each operand individually.
17081 Operate on constant selectors rather than rtxes.
17082 (altivec_expand_vec_perm_const): Likewise. Update call to
17083 altivec_expand_vec_perm_const_le.
17084 (rs6000_expand_vec_perm_const): Delete.
17085 (rs6000_vectorize_vec_perm_const_ok): Delete.
17086 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17087 reference to the SPE evmerge intructions.
17088 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17089 an element count and rtx array.
17090 (rs6000_expand_extract_even): Update call accordingly.
17091 (rs6000_expand_interleave): Likewise.
17092 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17093 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17095 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17099 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17100 vector mode and that that mode matches the mode of the data
17102 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17103 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17104 directly using expand_vec_perm_1 when forcing selectors into
17106 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17110 * optabs-query.h (can_vec_perm_p): Delete.
17111 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17112 * optabs-query.c (can_vec_perm_p): Split into...
17113 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17114 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17115 particular selector is valid.
17116 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17117 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17118 (vect_grouped_load_supported): Likewise.
17119 (vect_shift_permute_load_chain): Likewise.
17120 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17121 (vect_transform_slp_perm_load): Likewise.
17122 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17123 (vectorizable_bswap): Likewise.
17124 (vect_gen_perm_mask_checked): Likewise.
17125 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17126 implementations of variable permutation vectors into account
17127 when deciding which selector to use.
17128 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17129 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17130 with a false third argument.
17131 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17132 to test whether the constant selector is valid and can_vec_perm_var_p
17133 to test whether a variable selector is valid.
17135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17137 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17138 * optabs-query.c (can_vec_perm_p): Likewise.
17139 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17140 instead of vec_perm_indices.
17141 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17142 (vect_gen_perm_mask_checked): Likewise,
17143 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17144 (vect_gen_perm_mask_checked): Likewise,
17146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17148 * optabs-query.h (qimode_for_vec_perm): Declare.
17149 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17150 (qimode_for_vec_perm): ...this new function.
17151 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17153 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17155 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17156 does not have a conditional at the top.
17158 2018-01-02 Richard Biener <rguenther@suse.de>
17160 * ipa-inline.c (big_speedup_p): Fix expression.
17162 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17165 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17168 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17172 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17173 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17174 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17175 cond_taken_branch_cost 3->4.
17177 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17179 PR tree-optimization/83581
17180 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17181 TODO_cleanup_cfg if any changes have been made.
17183 PR middle-end/83608
17184 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17185 convert_modes if target mode has the right side, but different mode
17188 PR middle-end/83609
17189 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17190 last argument when extracting from CONCAT. If either from_real or
17191 from_imag is NULL, use expansion through memory. If result is not
17192 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17193 the parts directly to inner mode, if even that fails, use expansion
17196 PR middle-end/83623
17197 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17198 check for bswap in mode rather than HImode and use that in expand_unop
17201 Copyright (C) 2018 Free Software Foundation, Inc.
17203 Copying and distribution of this file, with or without modification,
17204 are permitted in any medium without royalty provided the copyright
17205 notice and this notice are preserved.